SEQH Capital Research · Trading & Research Desk
Edition 001 · Thursday, September 24, 2026 · Post-close
Prices as of 4:00 PM ET close unless stamped otherwise.

The Tape in 60 Seconds

Bonds set the terms; the physical uranium market kept tightening anyway.

The 10-year Treasury held near 5.10%, its highest level since 2007, and the 30-year printed 5.43%, its highest since 2004. Fed speakers kept an October hike live, and CME FedWatch put the odds at 69%. A Reuters report that U.S. and Iranian negotiators are working on a phased deal to reopen Hormuz pulled indexes off their lows. The S&P 500 closed −0.02%, the Nasdaq +0.01% and the Dow −0.31%.

For our coverage, it was day two of a rate-driven selloff. On Wednesday an equal-weight six-name nuclear/uranium basket fell 3.78%, against −0.75% for the S&P, a 303 bp underperformance. Underneath, UxC’s long-term U3O8 price sits at a record $96/lb, while SPUT trades 11% below its net asset value.

Three takeaways:

1. Rates, not fundamentals, drove the complex. Pre-revenue SMR developers and high-multiple optics are the longest-duration assets in our universe, and they traded that way.

2. The gap between physical uranium and the equities is widening. The term price is at an all-time high, while CCJ is 32.9% below its 52-week high and SPUT sits at a −11.04% discount.

3. The AI-power read-through turned negative intraday. Oracle sent a force-majeure notice on its 2.45 GW Project Jupiter campus in New Mexico.

Key numbers

• U3O8 spot (SPUT-implied): $89.82/lb · 9/23 close

• U3O8 long-term (UxC): $96.00/lb · record high · 8/31 print

• SPUT premium/(discount): −11.04% · 9/23, which works out to ≈$80.84/lb at market price

• US 10-year: ~5.10% · highest since July 2007

• WTI crude (Nov): $95.32 · ▲ +3.16 (+3.43%)

• October FOMC hike odds: 69% · CME FedWatch, 9/24 AM

01 · Macro Tape

• S&P 500: 7,704.13 · ▼ 1.90 (−0.02%)

• Dow Jones: 51,349.98 · ▼ 161.61 (−0.31%)

• Nasdaq Composite: 26,939.37 · ▲ 3.34 (+0.01%)

• Russell 2000: 2,835.58 · ▼ 3.08 (−0.11%)

• VIX: 15.64 · ▲ 0.46 (+3.03%)

• WTI crude, Nov-26: $95.32 · ▲ 3.16 (+3.43%)

• Gold (front): $4,306.70 · ▼ 11.70 (−0.27%)

• US 10-year yield: ~5.10% · intraday ~5.14% · 19-year high

• US 30-year yield: 5.43% · highest since 2004

Data and Fed:

• Initial jobless claims were 197K vs. 201K expected (−1K w/w), and continuing claims were 1.72M.

• The 30-year mortgage rate hit 7.37% (+11 bp, Mortgage News Daily), its highest since May 2024.

• Brent traded above $105 intraday.

• Philadelphia Fed President Paulson said modest further tightening may be warranted, and NY Fed President Williams called another hike by year-end reasonable.

• Treasury Secretary Bessent said the U.S.–China truce will be extended to Jan 10, with critical minerals on the summit agenda.

02 · Uranium & Fuel Cycle

Physical market: Sprott Physical Uranium Trust (U.U / SRUUF), 9/23 close

• NAV per unit: $21.57 (−$0.03, −0.13%)

• Market price: $19.19 (SRUUF −0.96%)

• U3O8 held: 81,697,348 lb

• Total NAV: $7.42B · uranium at market $7.338B

• Pounds per unit: 0.2374 · 344.16M units

• Implied $/lb at market price: $80.84, excluding cash

• NAV YTD: +8.97%

• Q2-26 sessions at a discount: 57 of 63 (90%)

The curve. The long-term price of $96.00 sits $6.18/lb (6.9%) above the SPUT-implied spot of $89.82. Spot remains 16.1% below its $107 cycle peak and 34.0% below the 2007 high of $136. In August spot rose 3.63%, while miners gained 17.29% and juniors 19.20%, roughly 4.8x torque to the commodity.

Contracting math (Sprott Uranium Report, published 9/24)

• Utility contract coverage is falling fast. U.S. coverage drops from 60% in 2030 to 9% in 2033, and European coverage from 100% in 2030 to 36% in 2034.

• The NNSA is seeking new supply. Its RFI asks for about 4 Mlb/yr of U.S.-origin U3O8 from around 2030, which is roughly 100% of current U.S. mine output. U.S. reactors need about 49 Mlb/yr.

• Supply has lagged price. Since 2020 the price has roughly tripled, while mine output rose only about 60 Mlb (~50%). The next supply increases have to come from higher-cost greenfield mines.

Equities (last verified print)

• CCJ: $90.80 · ▼ 4.01% · 9/23 close. Volume was 2.43M vs. a 3.21M average (0.76x). 52-week range is 77.70–135.24, and the consensus price target of $127.76 implies +40.7%.

• LEU: $151.31 · ▼ 4.13% · 9/23 close. Centrus is a HALEU enricher. An Oklo letter of intent covers fuel for up to five Aurora units, with deliveries from 2029.

• UUUU: $11.60 · ▼ 4.45% · 9/23 close. Biggest decliner among the producers on the day.

• UEC: $9.49 · ▼ 4.6% vs. the 9/23 close of $9.95 · 9/24 intraday. The range was 9.33–9.91 on 12.8M shares vs. a 9.04M average (1.42x). That leaves it 6.5% above its 52-week low and 53.3% below its January all-time high.

Desk read: Cameco’s 4% drop on Wednesday came on only 0.76x average volume, which looks like a de-rating on light participation rather than capitulation. UEC is the exception: turnover ran 1.42x average heading into its Sept 29 fiscal-year results, where options were pricing a move near 9% (TipRanks, 9/17).

03 · Reactors, SMRs & Power

• OKLO: $38.82 · ▼ 4.03% · 9/23 close. The stock is 30.2% below the ~$55.64 average price on its prior $1B at-the-market share sale (17.97M shares). A new $1B share-sale program has been live since 9/11.

• SMR: $8.68 · ▼ 2.36% · 9/23 close. NuScale’s Q2 revenue was $75K, and it has a $750M share-sale program on file. The stock gave back all of last week’s gains from the Ratepayer Protection Act rally.

XE · TRISO-X completes vertical construction at TX-1 (Oak Ridge)

• The 214,000 sq ft structure is complete, and work now shifts to interior buildout and fuel-fabrication equipment.

• Nameplate output is about 700,000 fuel pebbles a year (≈5 MTU), enough for up to 11 Xe-100 reactors.

• It holds the first NRC Category 2 license for HALEU processing, with a 40-year term.

• It would be the first new commercial-scale advanced fuel plant in the U.S. in more than 50 years.

• XE shares remain more than 55% below their IPO high.

CCJ · Westinghouse IPO math
Bloomberg reports a target valuation above $50B, with a public filing possible as soon as October. At $50B, Cameco’s 49% stake is worth about $24.5B, or roughly 62% of CCJ’s $39.5B market cap at Wednesday’s close. Cameco’s Q2 MD&A cites a Westinghouse pipeline of up to 91 potential AP1000 reactors (~105 GWe).

BWXT · Investment-grade rating ahead of Investor Day
Fitch assigned BWXT a BBB investment-grade rating on 9/23. Investor Day follows on Tuesday, 9/29, the next scheduled update on long-range targets.

AI power · Oracle invokes force majeure on Project Jupiter

• Oracle told the Blue Owl-affiliated developer it may delay payments if the New Mexico campus misses its 2028 start date.

• The site is designed around up to 2.45 GW of on-site fuel-cell power, and about $18B of related debt is quoted at 89–91 cents on the dollar.

• ORCL fell about 4%, and Bloom Energy fell 4.4% premarket.

Our read: problems securing site power and permits are now showing up as contract risk. That supports the long-run case for firm, licensed generation like nuclear, but it weighs on near-term AI-spending sentiment, which is what our SMR and optics names trade on.

04 · Photonics & Optical Interconnect

These prints are from 3:18 PM ET on 9/23, not the close.

• LITE: $934.01 · ▼ 1.23%. Up 153.3% YTD and 14.0% below its 52-week high of 1,085.68. The consensus price target of $1,149 implies +23.1%. Rosenblatt reiterated Buy with a $1,300 target on 9/23. Market cap is $83.8B.

• COHR: $299.94 · ▼ 3.37%. Last quarter’s revenue was $2.05B (+34% y/y), with datacenter and communications at 79% of the mix.

• AAOI: $101.70 · ▼ 4.95%. The most volatile name in the group, driven by its 800G/1.6T transceiver ramp.

• CRDO: $195.10 · ▲ 1.32%. The only gainer in the set, showing relative strength vs. optical components.

An equal-weight basket of these four names was down 2.06% late Wednesday.

Optics traded as the most volatile end of the AI-spending trade this week. Thursday added two negatives:

• Yields at 19-year highs pressured valuations on the most expensive growth names.

• Oracle’s force-majeure notice raised questions about timelines for large AI data center campuses.

Meta’s +4% on its Muse monetization plan partly offset those. Reported fundamentals are intact: Lumentum’s fiscal Q4 revenue was $1.006B (+109% y/y), and Ciena carries a backlog of about $10B. On the supply side, Tower Semiconductor and NewPhotonics began high-volume shipments of laser-integrated optical engines for 800G–1.6T modules on 9/17, with 6.4T near-packaged optics on the roadmap.

05 · Catalyst Calendar

• Tue 9/29: Uranium Energy FY2026 results, pre-market, with a call at 11:00 ET (UEC)

• Tue 9/29: BWX Technologies Investor Day (BWXT)

• Oct (TBD): Westinghouse public IPO filing possible, per Bloomberg (CCJ, BEP)

• Oct FOMC: Rate decision, with 69% hike odds priced as of 9/24

• Fri 10/30: Cameco Q3 results (CCJ)

• Sat 10/31: Fermi/TensorWave extended closing date. Fermi picked CBRE to run its first data center on 9/24 (FRMI)

• ~Thu 11/5: Lumentum fiscal Q1 FY27 results, estimated date (LITE)

• Sun 1/10/27: U.S.–China truce extension expires, with critical minerals on the agenda

06 · Desk Read: What We’re Watching

i. The gap between the term price and the equities. The long-term price is at a record $96, while CCJ is 32.9% off its high and SPUT trades at an 11% discount. The physical market is pricing a scarcity the equity market is not. A narrowing SPUT discount would be the first sign that equity money is returning.

ii. Rates are the swing factor. With the 10-year at 5.10% and an October hike about 69% priced, Fed speakers and Treasury auctions are likely to matter more than company news for pre-revenue developers (OKLO, SMR, NNE) and high-multiple optics.

iii. Execution over announcements. The milestones that turn policy support into actual fuel are TX-1’s construction, Centrus’s HALEU deliveries and UEC’s Burke Hollow ramp (reporting 9/29).

Sources, stamps & method

• Indexes, commodities, VIX: Yahoo Finance close data and live coverage, 9/24 (4:11 PM ET).

• Yields, claims, mortgage rates, Fed commentary, FedWatch odds: Yahoo Finance, 9/24.

• Physical uranium: Sprott Physical Uranium Trust daily data (9/23 close, posted 6:00 PM ET).

• Term price, contracting, NNSA and supply data: Sprott Uranium Report, 9/24 (UxC and Bloomberg data as of 8/31).

• Equity prints: Yahoo Finance, 9/23 close (LITE, COHR, AAOI and CRDO at 3:18 PM ET 9/23). UEC 9/24 intraday via Robinhood.

• Company news: GlobeNewswire (TRISO-X, 9/24); UEC release (9/22); Bloomberg, Reuters and CNBC (Oracle, 9/24); Bloomberg via The Motley Fool (Westinghouse); SEC filings (Oklo 10-Q; Tower 6-K).

Method:

• SPUT-implied spot = market value of U3O8 held ÷ pounds held.

• $/lb at market = unit price ÷ pounds per unit, excluding cash.

• Baskets are equal-weighted simple averages.

• Where a Thursday close was not verified at publication, the line shows the last verified print with its stamp.

For informational and educational purposes only. Nothing here is investment advice or a recommendation to buy or sell any security. Figures come from third-party sources believed to be reliable but not guaranteed; verify before acting. Past performance does not indicate future results.

© 2026 SEQH Capital Research · Next edition: Friday 9/25, post-close

WHAT THIS REPORT ARGUES

• NANO Nuclear is assembling an advanced-nuclear stack across reactor design, fuel, logistics, component engineering, and power offtake. The strategy is more coherent than the announcement flow suggests, but most links remain non-binding, undisclosed, or unbuilt.

• The key question is simple: can KRONOS move from a $300–350 million first-of-a-kind reactor to a serial product that can compete in real-world firm-power markets?

The commercial gap

• KRONOS is designed as a 15 MWe high-temperature gas-cooled microreactor. At the disclosed FOAK midpoint of $325 million, the implied capital intensity is about $21,667/kW.

• On the report’s assumptions, that requires roughly $282/MWh to recover capital, versus an indicative competitive firm-power range closer to $80–120/MWh.

• The serial-cost problem is therefore the thesis: installed cost needs to fall toward roughly $80–110 million per unit. NNE has not yet disclosed an nth-of-a-kind cost target, fuel cost, operating-cost estimate, capacity factor, or design-life assumption.

What NNE owns

• NNE owns the reactor-design and licensing path, plus the STS nuclear transport business acquired in May 2026.

• It has binding engineering relationships with Howden and Fortil, but fuel supply, conversion, deconversion, data-center integration, and major power offtake remain mostly memoranda or non-binding frameworks.

• The largest missing link is construction: no EPC partner, fixed-price construction arrangement, or nuclear construction-management structure has been publicly disclosed.

Balance sheet and runway

• NNE held approximately $580 million of cash and short-term Treasury investments at June 30, 2026, with minimal liabilities and no meaningful debt.

• That is a real strength, but operating costs are rising. Quarterly cash operating expense was about $13.6 million, or roughly $54 million annualized, before FOAK construction starts.

• In SEQH’s base case, liquidity turns negative in FY2030, around the company’s targeted first-operation date. A capital-light case works only if NNE funds less than two-thirds of FOAK cost and secures substantial non-dilutive support.

The commercial framework

• The Tillman framework targets up to 2 GW of future capacity, equivalent to about 133 KRONOS units, but is explicitly non-binding.

• NNE has agreed in principle to up to $100 million of milestone-based warrants and a $5 million stock grant, illustrating the value attached to securing committed demand.

• The eventual model likely requires outside project owners to finance and own power plants, leaving NNE to earn equipment, fuel, engineering, and services margins rather than utility-style power revenue.

Bottom line

• NNE has a strong balance sheet, a genuine licensing pathway, a scarce logistics asset, and a strategically sensible effort to control nuclear bottlenecks.

• But the full-circle story remains ahead of execution: only two major links are owned, commercial agreements are mostly not binding, and the central cost-down requirement is still undisclosed.

• The clearest milestones are a disclosed nth-of-a-kind cost target, a binding reactor order, an EPC partner, successful NRC permitting, non-dilutive FOAK funding, and a full quarter of STS performance near the acquisition-period run rate.

What readers get in the full PDF
Upgrade to access the complete project-level underwrite, including:

• The full supply-chain map showing which links are owned, contracted, exploratory, or missing.

• The FOAK capital-cost and required-power-price sensitivity model.

• Base, stress, and capital-light liquidity scenarios through FY2031.

• Deep dives on the Tillman framework, STS acquisition, fuel strategy, construction risk, and commercial conversion timeline.

• The full risk register, source register, and list of critical figures NNE has not yet disclosed.
  
  
  21-PAGE PDF ATTACHED BELOW:
  

Read more

WHAT THIS REPORT ARGUES

• This report takes a more skeptical, scenario-driven look at Sivers after its sharp rerating and subsequent drawdown. The conclusion is not that Sivers lacks real technology, real customers, or real exposure to AI optics, but that the current valuation still requires a level of revenue conversion, margin recovery, and execution that the reported financials do not yet support.

• At the report’s reference price of SEK 27.38, Sivers traded at roughly 31.6x trailing revenue and 73.4x trailing product revenue, despite declining product revenue, a shrinking Photonics segment, negative gross margin, and a large gap between pipeline value and disclosed firm orders.

Core thesis

• Sivers has genuine strategic assets: an InP laser platform, exposure to AI optical interconnect, SATCOM, LiDAR, and pluggable optics, a repaired balance sheet, and potential production ramps through customers such as ALL.SPACE, Jabil, SemiNex, Tachyon, and others.

• The bear case is not that the technology is fake or that the market opportunity does not exist. It is that the market has already priced in a large share of a successful multi-year conversion before that conversion has appeared in the reported revenue line.

• The report argues that the stock’s asymmetry is unfavorable at SEK 27 because the bullish operating outcome produces limited upside from the starting valuation, while further ramp delays, dilution, or margin weakness leave a very large downside range.

Product revenue: the headline problem

• Management highlighted Q2 product revenue growth of 18 percent in constant currency and 13.6 percent reported, but the report argues that this is a quarter-specific framing rather than the complete picture.

• Across H1 2026, product revenue fell from SEK 66.9 million to SEK 56.8 million, a 15.1 percent year-over-year decline, following a 33 percent decline in Q1.

• Product revenue increased as a share of total sales from 46.4 percent to 49.1 percent, but the report calls this an arithmetic rather than commercial mix shift: NRE revenue fell faster, down 23.7 percent, while product revenue also declined.

• The key proof point is therefore not product mix alone, but whether product revenue begins to grow in absolute SEK terms while NRE declines, which management has positioned as a Q4 2026 event.

The AI segment is shrinking

• Photonics, the segment carrying the AI-optics, InP laser, and co-packaged-optics narrative, generated SEK 33.7 million in H1 2026, down 23.2 percent year over year.

• Q2 Photonics revenue was SEK 15.9 million, down 7.6 percent, while the larger Wireless segment fell 18.5 percent in the first half.

• The report is not dismissing the possibility that current photonics revenue reflects sampling, qualification, and pre-production activity ahead of 2027 ramps. But it argues that this explanation cannot remain permanent: design wins ultimately need to show up as production revenue.

• Gross margin is presented as the biggest current financial gap. Group gross profit was negative SEK 19.7 millionin Q2, implying roughly negative 36.6 percent gross margin, and H1 gross profit was negative SEK 5.2 million versus positive gross profit a year earlier.

Pipeline versus backlog

• Sivers’ opportunity pipeline reached $1.2 billion, up 268 percent from December 2025, but management defines it as non-binding estimated revenue potential across 2026 to 2030 rather than as contracted backlog.

• The report identifies approximately $22.7 million of disclosed named orders and programmes, equal to only about 1.9 percent of the stated pipeline. These include ALL.SPACE, Microelectronics Commons, SemiNex, and Tachyon-related opportunities.

• The gap is not treated as proof that the pipeline is invalid. It is treated as the core underwriting issue: investors must decide what conversion rate, timing, and customer concentration are realistic before treating the pipeline as value.

• For the current enterprise value to produce a zero return over three years at a 6x FY2028 sales multiple, the report estimates Sivers would need to generate roughly SEK 1.44 billion of FY2028 revenue, about 4.8x FY2025 revenue and a 68 percent three-year CAGR.

What must go right

• Jabil needs to progress from beta builds in late 2026 through qualification and into production orders in the first half of 2027.

• The strategic LiDAR relationship needs to begin volume production around Q4 2026 and perform toward the higher end of the disclosed lifecycle potential.

• ALL.SPACE needs to convert its $8.2 million production order into recognized 2027 revenue and into a repeatable multi-year SATCOM program rather than a one-time award.

• Sivers also needs to convert pluggable CW laser samples into sockets and production share in a market where larger incumbents already have more installed manufacturing capacity.

• Each outcome is plausible individually. The report’s point is that the probability of all of them occurring close to schedule is lower than the narrative can make it appear.

Dilution and capital structure

• Registered shares rose from approximately 285.7 million to 356.7 million between June 2025 and August 2026, an increase of 24.9 percent.

• The report focuses on the Bootstrap convertible, which converted in July 2026 at a strike of SEK 4.77, issuing roughly 22.85 million shares in exchange for approximately SEK 109 million of principal.

• At the prevailing share price around conversion, the report estimates those shares were worth about SEK 1.29 billion, illustrating how expensive the realized cost of emergency-stage financing became after the share-price rally.

• The company is now debt-free after repaying its residual term loan, but the report argues that the historical funding structure matters because it demonstrates how dilution can become the cost of bridging an execution gap.

Cash runway and Glasgow

• Sivers’ post-raise liquidity is one of the strongest elements of the bull case. The report estimates pro forma cash of about SEK 704 million and net cash of roughly SEK 685 million, with no interest-bearing debt.

• The concern is timing. The announced $30 million Glasgow expansion adds approximately SEK 57 million of capex per quarter through Q4 2027, on top of recent quarterly operating cash burn of about SEK 70 million.

• Under the report’s modeled burn paths, cash is exhausted between Q1 2028 and Q3 2028, placing the next potential financing decision near the moment when the expected 2027 transformation has to prove itself.

• If the ramp works, the company can raise from strength. If the ramp slips, investors may face another financing event at a much lower share price and with a larger share count.

Accounting and governance overhang

• The report highlights a series of 2026 events: delayed annual reporting, two rounds of restatements, a PCAOB-related audit uplift, a going-concern qualification at the earlier stage of the year, short-seller allegations, law-firm investigations, and regulatory enquiries into a suspected information leak ahead of the U.S. listing announcement.

• It is careful not to allege wrongdoing or treat any regulatory enquiry as an adjudication. Its concern is the cumulative effect of multiple accounting corrections across revenue recognition, FX translation, employee accruals, leases, provisions, and other areas.

• The report argues that a company whose valuation depends heavily on management-defined forward metrics, opportunity-pipeline sizing, and future conversion should carry a higher discount rate when its reported historical financials have required repeated revision.

Scenario logic

• The report’s bear case assumes another delayed ramp, continued NRE runoff, limited Photonics scaling, and a future dilutive raise. It arrives at FY2028 revenue of SEK 380 million and an implied value of about SEK 2.28 per share.

• The base case assumes LiDAR and ALL.SPACE broadly convert, while CPO and pluggable revenue shifts later into 2028 and 2029. It reaches SEK 750 million of FY2028 revenue but still produces a negative return from the reference share price.

• The bull case requires Jabil qualification, continued supply scarcity, successful LiDAR execution, and full use of Glasgow capacity, reaching SEK 1.25 billion in FY2028 revenue. Even this outcome generates only modest upside from the report’s starting price.

• The blue-sky case, in which Sivers becomes a designated second source across multiple hyperscaler optical programs and capacity scarcity persists, is the only scenario where the current entry price works convincingly.

Bottom line

• The report’s conclusion is that Sivers is a real platform in a real AI-photonics opportunity, but the valuation embeds a level of near-flawless execution before product revenue, margin expansion, and backlog conversion have demonstrated it.

• On the report’s assumptions, risk and reward become more balanced in the mid-teens SEK and more attractive below roughly SEK 12, not because that level would make Sivers cheap on traditional metrics, but because it would create room for the ramp to disappoint without permanently impairing the investment case.

• The next critical evidence points are product revenue growth in absolute terms, a recovery in gross margin, firm LiDAR and Jabil production orders, disclosed backlog rather than pipeline alone, and proof that the 2027 transformation is translating into reported revenue.

What readers get in the full PDF
Upgrade to access the full risk/reward framework, including:

• The complete breakdown of product revenue, NRE revenue, Wireless, and Photonics across quarterly, half-year, and trailing figures.

• A full reconstruction of the share-count expansion, directed issues, Bootstrap convertible conversion, warrants, and capital-structure consequences.

• The complete pipeline-versus-backlog analysis, including named orders, required conversion math, and the FY2028 revenue hurdle implied by the current valuation.

• A detailed Glasgow capex and cash-runway model under multiple burn scenarios.

• The report’s full chronology of restatements, audit changes, legal investigations, regulatory enquiries, and governance considerations.

• The complete FY2028 bear, base, bull, and blue-sky scenario framework, plus the specific financial and commercial milestones that would invalidate the cautious view.
  
  
  FULL 14-PAGE MODEL AND PDF ATTACHED BELOW:
  

Read more

SEQH Capital Research | Week of September 14, 2026

The full Week Ahead report goes out this evening. Below is the research agenda we’re working against for the next five sessions, and why each item is on the board.

It’s a heavier week than usual. Four workstreams, one of which has been building for a while.

1. IPOs: Going Deeper

We’ve been tracking the new-issue calendar at a surface level for most of this year. That stops this week.

The deeper work covers the pieces that actually determine whether a listing is investable rather than just tradable:

• Lockup structure and the expiry schedule sitting in front of each name

• Float as a percentage of shares outstanding, and who holds the rest

• Underwriter syndicate composition and what that has historically signaled about aftermarket support

• Use-of-proceeds language versus the actual cash burn in the filing

• Where pricing landed relative to the initial range, and what the revision tells you

The goal is a repeatable framework, not a hot list. Once the framework is built, every subsequent listing gets run through the same filter.

2. SIVE: Risk/Reward at Current Levels

Sivers Semiconductors (Nasdaq Stockholm: SIVE, OTC: SIVEF) is a Swedish photonics and wireless semiconductor business split across two units: a Photonics segment building laser and optical amplifier technology aimed at AI datacenter interconnect, and a Wireless segment in mmWave RFICs and beamforming.

The reason it’s on the list is the gap between narrative and financials. The company has real exposure to one of the strongest secular demand curves in the market, has expanded its stated opportunity pipeline substantially, and has raised capital to fund manufacturing expansion. It is also unprofitable, has used directed share issues to fund itself, and has traded across an extraordinarily wide range over the trailing twelve months.

That combination is exactly where risk/reward work earns its keep. This week’s piece examines:

• Dilution history and what the capital structure looks like after the recent raise

• Product revenue growth versus total revenue, and the quality of the mix shift

• What has to be true on the design-win-to-revenue conversion for current levels to be justified

• The downside case, sized honestly, including cash runway

• Where the asymmetry sits, if it sits anywhere

No price target. A range, an assumption set, and the conditions that would invalidate it.

3. NNE: A Full-Circle Nuclear Buildout

NANO Nuclear Energy (Nasdaq: NNE) is the most complete expression we’ve found of a thesis we’ve been building toward for months: the nuclear trade is not a reactor trade. It’s a supply chain trade.

Over the past year NNE has moved to occupy multiple links in that chain rather than one. Reactor development continues across the KRONOS MMR, ZEUS, and ODIN platforms. On top of that sits a fuel-cycle effort, a transportation and logistics arm following the Secured Transportation Services acquisition, and a widening set of engineering partnerships. Recent public announcements include a detailed-design collaboration with Howden, a Baker Hughes business, on the KRONOS primary helium circulator, a fuel handling and storage collaboration with Fortil, a commercial framework with Tillman Digital Cities, a non-binding memorandum of understanding with Enveniam covering fuel conversion and deconversion and data center power, and a Department of the Air Force award.

Read individually, each is a press release. Read together, it’s an attempt to own the full circle: fuel in, reactor built, power delivered, material moved.

Our work this week underwrites that as a project, not as a ticker. What does the buildout actually cost. What is the realistic timeline from memorandum to binding contract to revenue. What does the balance sheet support without another raise. Where does vertical integration create genuine margin capture versus where does it just add execution risk.

Every number in that piece will be tied to a primary source with a date stamp. Where a figure doesn’t exist publicly, we’ll say so rather than model around it.

4. Macro and Broad Market Sector Forecasts into 2027

The final workstream is the widest. We’re publishing our sector framework for 2027, built off the macro conditions we think are most likely to hold rather than the ones that generate the best headlines.

This covers the rate path and what it does to duration-sensitive equities, where capital expenditure is still accelerating and where it has already peaked, the sectors carrying the heaviest embedded expectations, and the ones where positioning has gotten light enough to matter.

Forecasts get published with their assumptions attached. If the assumption breaks, the forecast goes with it, and we’ll say that in writing.

How We Work

SEQH Capital Research publishes underwriting, not opinions. Every piece states its assumptions up front, sources its inputs, and defines what would prove it wrong. We would rather be specific and revisable than vague and permanently defensible.

All research, including this week’s full reports, is at seqhresearch.com.

SEQH Capital Research publishes informational and educational content only. Nothing in this newsletter is investment advice, a recommendation, or an offer to buy or sell any security. We are not a registered investment adviser or broker-dealer. Securities discussed may be volatile and unsuitable for many investors. Do your own research and consult a licensed professional before making any investment decision.

WHAT THIS REPORT ARGUES

• Nuclear and fuel-cycle equities are often valued as if federal policy has already lowered their cost of capital through DOE loans, Price-Anderson liability protection, and IRA tax credits.

• SEQH’s central finding is that, across Energy Fuels, Oklo, NuScale, NANO Nuclear, and ASP Isotopes, most of that support remains prospective rather than realized. None holds a DOE Title 17 or Energy Dominance Financing loan today, and only Oklo has any live component of the three-pillar policy framework.

The three pillars

• The report evaluates three mechanisms commonly embedded in nuclear-equity narratives: DOE Title 17 or Energy Dominance Financing, Price-Anderson Act liability protection, and IRA Sections 45U and 48E.

• Title 17 and Energy Dominance Financing can lower funding costs through long-term federally backed financing, but recent nuclear deals have gone to established utilities, reactor restarts, or large supply-chain projects rather than to pre-revenue reactor or fuel-cycle developers.

• Price-Anderson is a liability and bankability framework, not simply a lower-interest-rate benefit. It matters because the private market cannot readily insure catastrophic nuclear liability at scale.

• Sections 45U and 48E can potentially improve project economics through production or investment tax credits, but they are not broadly available to the five companies in this report today.

Title 17 reality check

• None of the five covered names holds a Title 17 loan, federal guarantee, or conditional commitment as of the report date.

• The recent federal financing winners include Palisades at $1.52 billion, Constellation’s Crane restart at $1 billion, Southern Company at $26.5 billion, Duane Arnold at $1.9 billion, and a conditional $17.5 billionAP1000 supply-chain package.

• These transactions share a common profile: existing assets, investment-grade utilities, or industrial-scale buildouts with clearer cash-flow visibility than early-stage developers.

• The report’s conclusion is that the market is not pricing a current basis-point funding advantage for the five names. It is pricing the possibility that one or more eventually becomes bankable enough to access a facility that does not yet exist.

Price-Anderson is narrower

• Only Oklo has a current Price-Anderson component, and even that is not full commercial-reactor coverage. It is a DOE contractor indemnity associated with its Reactor Pilot Program activity at Idaho National Laboratory.

• NuScale has a design approval, but its eventual Price-Anderson coverage would attach to a future plant owner or customer licensee, not to NuScale itself as the technology vendor.

• NANO Nuclear has only a third-party pathway through the University of Illinois KRONOS construction permit application. The license belongs to the university, not to NNE.

• Energy Fuels is outside the framework because its operations rely on materials licenses rather than an NRC reactor license.

• ASP Isotopes is structurally excluded even in a hypothetical future enrichment-plant scenario, because post-1992 uranium enrichment facilities are excluded from Price-Anderson coverage under the relevant statute.

Tax-credit reality

• Section 45U is unavailable to all five companies because it applies only to nuclear reactors placed in service before August 16, 2022.

• Section 48E is more relevant to future nuclear builds, offering a 6 percent base investment credit that can rise to roughly 30 percent with labor compliance and potentially toward 50 percent with additional incentives.

• But among the five names, only Oklo has made an explicit first-party argument that it could benefit as a future owner-operator. Even then, the report finds no disclosed monetization strategy, no realized credit proceeds, and no market-clearing reference transaction for a first-of-a-kind new nuclear project.

• NuScale’s potential tax-credit benefit would accrue to a future customer or plant owner rather than to NuScale itself, while Energy Fuels, NANO Nuclear, and ASP Isotopes do not generate electricity and therefore do not qualify under the relevant power-generation credits.

Company implications

• Energy Fuels has a real $725 million federal facility, but it comes from the Department of War’s Office of Strategic Capital rather than DOE Title 17. That creates a separate policy dependency tied to defense budgets, industrial policy, and critical-minerals strategy.

• Oklo has the strongest partial factual basis for a federal-backstop narrative through pilot-phase DOE indemnity and its stated future 48E eligibility. However, the central risk is that neither benefit has become a commercial-scale, monetized asset.

• NuScale is strongly associated with federal nuclear support in investor narratives, yet the report finds no current first-party Title 17 financing, Price-Anderson coverage, or tax-credit benefit. Its prior Title 17 application was tied to the terminated Carbon Free Power Project.

• NANO Nuclear benefits from a real regulatory milestone at the University of Illinois, but the license and any associated liability framework belong to the university rather than to NNE directly.

• ASP Isotopes sits outside all three mechanisms. Its enrichment and isotope ambitions are financed through industrial partnerships and private financing rather than current DOE loan support, Price-Anderson coverage, or relevant electricity tax credits.

The real valuation question

• The report’s most important distinction is between existing legal support and future policy optionality.

• Federal programs may still matter materially to these companies over time, especially as licensing, commercial deployment, customer commitments, and financing structures evolve.

• But as of the report date, the supposed federal discount is mostly not a current balance-sheet benefit. It is a forward-looking expectation that policy pathways will mature, remain available, and ultimately convert into financeable projects.

• That makes the sector sensitive not only to execution and licensing risk, but to legislative, regulatory, and financing-framework risk.

Bottom line

• The cleanest framing is that federal nuclear support is mostly a future option, not a current balance-sheet asset, for this peer group.

• Oklo has the most support in progress and therefore the most to gain if policy pathways mature, but it also has the most exposure if future credits, financing access, or commercial liability coverage fail to materialize.

• Energy Fuels, NuScale, NANO Nuclear, and ASP Isotopes have less current support from this specific policy triad, which means they also have less direct exposure to a reversal of benefits they do not yet hold.

• The report does not argue that policy does not matter. It argues that investors should distinguish between realized financing, existing legal coverage, and narrative optionality before assigning a cost-of-capital discount.

What readers get in the full PDF
Upgrade to access the complete policy-support analysis, including:

• A detailed walkthrough of DOE Title 17 and Energy Dominance Financing, including the current pricing structure, recent closed transactions, and why pre-revenue developers have not yet received comparable support.

• The full Price-Anderson liability map, including the difference between large-reactor coverage, sub-100 MW coverage, DOE contractor indemnity, materials licenses, and enrichment-plant exclusions.

• A deeper analysis of Sections 45U, 48E, 6417, and 6418, including what direct pay, transferability, and stacked credit value could mean if future projects become eligible.

• Individual deep dives on UUUU, OKLO, SMR, NNE, and ASPI, with balance-sheet context, current legal status, and company-specific reversal risks.

• The full policy-support framework, catalyst matrix, and the exact events that could expand or reduce each company’s access to federal financing, liability protection, or tax-credit support.
  
  
  FULL 20-PAGE PDF BELOW:
  
  

Read more

WHAT THIS REPORT ARGUES

• The report corrects a growing market misconception around hollow-core fiber: Microsoft, Corning, and Heraeus announced their outsourced manufacturing ramp on September 23, 2025, not in July 2026.

• More importantly, hollow-core fiber does not make erbium-doped fiber amplifiers obsolete. The amplifier mechanism that genuinely loses its role inside a hollow span is distributed Raman gain, while solid-core EDFA modules can still be spliced into the network at amplification sites.

Core thesis

• Hollow-core fiber guides more than 99 percent of optical power through air rather than solid glass, producing lower attenuation, far lower dispersion, and a three-to-four-order-of-magnitude reduction in nonlinearity versus conventional solid-core fiber.

• Those advantages make hollow-core fiber important for AI networking, especially where low latency, high launch power, and long-distance optical transport matter.

• But the report’s key point is that the transition should not be interpreted as an imminent replacement cycle for all optical amplifiers. It is a multi-year qualification and ecosystem transition, not a near-term revenue reset.

What hollow-core changes

• Microsoft’s DNANF hollow-core fiber has demonstrated record attenuation of about 0.091 dB/km at 1550nm, below the commonly cited Rayleigh-scattering floor for standard solid silica fiber.

• The same architecture offers materially lower dispersion, a group index closer to air than glass, and roughly 30 to 33 percent lower latency than conventional single-mode fiber.

• Lower nonlinearity gives network designers more headroom for higher per-channel launch power, which can increase the value of high-power booster amplifiers over time.

• Microsoft has stated that its hollow-core fiber can deliver up to 47 percent faster data transmission, and more than 1,280 kilometers of hollow-core fiber had reportedly been deployed in Azure by mid-2026.

The Raman distinction

• An EDFA works by amplifying light inside a discrete section of erbium-doped solid silica fiber. That module can remain intact and can be connected to a hollow-core span through standard optical interfaces.

• Distributed Raman amplification is different. It requires the pump and signal to overlap inside the silica transmission fiber itself, which becomes largely impossible when the optical mode travels through an air-filled hollow core.

• This makes Raman gain the true structural casualty of hollow-core fiber, not the broader EDFA product category.

• The report identifies hybrid Raman-EDFA products as the most directly exposed amplifier line, because part of their value proposition comes from the incremental reach delivered by distributed Raman gain in conventional solid-core spans.

Why the timeline matters

• Hollow-core technology is advancing quickly, but it is still early from a standards and qualification perspective. No formal ITU-T recommendation yet exists for hollow-core fiber.

• Connector, splice, amplifier-control, gain-flattening, transceiver, and test-equipment vendors all need to qualify against a different loss curve, different dispersion profile, and different field-installation economics.

• The report estimates a full hollow-core-specific equipment refresh at three to five-plus years, based on standards timing, Telcordia GR-468 reliability qualification, and comparable optical-component commercialization cycles.

• The economics reinforce that timing: hollow-core splices are estimated near $150 each versus roughly $15 for standard single-mode splices, while hollow-core connectors are estimated near $200 versus around $10 for standard connectors.

Lumentum read-through

• Lumentum is the more amplifier-levered of the two names, with more than 200,000 Lumentum-designed EDFAs deployed in commercial networks and a broad portfolio spanning Raman amplifiers, EDFAs, hybrid Raman/EDFA systems, high-power EDFAs, L-band amplifiers, and compact amplifier arrays.

• Its current growth cycle is primarily tied to pump-laser and multi-rail C+L demand, not hollow-core fiber. Management has guided to a fourfold increase in pump-laser shipments, a tripling of pump revenue over three years, and a fivefold rise in pump output.

• Lumentum does sell a product called Hollow-Core Anti-Resonant Fiber, but the report classifies it as a false positive for telecom exposure. It is designed for high-power laser delivery, spectroscopy, sensing, and medical applications, with loss below 50 dB/km, roughly 500 times higher than telecom-grade hollow-core fiber.

• The report finds no disclosed Lumentum roadmap, filing language, or earnings-call discussion linking telecom hollow-core fiber to its amplifier business.

Coherent read-through

• Coherent has direct exposure to the same multi-rail amplifier refresh through its variable-gain EDFAs, fixed-gain EDFAs, arrayed EDFAs, pump lasers, ROADM line cards, and hybrid Raman-EDFA products.

• Its Hybrid Raman-EDFA, which combines Raman and EDFA gain for ultra-long-haul and flex-spectrum systems, is identified as the narrowest direct product exposure if hollow-core deployment reduces demand for distributed Raman contribution.

• At the same time, Coherent is positioned for the opportunity side of the transition: higher-power booster amplifiers, C+L systems, dynamic gain equalizers, and multi-rail amplification architectures.

• Coherent launched a four-rail, 1RU in-line amplifier platform in March 2026 and expects initial multi-rail revenue in the first half of calendar 2027, with management sizing the multi-rail opportunity at more than $2 billion by 2030.

• The report also flags an unconfirmed forward signal from a September 2026 industry session in which Coherent reportedly identified hollow-core and multicore fiber as next-generation technology areas and hinted at possible future announcements.

The refresh cycle actually underway

• The report’s main practical conclusion is that the current refresh is multi-rail C+L amplification, driven by higher wavelength counts, channel density, scale-across AI networking, and greater demand for amplifier capacity.

• Both Lumentum and Coherent have already launched or sampled products aimed at this cycle, including multi-rail channel monitors, C+L tunable lasers, wavelength-selective switches, dynamic gain equalizers, high-power pump lasers, and multi-rail in-line amplifier cards.

• This demand exists regardless of whether a customer deploys solid-core or hollow-core fiber. Hollow-core is an adjacent future technology shift, not the core explanation for today’s amplifier spending.

Bottom line

• The cleanest framing is that hollow-core fiber changes the physics of the span, not the relevance of optical amplification itself.

• EDFAs remain part of the deployed answer, while distributed Raman gain becomes less useful inside a hollow-core link.

• Coherent faces a modest Raman-specific product risk but also has a potential high-power-booster opportunity. Lumentum has stronger existing amplifier and pump-laser scale, but no confirmed telecom hollow-core positioning.

• The near-term revenue catalyst is the multi-rail C+L refresh. A true hollow-core-specific component cycle is likely years away, pending standards, qualification, connector economics, deployment scale, and confirmed vendor roadmaps.

What readers get in the full PDF
Upgrade to access the full report, including:

• The corrected Microsoft, Corning, and Heraeus announcement timeline, including what was actually disclosed and what remains undisclosed.

• A detailed physics comparison of hollow-core versus solid-core fiber, covering attenuation, latency, dispersion, nonlinearity, splice economics, and connector costs.

• The full analysis of why Raman gain disappears while EDFAs remain deployable, including the mechanics of hybrid-span architectures.

• Product-level exposure maps for Lumentum and Coherent, including Raman/EDFA hybrids, high-power booster EDFAs, pump lasers, specialty fiber, and multi-rail systems.

• A complete qualification and standards calendar, including ITU-T status, GR-468 timing, and the catalysts that would signal a real hollow-core component refresh.
  
  
  FULL 19-PAGE PDF LOCATED BELOW:

Read more

SEQH Capital Research | Week of September 7, 2026

Happy Labor Day. We’re publishing this week’s edition on Monday instead of our usual Sunday slot in observance of the holiday, normal Sunday cadence resumes next week. We hope you’re taking the long weekend to recharge before markets reopen tomorrow.

Here’s what’s on our desk this week.

1. Research Releasing This Week

Wednesday: Hollow-Core Fiber and the Amplifier Refresh Cycle

Microsoft ramped hollow-core fiber production with Corning and Heraeus in July 2026, claiming up to 47% faster transmission and roughly a third lower latency versus standard single-mode fiber. Hollow-core fiber carries a different attenuation and dispersion profile than glass fiber, meaning the amplifiers and transceivers tuned for legacy EDFA designs are not drop-in compatible. We see this setting up a coming refresh cycle in amplifier and optics content specifically for campus and metro hyperscaler links. Wednesday’s report maps that opportunity against Lumentum’s (LITE) and Coherent’s (COHR) fiber-optics and amplifier product lines before it becomes market consensus.

Saturday: Nuclear’s Hidden Balance Sheet

Our flagship weekend release is the first systematic comparison of DOE Title 17 loan guarantees, Price-Anderson liability caps, and IRA 45U/48E credit monetization across Energy Fuels (UUUU), Oklo (OKLO), NuScale Power (SMR), NANO Nuclear Energy (NNE), and ASP Isotopes (ASPI). We quantify the effective cost-of-capital advantage these federal backstops confer, and identify which names carry the most exposure if that policy support shifts.

2. Major Market Events to Watch

Equity, bond, and futures markets are closed today for Labor Day. Once trading resumes, the calendar is front-loaded toward inflation data heading into next week’s Fed decision.

• Tuesday, September 8: July Consumer Credit; Treasury announces 3-year note and 26-week bill auctions. Capital Markets Day (ASPI)

• Wednesday, September 9: 10-year note auction. The prior 10-year sale in August priced at the highest yield since the financial crisis, so demand here is worth watching given ongoing supply pressure.

• Thursday, September 10: August Producer Price Index (8:30 AM ET), Existing Home Sales, and a 30-year bond auction.

• Friday, September 11: August Consumer Price Index (8:30 AM ET), the marquee print of the week. Consensus expects headline CPI up 0.4% month over month, holding the annual rate near 3.4%, with core prices up a slower 0.2%, cooling the annual core rate to roughly 2.3% (Cleveland Fed inflation nowcast). The preliminary University of Michigan Consumer Sentiment read for September also lands Friday.

The setup matters because last Friday’s August jobs report ran hot, payrolls up 162,000 versus a 55,000 consensus estimate, with unemployment holding at 4.1% (Reuters). That print pushed short-term rate futures to price roughly 59% odds of a rate increase at the Fed’s September 15-16 meeting, up from about 55% beforehand. This week’s CPI print is the last major data point before that decision, so a hot or cool surprise on Friday should move the hike odds meaningfully in either direction. The FOMC meeting itself, with a fresh Summary of Economic Projections and press conference on Wednesday, September 16, sits just outside this week’s window but is very much on our radar.

3. Quant / Advanced Analytical Data for the Upcoming Week

A data snapshot on the names underpinning this week’s two research releases, plus broader positioning context heading into the CPI print and next week’s Fed decision.

Prices reflect the latest available close. Percentage changes are single-session moves, not weekly performance.

A few things stand out. LITE and COHR have already re-rated sharply, both are trading well off their 52-week lows, with LITE’s average sell-side target now sitting only modestly above spot, so the fiber-optics thesis is arguably more about which specific product lines capture the hollow-core refresh than about the sector re-rating further from here. The nuclear complex tells a different story: every name in Saturday’s coverage list except SMR carries a Buy-or-better consensus, and the average targets across UUUU, OKLO, and ASPI still imply substantial upside from current levels, meaning the market has not yet fully priced the federal backstop dynamics we’re about to quantify.

On the volatility side, the VIX is sitting at 15.30 (up 5.30% on the session), still near the low end of its 52-week range of 13.38 to 35.30. That is a fairly complacent starting point for a week carrying an inflation print, two long-duration Treasury auctions, and a live Fed decision the following week. The Global X Uranium ETF (URA), a reasonable proxy for nuclear-sector sentiment, is at $46.06 (+0.79%), well above its 52-week low of $37.18 but still off its high of $62.28, leaving room to react to Saturday’s report either way.

4. Upgrade to a Paid Membership

Free subscribers get our takeaways. Paid members get the full research, the data behind it, and direct access to the desk.

Paid ($20/month) includes:

• Full-length publication of every report, including Wednesday’s amplifier refresh cycle note and Saturday’s Nuclear’s Hidden Balance Sheet, on release day rather than in preview form

• The underlying quant models and data files behind flagship reports, including the Title 17, Price-Anderson, and 45U/48E comparison framework

• Complete access to the SEQH research archive

• Our running, dated model portfolio for full accountability

• Direct access to the analyst desk for questions on published research

• Custom research requests within our coverage universe

• Priority scheduling for bespoke sector briefings

If you’re ready to move from free previews to full access, upgrade here: Upgrade to SEQH Capital Research

SEQH Capital Research
Institutional-grade research for the nuclear, uranium, and technology infrastructure sectors.

WHAT THIS REPORT ARGUES

• Erbium is a small but strategically important component of the AI networking stack because it is the active dopant inside erbium-doped fiber amplifiers, or EDFAs, which regenerate and boost optical signals across long-haul, metro, and data-center-interconnect networks.

• SEQH’s central finding is that China’s erbium export-control regime is suspended, not resolved. The relevant control was paused before taking effect, but the legal authority remains in place and the current suspension is scheduled to expire on November 10, 2026.

Core thesis

• China added erbium, ytterbium, thulium, holmium, and europium to its export-control list in October 2025, before suspending implementation one month later under a trade-truce framework.

• The risk is therefore not that EDFAs are currently blocked from supply. The risk is that a future reactivation could introduce export licensing, customs delays, supply uncertainty, and a multi-quarter requalification problem across the optical networking chain.

• The report’s key distinction is that this is not primarily a raw-material cost problem. It is a lead-time, licensing, and qualification problem for manufacturers and network customers.

Why erbium matters

• Erbium-doped fiber amplifiers are foundational to modern optical networking because they allow telecom-wavelength light signals to be amplified without converting them into electrical signals first.

• This makes EDFAs critical to the long-distance optical links that connect AI data centers, cloud campuses, metro networks, and broader internet infrastructure.

• The physical amount of erbium in an EDFA is tiny, likely in the sub-microgram to low-microgram range based on the report’s illustrative calculation. That means a major rise in erbium price would have limited direct bill-of-materials impact.

• But replacing or requalifying an optical-fiber supplier can take multiple quarters, while a China export license could take roughly 10 weeks in a best-case scenario and potentially 12 to 16 weeks in practice.

The supply-chain issue

• The report identifies what it calls a zero-producer problem outside China for commercial erbium separation.

• The major non-China heavy-rare-earth projects currently focus on elements such as dysprosium, terbium, samarium, gadolinium, yttrium, lutetium, neodymium, and praseodymium.

• Lynas, MP Materials, and Caremag are all expanding or developing non-China heavy-rare-earth capability, but none is disclosed as targeting commercial-scale erbium separation.

• This means the optical industry has little visible Western redundancy for erbium specifically, even as supply-chain independence improves for other rare earths.

Company exposure map

• Lumentum has direct EDFA exposure through its optical-amplifier product portfolio, 980nm pump lasers, gain-flattening filters, and its own erbium-doped fiber product line.

• Coherent also has direct EDFA exposure through its Nufern subsidiary, which manufactures erbium-doped and erbium/ytterbium co-doped specialty fiber in Connecticut.

• Sivers Semiconductors is structurally insulated from this specific risk because its business centers on indium phosphide lasers, laser arrays, and semiconductor optical amplifiers rather than erbium-doped fiber.

• Applied Optoelectronics also screens as insulated in the report’s filing review. Its CATV amplifiers are RF devices in the coaxial portion of a network and do not incorporate optical amplification or erbium-doped fiber.

The disclosure gap

• The report’s most important research finding is that both Lumentum and Coherent appear to have real product-level erbium exposure, yet neither company’s most recent annual filing specifically names erbium in its rare-earth or critical-materials risk language.

• Lumentum discloses generic risk from rare earth metals and other critical minerals, but does not quantify or identify its erbium-related supply exposure.

• Coherent explicitly discusses its much larger and more visible indium phosphide exposure, including supply and licensing issues, while its Nufern erbium-doped-fiber exposure receives no similar discussion in reviewed filings or earnings calls.

• SEQH does not claim this represents an intentional omission or proves material exposure. The report’s narrower conclusion is that the disclosure gap exists, and it may become relevant if the suspended controls return.

What to watch

• November 10, 2026 is the key date: the current expiration of China’s suspension of the erbium control regime.

• Investors should watch for any Chinese announcement that extends, narrows, or permanently repeals the suspension.

• Further movement in erbium oxide pricing is also relevant. The report cites a secondary-market reading near $82 per kilogram in August 2026, up about 17.6 percent month over month, though it cautions that the intra-year series is not primary-verified.

• Another important signal would be any Lumentum or Coherent filing, earnings-call comment, or investor presentation that names erbium specifically. Such a disclosure could indicate that management considers the exposure material enough to discuss directly.

• Finally, any Western company announcing commercial-scale erbium separation capacity would materially change the supply-chain thesis, because no such project is currently identified in the report.

Bottom line

• The cleanest framing is that erbium is cheap, tiny, and easy to ignore until it is unavailable.

• Lumentum and Coherent appear to have direct exposure through their EDFA and doped-fiber businesses, while Sivers and AAOI appear structurally insulated from the specific erbium chokepoint.

• The relevant question is not whether erbium will suddenly make optical hardware expensive. It is whether a reactivated licensing regime can slow qualification, disrupt fiber availability, and create an unexpected bottleneck in the optical layer of the AI buildout.

What readers get in the full PDF
Upgrade to access the full report, including:

• A complete China rare-earth export-control timeline, from the 2025 policy package through the November 2026 suspension-expiry date.

• A detailed review of erbium’s role inside EDFAs, including the dosing chemistry, active-fiber structure, and why qualification time matters more than material cost.

• The full non-China alternative-supply survey covering Lynas, MP Materials, Caremag, and the absence of commercial erbium separation plans.

• Dedicated company deep dives on Lumentum, Coherent, Sivers, and AAOI, including product architecture, manufacturing locations, filing language, and material-risk disclosures.

• The complete component-level exposure matrix showing erbium, EDFA, indium phosphide, manufacturing geography, and disclosure status across all four names.
  
  

Read more

SEQH CAPITAL RESEARCH
AUG 28, 2026 ∙ PAID

Last September, Oklo was the hottest name in the microreactor trade, up over 470% on the year, when the wheels wobbled in the span of 48 hours. CEO Jacob DeWitte gifted $3 million of stock. Director Michael Klein sold $6.7 million. CFO Craig Bealmear offloaded $9.4 million. Goldman Sachs picked that exact week to initiate at Neutral, flagging valuation and execution risk. The stock dropped 16% in two sessions, not because the reactor got worse, but because insiders and skeptics showed up on the tape at the same time.

Eleven months later, on August 26, 2026, Oklo did it again on a smaller scale: shares fell 6.3% to $41.49 on volume 59% above average, after two more insiders filed Form 4s. Nobody trades these names on discounted cash flow. They trade on the delta between locked-up supply and the news flow that’s supposed to absorb it.

Which brings us to a name almost nobody is running this analysis on: Hadron Energy (NASDAQ: HDRN), a $2-and-change light-water micro modular reactor developer trading under a market cap of roughly $156.6 million. Seventy-seven percent of the company closed the merger locked behind a hard expiration date; that lock-up block is already shrinking as a share of the float before a single restricted share trades, and the regulatory calendar that’s supposed to justify a re-rating runs on a clock measured in years, not months. Sell-side coverage on HDRN is essentially nonexistent and retail commentary treats the lock-up, the warrant stack, and the NRC docket as three separate stories. We think that’s the mistake. They’re one story, and the calendar tells you when it resolves.

Our view: Hadron Energy’s insider lock-up cliff lands on or around November 22, 2026, freeing a block of stock currently worth roughly $118 million into a regulatory and commercial calendar that, on the company’s own disclosed roadmap, doesn’t produce its next hard catalyst (a filed licensing application) for another year and doesn’t reach first-of-a-kind deployment until 2029. Layered underneath, a second, quieter mechanism most coverage misses entirely (a warrant strike that resets down, not up) sits closer to the money than the headline $11.50 figure everyone quotes. This is a structural mismatch, not a coincidence, and it is the dominant variable for anyone holding or shorting this stock into year-end.

Before going further, it is worth being unambiguous about one thing, because it is the most common point of confusion on this name: Hadron Energy has already completed the entire path to being a publicly traded company. There is no pending IPO to wait for and no roadshow ahead. The ticker’s history runs in two distinct legs. GigCapital7 Corp., the blank-check shell, priced and closed its own underwritten IPO on August 30, 2024, selling 20,000,000 units at $10.00 apiece for $200 million of gross proceeds and beginning to trade on Nasdaq under “GIGGU”. That IPO is the origin of every Public Warrant discussed below. Nearly two years later, on May 22, 2026, GigCapital7 completed its reverse merger with the operating nuclear business, and the surviving public company began trading as Hadron Energy under “HDRN”. Every lock-up date, warrant strike, and share count in this piece descends directly from that structure. The company is not pre-IPO by any definition; it is a fully listed, twice-public entity whose float mechanics are simply unusual.

The Shares Nobody Can Sell (Yet)

TLDR:

• Hadron Energy completed its SPAC merger with GigCapital7 on May 22, 2026, converting a nuclear microreactor developer into a Nasdaq-listed shell with 70,173,146 shares outstanding, of which directors, officers, and their affiliated entities held roughly 77.2% at closing.

• That block unlocks on the earlier of three triggers defined in the company’s lock-up agreement. Critically, one of those triggers (six months post-closing) is a fixed calendar date: November 22, 2026.

• A fact almost nobody is tracking: the share count has already grown by nearly 1.9% since closing, from 70,173,146 to 71,498,842 as of August 10, 2026, quietly eroding the insiders’ proportional grip on the company before the lock-up even expires.

• Layered on top: 28,719,000 warrants, structurally “underwater” at current prices, that HDIN Research has flagged as a $333 million liquidity injection the company can’t access unless the stock more than quadruples, and a second, closer-to-the-money trigger buried in the smallest warrant class that the market has priced as essentially worthless.

• Meanwhile, Hadron’s own regulatory roadmap places its next licensing milestone roughly a year out and first-of-a-kind deployment three to four years out, well past the lock-up cliff, even as the company just added a credible operator to the bench.

The plan for this piece looks like this:

First, The Float: We size exactly how much stock becomes sellable on November 22, correct a subtle but real math error the market keeps making about how big that block actually is, and stress-test what even modest insider selling does against the stock’s actual liquidity.

Second, The Overhang: We walk through the full three-class warrant structure, including a downward-resetting strike almost nobody is discussing, and why the “upside case” for HDRN is also, mechanically, a dilution event.

Third, The Calendar: We map Hadron’s NRC pre-application docket, a fresh executive hire, and commercial pipeline against the lock-up date to see whether any binding catalyst is realistically positioned to offset the unlock.

Fourth, The Model: We put it together into a probability-weighted timeline, the Float-Collision Model, and assign rough odds to whether supply and catalyst collide, miss each other, or arrive so far apart it barely matters.

If insiders are structurally unable to be paid in anything but stock for another year of pre-revenue development, if their share of the company is already eroding through routine issuance before the freeze even lifts, and if that stock becomes freely tradable months before the company has anything binding to show for it, the plan writes itself: this is a name where the calendar, not the technology, sets the trade.

Part I – Seventy-Seven Percent Frozen (and Already Shrinking)

The lock-up language in Hadron’s resale prospectus is almost comically specific about when the freeze thaws:

“The Lock-Up Agreement provides that, subject to certain exceptions, each of such stockholders will not transfer any shares of the Common Stock beneficially owned or owned of record by such of the stockholders until the earlier of (a) six months following the Closing; (b) subsequent to the Closing, the date on which the reported closing price of one share of the Common Stock quoted on the Nasdaq equals or exceeds $11.50 per share […] for any twenty trading days within any thirty consecutive trading day period commencing at least ninety days after the Closing; and (c) subsequent to the Closing, the date on which the Company completes a liquidation, merger, stock exchange or other similar transaction […]”

Jargon translation: there are three doors out of the lock-up, and insiders only need one of them to open. Door (c) requires a corporate transaction that isn’t on the table. Door (b) requires the stock to trade at $11.50 for 20 of 30 sessions in a window that can’t start before roughly August 20, 2026, 90 days after the May 22 close. With HDRN trading around $2.19 pre-market as of this writing, more than 400% below that trigger, door (b) is functionally dead for any period we can reasonably forecast. That leaves door (a): six months after closing, flat calendar math, no conditions attached. Six months from May 22, 2026 is November 22, 2026. Barring a negotiated extension nobody has announced, that is the date.

Here’s the worked example, and here is where we’d flag a distinction that most retail coverage of this name gets wrong. Directors, officers, and their affiliated holding vehicles, chiefly founder and CEO Samuel Gibson, who controls 43,474,075 shares across his direct stake, Gibson Family Holdings LLC, and the SG 2026 Irrevocable Exempt Trust (good for 60.8% of the vote on his own), held approximately 77.2% of the company at the moment the merger closed, against a base of 70,173,146 shares then outstanding. That works out to roughly 54.2 million shares under the lock-up, not the larger figure you get by applying 77.2% to today’s outstanding share count, which is a different, larger number for a specific reason worth dwelling on.

Hadron’s own SEC filings show the share count has not sat still since closing: 70,173,146 shares outstanding on May 22, 2026, rising to 70,220,020 by June 30, and jumping to 71,498,842 by August 10, an increase of roughly 1.9% in eleven weeks, with the bulk of that increase (about 1.28 million shares) landing in just the six weeks between the quarter-end and the mid-August filing. None of the Form 4s we reviewed from Hadron’s directors show open-market dispositions; they show RSU vesting awards, including 67,395-share grants each to directors Raanan Horowitz and Robert James Lewis in mid-August, paid in stock rather than cash. Assuming Gibson’s block is effectively unchanged, and adjusting for the modest RSU issuance to independent directors that would partially offset this in insiders’ favor, the arithmetic still points the same direction: the 77.2% insider stake at closing has already drifted down to roughly 75.8% of the company as of August 10, purely from routine share issuance, before a single lock-up share has become legally transferable. Nobody is watching this because it doesn’t show up in a press release. It shows up only if you diff two cover pages three months apart.

The stock’s average daily volume is about 265,789 shares. Run the corrected math forward, using the 54.2 million-share lock-up block rather than the larger figure the market has been implicitly using:

If insiders sell this share of the unlocked block…Shares hitting the tapeEquivalent trading days at average volume5%~2.71 million~10 trading days (2 weeks)10%~5.42 million~20 trading days (1 month)20%~10.83 million~41 trading days (2 months)30%~16.25 million~61 trading days (3 months)

Even a conservative 10% liquidation from the unlocked block, well below what founder-heavy de-SPACs have historically sold in the first quarter after a cliff, represents a full month of the stock’s current average volume landing on the market with no offsetting buyer of record. Nobody has to sell all 54 million shares to move this stock. They just have to sell a fraction of it into a market this thin, and the fraction of the company they’re sitting on is already, quietly, getting smaller.

Part II – The Warrants Nobody Wants (Yet), and the One That Resets Down

If the lock-up is the supply side, the warrant stack is the option value everyone is ignoring because it’s out of the money. Hadron carries three warrant classes:

• 20,000,000 Public Warrants at $11.50/share, originally issued in GigCapital7’s 2024 IPO, exercisable and expiring five years after the May 22, 2026 closing

• 3,719,000 Private Placement Warrants at $11.50/share, same exercisability and expiration terms

• 5,000,000 Hadron Private Warrants at $12.00/share, terminating on the fifth anniversary of closing, but carrying a mechanic none of the other two classes have

That’s 28,719,000 total warrants. Full exercise at current strikes would hand the company roughly $332.8 million in proceeds, which lines up almost exactly with HDIN Research’s characterization of the situation:

“28.7 Million Underwater Warrants Signal Structural Liquidity Deficit […] Hadron’s warrants are functionally neutralized [at $2.74], choking off an anticipated $333 million liquidity injection unless the stock surpasses $11.50/$12.00 thresholds.”

Here’s the part that HDIN, and everyone else we’ve read on this name, treats as a footnote but that we think deserves its own line item. The 424B3 spells out a reset mechanic buried in the Hadron Private Warrants:

“The exercise price is subject to reset on the first anniversary of the Closing to the greater of (i) $6.00, or (ii) the lower of (a) the 30-trading-day volume-weighted average price of the common stock prior to such anniversary, or (b) $12.00.”

Jargon translation: on May 22, 2027, this specific 5-million-warrant slice effectively re-strikes itself downward, to a floor of $6.00, unless the stock is trading meaningfully above that level in the 30 days heading into the anniversary. From $2.19, reaching $6.00 requires a 174% rally; reaching the headline $11.50 requires 425%; reaching the $18.00 forced-redemption trigger on the Public Warrants requires 722%. In other words, the market is fixated on a $11.50 bogey when a mechanically closer, lower-bar dilution trigger is quietly being manufactured by the reset formula itself, one that activates less than nine months from today regardless of what the other 23.7 million warrants do.

Hadron’s own accountants already treat this slice as close to worthless. At the Closing, with the stock at $5.16, the company fair-valued all 5,000,000 Hadron Private Warrants at a combined $6,000 using a Monte Carlo simulation, or about $0.0012 per warrant. By June 30, 2026, with the stock down to $2.03, that combined fair value had fallen further, to $1,515, or roughly three-hundredths of a cent per warrant. That is not a rounding error; it is the company’s own GAAP disclosure confirming that the market is pricing this instrument as a lottery ticket, which is precisely why nobody is watching the reset date. The prospectus also buries a separate redemption mechanic worth spelling out in plain English, because it flips the Public Warrant story from “dead weight” to “double-edged” the moment the stock actually works:

“Outstanding Public Warrants may be redeemed at any time after they become exercisable and prior to their expiration, at a price of $0.01 per Public Warrant, provided that the last reported sales price of our Common Stock equals or exceeds $18.00 per share […] for any twenty (20) Trading Days within a thirty (30) Trading-Day period […]”

Translating the two-price structure: $11.50 is where warrant holders start wanting to exercise; $18.00 is where the company can force the issue by calling the warrants for a penny apiece. A rally that finally justifies the $11.50 to $18.00 range simultaneously injects up to 28.7 million new shares, about 29% dilution against the current share count, alongside whatever the newly-unlocked insider block decides to do with a stock that’s suddenly quadrupled. And even short of that, the $6.00 reset floor means a comparatively modest rally, less than a triple from here, is enough to put the smallest warrant class back in play well before the other two matter at all. The overhang doesn’t cap the stock so much as it guarantees that any real re-rating comes stapled to its own supply shock, in two separate stages.

For now, though, the warrants are a side story. The main event is November 22.

Part III – A Regulator That Moves in Years, Not Weeks

Here’s where the model gets interesting, because the honest answer to “what offsets the lock-up unlock” is: probably nothing that’s binding, though the bench just got stronger.

Hadron’s Halo micro modular reactor has been in NRC pre-application engagement since May 2025, and the docket shows real, if incremental, progress. The timeline as disclosed:

• December 2025: NRC pre-application meeting; the agency gave what the company characterized as favorable feedback on Hadron’s proposed regulatory approach

• April 2026: Hadron submitted its Principal Design Criteria White Paper under 10 CFR Part 52

• April 28, 2026: A non-binding uranium conversion agreement with ConverDyn to secure domestic UF₆ supply for the Halo reactor’s eventual first-of-a-kind unit

• June 4, 2026: The NRC staff issued a final safety evaluation accepting Hadron’s Quality Assurance Program Description for reference in future licensing applications, the first time a light-water microreactor developer has cleared this step

• August 4, 2026: Hadron disclosed the hire of Eric Williams, 51, as Executive Vice President of Engineering, effective upon the start of his employment on August 31, 2026. Williams joins directly from TerraPower, where he most recently served as Executive Vice President and Chief Operating Officer, following prior roles as Senior Vice President and Design Authority (2022 to 2025) and Vice President of Engineering (2020 to 2022)

Every one of the regulatory items is real and every one of them is procedural. None of them is a manufacturing license, a construction permit, or a Combined License Application (COLA), and the NRC’s own public tracking page for Hadron confirms none of the three has been filed, let alone approved. The Williams hire is a genuinely different category of news: pulling a three-decade TerraPower veteran with first-of-a-kind reactor commercialization experience directly onto the engineering org chart is the kind of hire that reduces execution risk on the roadmap Hadron has already published. It is also, mechanically, not a catalyst that resolves the float problem. It doesn’t accelerate an NRC filing date, and it doesn’t convert pipeline into revenue. It is a quality-of-management signal in a stock that trades on a calendar mismatch, not an offset to the mismatch itself.

The company’s own August 12, 2026 stockholder letter lays out the roadmap explicitly, staged from “Today” in increments: pre-application (today) → licensing application (+1 year) → manufacture and test (+2 years) → first-of-a-kind deployment (+3 years) → subsequent application and nth-of-a-kind manufacture (+4 years). Read that roadmap literally: the next hard regulatory catalyst, an actual licensing application filed with the NRC, lands around mid-to-late 2027. First-of-a-kind deployment, the point at which Hadron would plausibly generate real offtake revenue, lands around 2029.

The commercial side tells the same story. Hadron has repeatedly cited an “indicative commercial pipeline” of roughly 8.1 GW by 2035 and more than six data-center prospects at the letter-of-intent stage, including a non-binding memorandum of understanding with Smartland Energy covering up to five behind-the-meter projects and roughly 1.8 GWe of aggregate potential capacity. None of it is a signed, binding, revenue-generating offtake contract. It’s pipeline, not backlog.

One more data point worth putting on the same page as the regulatory clock: cash. Hadron held $22.2 million in cash and no debt as of June 30, 2026, against $4.4 million of net cash used in operations over the first six months of the year, roughly $730,000 a month. At that run rate, the balance sheet funds about two and a half years of operations on its own, comfortably past the lock-up date and close to the licensing-application milestone, which matters because it means Hadron does not have a near-term funding cliff forcing it to tap the warrant proceeds or issue new equity on unfavorable terms. That’s a genuine point in the bull case. It also means there is no balance-sheet pressure valve that would force insiders to hold rather than sell into strength on November 22; if anything, a company that doesn’t need the cash has less reason to coordinate a lock-up extension than one that does.

So line up the calendars. The lock-up cliff is a fixed date: November 22, 2026. The warrant reset is a fixed date: May 22, 2027. The nearest thing to a binding regulatory or commercial catalyst of comparable magnitude is, on the company’s own disclosure, roughly a year further out than the first of those, and Williams doesn’t move that date. There’s no scheduled event between now and year-end that plausibly absorbs a 54-million-share unlock with fresh demand of similar size.

Part IV – The Collision Model

This is the part nobody else is running, because it requires holding the securities-law calendar, the warrant-reset calendar, and the nuclear-licensing calendar in the same spreadsheet.

We assign three scenarios, weighted by how the evidence actually points rather than how the story would read best:

Scenario A, The Gap (our base case, ~65% probability): November 22, 2026 arrives with no binding regulatory or offtake catalyst large enough to offset it. Insiders, who have been compensated almost entirely in RSUs and stock awards rather than open-market purchases since the merger closed, per the Form 4 filings we reviewed, face their first real opportunity to convert years of illiquid equity into cash, on a stake that has already begun eroding as a share of the company through routine dilution. Even modest selling (our 10% to 20% liquidation scenarios above) represents one to two months of the stock’s current trading volume landing with no offsetting catalyst. The stock’s next quarterly filing, likely due in mid-November 2026 based on the reporting cadence of the June 30, 2026 10-Q, would be the last clean data point before the flood, and on the current burn rate it will show a pre-revenue nuclear developer with a healthy but finite cash cushion and nothing to report on the regulatory front beyond “on track.”

Scenario B, The Pull-Forward (~25% probability): A binding catalyst, whether a converted data-center offtake agreement, a Department of Defense contract, or an accelerated NRC step such as an actual licensing-application filing, lands in the September-to-November window and gives holders of the unlocking block a reason to hold rather than sell. The Williams hire is exactly the kind of signal that could precede this: it doesn’t move the regulatory date, but it’s consistent with a management team building toward one. Nothing in the current pipeline is contractually close enough to call this the base case, but Hadron’s rate of procedural and personnel announcements (five or six in the past five months alone) means it can’t be dismissed either.

Scenario C, The Voluntary Extension (~10% probability): Gibson or the other major holders negotiate or self-impose an extension to the lock-up, signaling confidence and buying time for the regulatory calendar to catch up, a move other de-SPAC founders have made when they didn’t want to be the first sellers into a thin tape. There’s no indication this is being discussed, Gibson’s control position gives him little incentive to do it unilaterally, and a company sitting on 30 months of runway has one less reason to feel pressure to reassure the market this way, which is why we weight it lowest.

Netting it out: the model leans decisively toward a supply-side air pocket arriving months ahead of any demand-side catalyst large enough to matter, compounded by a warrant structure whose nearest trigger sits far closer than the headline number suggests, with the stock’s own recent behavior arguably front-running the risk rather than the reward. HDRN fell from its $5.48 intraday high on the day the merger closed to a post-merger closing low of $1.47 on August 3, 2026, having actually touched $1.19 intraday on July 30, a level deeper than the closing-price low most commentary cites, before rallying 51% to close at $2.22 on August 27. That’s a stock that’s already round-tripped through a de-SPAC redemption unwind and found a bid well before the real overhang, in either of its two forms, even shows up.

The Trade: Positioning Into November 22 and May 22

The Float-Collision Model transmits into price action through four channels:

Volume Compression Before the Date: Expect bid-side liquidity to thin further into early November as market makers and existing holders price in the unlock; this is the highest-conviction, lowest-cost window to be positioned ahead of the event rather than reacting to it.

The Post-10-Q Air Pocket: If the next quarterly filing (expected mid-November) shows continued cash burn with no binding offtake or licensing news, it sets up the worst possible sentiment backdrop right before the unlocked block becomes sellable.

The Warrant Trap on Any Rally: Any sharp HDRN rally toward the $6.00 zone, not just the $11.50 zone everyone else is watching, should be read skeptically. Crossing $6.00 into the May 22, 2027 reset window starts pulling the Hadron Private Warrants toward relevance nearly a year before the Public and Private Placement Warrants matter, meaning the rally itself manufactures a supply event well before the headline strike is even close.

The Second Anniversary: May 22, 2027 is now a date worth marking on the same calendar as November 22, 2026. It is when the market finds out whether the smallest warrant class resets to a hard $6.00 floor or to something tied to the 30-day VWAP heading into that date, and either way it is the first moment the “underwater warrant” story stops being uniformly true across all three classes.

Below the paywall, we detail our specific options structure for expressing this into both the November 22 and May 22 dates, our downside price targets under each of the three scenarios above, and the three other 2026-vintage de-SPAC nuclear names (beyond Oklo) where an identical lock-up and reset-mechanic mismatch is quietly building.

WHAT THIS REPORT ARGUES

• SEQH reconstructs the economics of 12 hyperscaler-linked nuclear transactions involving Microsoft, Meta, AWS, and Google, representing more than 10 GW of restarted, uprated, or pre-ordered nuclear capacity.

• The core conclusion is that nuclear pricing is rising, but the investment implications are more nuanced: the premium increasingly reflects asset scarcity and the repriced PJM capacity market, while public equities have become far less responsive to announcement headlines.

Core thesis

• Only a small subset of the deal book discloses actual dollar economics, so this report separates company disclosures, named third-party price estimates, and explicit SEQH calculations rather than treating all announced gigawatts as equally valuable.

• Existing-fleet restart and uprate transactions generally back-solve into a $65 to $115 per MWh range, while the newest PJM and ERCOT deals cluster closer to $90 to $120 per MWh.

• SEQH’s broader read is that clean, firm nuclear power is no longer being priced against cheap legacy renewables or old energy-only power curves. It is increasingly being priced against the full cost of reliable grid power for hyperscale data centers.

The real driver: PJM repricing

• PJM’s all-in wholesale power cost reached about $114.50 per MWh in the first half of 2026, up more than 50 percent year over year, while load-weighted energy pricing also moved sharply higher.

• Capacity pricing drove much of that change: the PJM base residual auction increased from $28.92 per MW-dayin 2024/25 to nearly $270 per MW-day in 2025/26, then rose to administrative-cap levels above $329 per MW-day for later delivery years.

• That changes the comparison for hyperscalers. A roughly $90 to $115 per MWh nuclear contract can still carry a premium to energy-only pricing, but it is no longer necessarily expensive versus firm, unhedged grid power with capacity costs included.

Deal-book pricing

• Microsoft and Constellation’s Crane / Three Mile Island restart remains the highest-profile transaction, with implied pricing estimated around $101 to $115 per MWh for 835 MW of restarted capacity.

• Meta and Constellation’s Clinton deal is estimated near $70 per MWh for 1,121 MW of uprated output, materially lower than the Crane restart despite being signed into a stronger power market.

• Meta and Vistra’s PJM uprate agreements, covering 2,609 MW across Perry, Davis-Besse, and Beaver Valley, are estimated around $101 per MWh at full run rate by BMO, with other estimates in an $85 to $100 per MWhrange.

• AWS and Vistra’s Comanche Peak agreement for 1,200 MW is estimated between $90 and $120 per MWh, while AWS and Talen’s restructured Susquehanna PPA implies roughly $65 to $92 per MWh, depending on whether the calculation uses term-average or full-volume economics.

Who has pricing power

• The report’s most counterintuitive finding is that Constellation did not sustain its own pricing power across its two headline deals. Its 2024 Crane restart carried the premium valuation, while the 2025 Clinton uprate priced materially lower despite the market’s broader move higher.

• Vistra appears to have captured the strongest economics in the newest transaction cohort, with Meta’s PJM agreements matching or exceeding Constellation’s earlier peak on some estimates.

• SEQH argues that the premium attaches less to any single company’s negotiating skill and more to specific asset scarcity, such as cold restarts, high-value uprates, location, and the prevailing capacity-price regime when the contract is signed.

SMR pre-orders are not PPAs

• The report draws a sharp line between signed, operating-fleet transactions and new-build SMR or advanced-reactor frameworks. Meta-TerraPower, Meta-Oklo, Google-Kairos, Standard Power-NuScale, Switch-Oklo, and Equinix-Oklo do not provide firm, comparable power-price disclosure.

• In SEQH’s view, these should be treated as options on future clean-firm capacity, not as revenue-equivalent contracts.

• The key precedent is the 1,848 MWe Standard Power and NuScale framework, which never became a binding PPA. The report does not claim current pre-orders will fail, but it argues that investors should apply a conversion discount to unpriced, non-binding gigawatt headlines.

• For Oklo specifically, the more meaningful unit of analysis is cash and optionality: Equinix’s $25 million prepayment and related most-favored-nation discount right, rather than the headline future MW figure.

Why stocks are not reacting

• The report finds that announcement alpha has decayed. The largest reactions occurred in the 2024 cohort, including Constellation up 22.3 percent on the Microsoft / Crane announcement and Talen up 12.7 percent on the AWS Cumulus transaction.

• By 2025 and 2026, price reactions became mixed. Vistra fell 4.5 percent on the Comanche Peak news but rose 10.5 percent on the larger Meta PJM deal announcement.

• Despite record nuclear-hyperscaler deal activity, Constellation, Vistra, Talen, Oklo, NuScale, and X-energy were all down or materially below prior highs in 2026 according to the report’s August 17 snapshot.

• The practical conclusion is that stocks are now responding more to deal scale, scarcity narrative, and credibility than to back-solved per-MWh economics that the market often cannot observe directly.

What matters next

• The next major pricing signal is whether PJM capacity auctions stabilize or remain near the administrative cap as more supply comes online.

• Crane / Three Mile Island’s restart schedule and actual first power will be the live test of whether restart economics can match the premium embedded in the contract.

• Any disclosure of actual Meta-Vistra, Meta-TerraPower, or Meta-Oklo pricing would materially improve the market’s ability to value the sector.

• The most important risk monitor for SMR names is whether pre-orders convert into binding PPAs with disclosed terms, rather than quietly lapsing or being revised.

What readers get in the full PDF
Upgrade to access the full nuclear PPA economics framework, including:

• The complete 12-deal hyperscaler nuclear scorecard, with capacity, term, transaction type, and price-source methodology.

• A full pricing-power ranking across Constellation, Vistra, Talen, Oklo, and the key hyperscaler counterparties.

• The detailed PJM repricing bridge, including energy, capacity, all-in wholesale-cost, and renewable-PPA comparisons.

• The complete public-company reaction table across CEG, VST, TLN, OKLO, SMR, and XE.

• A deeper analysis of why SMR pre-orders should be modeled as options rather than contracts, plus the catalysts that could turn announced gigawatts into financeable future revenue.
  
  
  FULL EXCEL FILE DATA-SHEET AND 20-PAGE PDF REPORT ATTACHED BELOW:
  
  

Read more

WHAT THIS REPORT ARGUES

• SEQH is constructive on photonics heading into Q4 2026, arguing that the sector remains early in a multi-year AI bandwidth upgrade cycle rather than late in a fully priced one.

• The report ranks 25 companies across the optical interconnect stack using a proprietary AI-Bandwidth per Dollar framework designed to separate genuine technology and capital-efficiency leaders from names trading mainly on AI optics narrative.

Core thesis

• Every new unit of GPU compute needs more optical bandwidth between accelerators, switches, and storage. As AI clusters scale, transceivers, coherent DSPs, silicon photonics, co-packaged optics, lasers, and specialty fiber are becoming performance bottlenecks rather than commodity components.

• The report measures two distinct attributes: Tech Frontier, based on disclosed channel bandwidth relative to the sector’s best public benchmark, and Capital Efficiency, based on revenue generated per dollar of capital expenditure.

• The combined score identifies companies that not only participate in the optical upgrade cycle but are positioned to monetize it efficiently.

The top-ranked names

• Lightmatter ranks first overall, driven by its photonic-compute-fabric and co-packaged-optics positioning, though it remains private and lacks public capital-efficiency disclosure.

• Nokia / Infinera ranks first among public companies at 89.6, followed closely by Ciena at 89.3. Ciena combines a perfect 100.0 Tech Frontier score with a 78.7 Capital Efficiency score, placing it ahead of Broadcom, NVIDIA, and other large semiconductor names.

• Broadcom ranks fifth at 74.6, supported by exceptional capital efficiency, while Semtech, Corning, Eoptolink, Cisco / Acacia, and Coherent round out the top public cohort.

• Sivers Semiconductors ranks 14th overall with a 52.1 composite score, driven by a top-quartile 79.3 Capital Efficiency score rather than a headline bandwidth specification.

The real sector signal

• Across the 20 companies with enough disclosed generational data, the median growth rate in per-channel bandwidth is 34.9 percent annually.

• The 800G-to-1.6T transition is especially important. Coherent, Marvell, Lumentum, and Semtech are all moving through this generation shift at a tightly grouped pace of roughly 41 percent annual bandwidth growth.

• SEQH interprets this clustering as evidence that 1.6T is becoming an industry-wide deployment checkpoint, not a single-company technical achievement.

Coherent deep dive

• Coherent ranks 11th overall with a 55.6 composite score and sits in the report’s Tech Pioneer quadrant: ahead of the market technologically, but not yet a top-quartile capital-efficiency operator.

• Its 1.6T transceiver portfolio spans silicon photonics, externally modulated lasers, and VCSEL technologies, while its 800G-to-1.6T transition implies a 41.1 percent annual growth rate in per-channel bandwidth.

• The company generated $5.81 billion in FY2025 revenue, carries a reported 35.2 percent gross margin, and has consensus estimates calling for revenue to rise from about $7.06 billion in FY2026 to $13.43 billion by FY2028.

• The report also highlights Coherent’s NVIDIA relationship, including a $2 billion equity investment and multibillion-dollar purchase commitment, as a central demand and execution variable.

Sivers deep dive

• Sivers is positioned as an Efficient Operator, not a technology laggard. Its InP laser and photonic-integrated-circuit products are specified in optical power rather than a directly comparable Gbps metric, so its Tech Frontier score is conservatively capped at 25.0.

• Its core strength is financial conversion: FY2025 revenue of SEK 304.1 million, gross margin of 87.2 percent, and a two-year revenue CAGR of 13.4 percent support its 79.3 Capital Efficiency score.

• The company’s platform includes InP100 continuous-wave DFB laser arrays and external laser source products aimed at co-packaged optics, with ecosystem relationships spanning Ayar Labs, GlobalFoundries, O-Net, Enablence, and Jabil.

• The key re-rating question is operating leverage. Sivers has a very high gross margin but a reported negative 46.5 percent operating margin, meaning the path to breakeven depends on laser-platform revenue scaling faster than fixed R&D and SG&A.

The risk investors cannot ignore

• The report treats customer concentration as an independent risk overlay rather than something solved by a high composite score.

• Fabrinet, Credo, and Zhongji Innolight carry Tier 1 severe concentration risk. Fabrinet’s two largest customers represented 45.8 percent of FY2025 revenue, Credo had one customer represent 86 percent of a recent quarter, and Zhongji’s four disclosed hyperscaler customers accounted for 89.4 percent of revenue.

• By contrast, Sivers is placed in the report’s Tier 4 diversified group, while Coherent is Tier 3 moderate. That does not eliminate business risk, but it gives both different exposure profiles than the high-concentration transceiver and connectivity names.

What matters next

• The major Q4 watch items are the pace of 1.6T qualification and volume deployment, the commercialization timeline for co-packaged optics, execution on Coherent’s forward revenue ramp, and Sivers’ path to operating breakeven.

• The report also flags ongoing consolidation in the sector, including Marvell’s acquisition of Celestial AI and Ciena’s acquisition of Nubis Communications, as evidence that strategic buyers are increasingly willing to own key optical-engine and photonic-fabric layers.

• The clean takeaway is that AI optics is broadening beyond one or two semiconductor winners: the sector is becoming a multi-layer buildout spanning switch silicon, DSPs, lasers, silicon photonics, module assembly, optical engines, and fiber infrastructure.

What readers get in the full PDF
Upgrade to access the full 25-name ranked watchlist, including:

• The complete AI-Bandwidth per Dollar ranking table across public, private, and recently acquired photonics assets.

• The full Tech Frontier versus Capital Efficiency matrix, showing which names screen as Leaders, Tech Pioneers, Efficient Operators, or Laggards.

• A detailed bandwidth-velocity table tracking the generational transition pace from 800G to 1.6T and beyond.

• Full deep dives on Sivers Semiconductors and Coherent, including product positioning, financial profiles, customer ecosystems, and forward catalysts.

• The complete customer-concentration framework, risk tiers, market-cap data, and sector watch items through Q4 2026 and into 2027.
  
  
  FULL PDF AND EXCEL DATA-SHEET ATTACHED BELOW:
  

Read more

WHAT THIS REPORT ARGUES

• This note is a research-only comparability analysis, not a rating or price-target piece, and its purpose is to show that Sivers’ recent SEK-reported revenue trend materially overstates underlying deterioration because of currency translation.

• SEQH’s core point is that investors looking only at headline Swedish-krona growth may be reading a mixed signal, where business performance and FX movement are bundled together and made to look like one operating number.

Core thesis

• Sivers reported Q1 2026 revenue of SEK 61.9 million, down 21.1 percent year over year, but management separately said the constant-currency decline was only about 11 percent.

• SEQH uses that one disclosed anchor point to build an independent FX-neutralization model and then reconstruct the prior four quarters on the same basis.

• The conclusion is that Sivers’ recent deceleration was real, but the shape and severity of the slowdown look meaningfully different once FX is stripped out.

What the model says

• SEQH calibrates an FX-sensitivity coefficient of 0.894 using the Q1 2026 gap between reported and constant-currency growth and applies it to a simple equal-weight USD/SEK and GBP/SEK basket.

• Under that model, underlying growth actually appears to have peaked in Q2 2025, not Q1 2025, which means the headline series makes the business look like it rolled over earlier than it likely did.

• The reconstructed implied constant-currency growth series runs at about 37.8 percent in Q1 2025, 44.6 percent in Q2 2025, 29.8 percent in Q3 2025, 15.2 percent in Q4 2025, and minus 11.0 percent in Q1 2026.

How much FX mattered

• In absolute terms, SEQH estimates that if Q1 2026 had translated at the prior year’s average FX rates, Sivers would have reported about SEK 69.9 million of revenue instead of SEK 61.9 million, an SEK 8.0 million gap attributable to translation rather than operations.

• Across the four quarters from Q2 2025 through Q1 2026, the model implies total FX drag of about SEK 22.5 million, equal to roughly 7.8 percent of trailing four-quarter revenue.

• The quarter-by-quarter drag becomes progressively larger as the krona strengthens, moving from about SEK -3.2 million in Q2 2025 to SEK -8.0 million in Q1 2026.

Why this matters for comparability

• The report argues that Sivers is especially easy to misread because it reports in Swedish kronor, while peers like Coherent, Lumentum, and Applied Optoelectronics report in U.S. dollars.

• That means a period of SEK strength can mechanically make Sivers’ headline growth look worse than a dollar-reporting peer’s growth, even if the underlying commercial environment is closer than the raw numbers imply.

• SEQH’s cleanest example is Coherent in Q1 2026: the headline growth gap versus Sivers is about 41.7 percentage points, but falls to roughly 31.5 points once Sivers is restated on an implied constant-currency basis.

What the model does not say

• SEQH is explicit that this does not eliminate the performance gap versus peers, especially against stronger AI-optics names like Lumentum and AAOI.

• The report also stresses that the model has important limitations: it is calibrated from one disclosed quarter, uses a fixed 50/50 USD and GBP basket, and cannot rely on company-disclosed currency or segment revenue splits because those are not published.

• So the analysis should be read as a useful correction to the headline, not as a definitive restatement of company economics.

Broader read-through

• The main takeaway is that recent SEK-denominated softness at Sivers was partly operational and partly translational, and those two effects should not be conflated when comparing the company to itself or to its photonics peers.

• In practical terms, the note gives investors a more defensible way to discuss whether Sivers’ slowdown is being overstated by reporting currency rather than by business fundamentals alone.

• The simplest framing is that FX did not fix the quarter, but it did distort the quarter.
  
  FULL 20-PAGE REPORT WITH MODELS ATTACHED BELOW:
  

Read more

WHAT THIS REPORT ARGUES

• This note treats York Space Systems’ acquisition of ALL.SPACE not as a standalone M&A story, but as a supply-chain event with direct relevance to Sivers Semiconductors’ SATCOM business.

• SEQH’s core point is that when a named terminal customer becomes part of a capitalized, backlog-funded defense prime, the meaning of a chip supplier’s production order changes as well.

• The near-term effect is validation and stronger counterparty quality. The longer-term effect is a new strategic question about whether the same prime could someday try to own more of that subsystem stack itself.

Core thesis

• York closed the ALL.SPACE acquisition on July 8, 2026 for about $300 million in cash and stock, down from the originally announced $355 million because York’s own share price fell between announcement and closing.

• At the time of close, ALL.SPACE was already six weeks into a named $8.2 million production order with Sivers for Ka-band beamforming ICs used in the Hydra terminal family.

• SEQH argues that this changes Sivers’ SATCOM exposure from a relationship with a single terminal company into indirect exposure to a public defense prime’s balance sheet, backlog, and procurement channels.

Why York matters

• York is not framed here as just another satellite company. It is presented as a PWSA-anchored defense primewhose core franchise has generated more than $1.2 billion in cumulative SDA-related awards since 2020.

• York also entered public markets with significant scale, raising $629 million in its January 2026 IPO, even though the stock later fell sharply from the offering price.

• That matters because an embedded supplier serving a venture-backed terminal maker faces a very different risk profile than one whose customer now sits inside a NYSE-listed, backlog-funded prime contractor.

Why ALL.SPACE matters

• ALL.SPACE’s Hydra terminal range is the real operating bridge between York and Sivers. The company had already built a multi-orbit, multi-band SATCOM product with defense and commercial relevance before York acquired it.

• Hydra MAX had reached TRL 6 under the U.S. Army’s Next Generation Tactical Terminal program, achieved Viasat GX Category 4 certification, and engaged with the U.S. Navy and Royal Canadian Navy before the deal closed.

• SEQH’s read-through is that York did not acquire a speculative concept. It acquired a validated, already-qualified terminal supplier, which makes the supplier stack around Hydra much more strategically relevant.

What it means for Sivers

• Sivers’ role is described as an embedded, jointly developed, and now production-validated chip supplierinside the Hydra architecture.

• Using SEQH’s prior internal estimate of about 1,800 Sivers chips and roughly $9,000 of Sivers content per Hydra terminal, the $8.2 million production order implies something like 900 terminal-equivalents, though the note is careful to label that as illustrative rather than a disclosed unit count.

• The important point is not the exact unit math. It is that Sivers’ largest named SATCOM production order now sits behind a customer base that includes York’s defense infrastructure, ALL.SPACE’s military terminal programs, and end users across Army, Navy, allied defense, and GX-linked networks.

Validation and risk

• SEQH sees the acquisition as a near-term validation event because York has effectively endorsed a terminal platform that already includes Sivers technology at the chip layer.

• But the report also flags a longer-term strategic risk. York’s 2026 acquisition sequence, including Orbion, ALL.SPACE, and Solestial, suggests a pattern of buying already-validated critical subsystems rather than merely sourcing them forever.

• That creates a real, even if not imminent, question for Sivers: if York prefers to own validated layers of its mission stack, could beamforming IC capability one day appear on a future integration shortlist as well.

Market and volume read-through

• The broader defense SATCOM market is large enough that this relationship can matter if it scales. The note cites a global tactical SATCOM terminals market of about $4.2 billion in 2024 and a land-based terrestrial military SATCOM market of $9.17 billion in 2025, with a path to more than $22 billion by 2034.

• Individual U.S. Army terminal programs already run into the tens to low hundreds of millions of dollars, which means chip content can compound meaningfully if Hydra-class systems win share inside larger procurement channels.

• SEQH’s bottom line is that Sivers still supplies only a fraction of terminal value, but York’s ownership gives ALL.SPACE access to a much larger commercial and defense demand base than it had as a standalone company.

Bottom line

• The clean takeaway is that York’s acquisition of ALL.SPACE is bullish near term for Sivers’ SATCOM credibility, because it upgrades the quality of the counterparty behind Sivers’ most important named SATCOM production order.

• At the same time, it introduces a more sophisticated long-term question about vertical integration risk, because the same logic York used to buy validated propulsion, power, and terminal assets could eventually reach deeper into the subsystem stack.

• So this is best read as validation now, strategic watchpoint later.

What readers get in the full PDF
Upgrade to read the full report, including:

• A full timeline of York’s 2026 acquisition sequence and why Orbion, ALL.SPACE, and Solestial fit the same ownership logic.

• A deeper breakdown of Hydra terminal architecture, defense certifications, and customer set.

• The detailed chip-content economics behind the Sivers order and what changes when the buyer sits inside a defense prime.

• SEQH’s full discussion of the integration shortlist, including why vertical integration is both a validation signal and a future supplier risk.

• Market sizing tables, supporting exhibits, and the full data provenance and methodology behind the SATCOM demand analysis.
  
  FULL 20-PAGE PDF ATTACHED BELOW:
  

Read more

WHAT THIS REPORT ARGUES

• This note argues that Sivers’ early July share count jump should not be read as one generic dilution event, but as two distinct capital actions with different purposes, mechanics, and governance implications.

• SEQH’s core point is that the company used a market-priced directed equity raise to fund growth while a pre-existing fixed-price debt conversion simultaneously delevered the balance sheet, creating a coordinated capital-structure reset rather than a single opportunistic financing.

Core thesis

• Between June 15 and July 3, 2026, Sivers moved from about 320 million shares to more than 355 million shares, an increase of roughly 11 percent in just over a week.

• But SEQH argues the market is missing the key distinction: one leg was board-controlled and discretionary, while the other was contractual and lender-controlled, meaning they should not be analyzed as if management made both decisions at the same time on the same terms.

• The report is not trying to say dilution did not happen. It is trying to show that the two events served different balance-sheet functions and therefore send different signals about future capital strategy.

What happened

• The first action was the SEK 700 million directed share issue completed on July 1 at SEK 57.00 per share, priced at a 9.7 percent discount to the prior close and upsized from the originally indicated SEK 600 million.

• The second action came on July 3, when Bootstrap Europe converted $12 million of debt into 22.85 million new shares at a fixed legacy strike price of SEK 4.77, settled entirely by set-off rather than cash.

• SEQH treats these as linked but different: the first increased cash for expansion, while the second reduced debt and future interest burden.

Why the distinction matters

• The directed raise was a forward-looking capital allocation decision. The board chose the timing, size, investor book, and discount in order to raise fresh growth capital.

• The Bootstrap conversion was different because the pricing had effectively been locked in months earlier, when the refinancing was arranged in February and March 2026 near the stock’s low.

• That means the steep discount embedded in the conversion was not a new July pricing choice, but the delayed consequence of distress-era financing terms struck before the stock rerated.

Headroom and forward optionality

• One of SEQH’s most important findings is that the July 1 placement only used 12.28 million shares of the AGM’s 53.84 million-share discretionary issuance ceiling.

• That means only about 22.8 percent of the board’s authorized issuance capacity was consumed, leaving roughly 41.56 million shares still available without another shareholder vote.

• In practical terms, SEQH is telling readers that Sivers still has meaningful board-level financing flexibility left, independent of any future contractual debt conversions.

Balance sheet effect

• The report reconstructs the balance sheet to show why the second leg mattered. Total debt had risen to about SEK 177.3 million by March 31, 2026, while cash stood at only SEK 26.6 million, leaving the company more levered going into July than many investors appreciated.

• By converting about SEK 109 million of debt into equity, Sivers materially reduced its interest-bearing obligations at the same time that the directed raise added SEK 700 million of gross cash.

• SEQH’s framing is that the company addressed two constraints at once: cash for expansion and debt for simplification.

Governance and signaling

• The note also spends time on what it calls lock-up choreography, including the waiver and reset of the April placement restriction and the separate insider personal lock-up expiry scheduled for July 16, 2026.

• SEQH argues that these dates matter because they create distinct supply-side events for the stock, separate from the fundamental question of whether the raise itself was strategically rational.

• More broadly, the report connects the financing window to a wider governance overhang that included the FY2025 restatement, short-seller pressure, regulatory inquiry, and open legal investigations, without taking a view on the merits of any of those matters.

Strategic read-through

• The bullish interpretation is that Sivers raised capital into strength rather than weakness, locking in funding while the stock still traded far above its February lows and before governance uncertainty had a chance to further raise its cost of capital.

• The more cautious interpretation is that shareholders should now assume management has both the willingness and the remaining authorization to continue using equity proactively if it believes the strategic payoff justifies the dilution.

• SEQH’s overall read seems to be that this was a rational but expensive reset: painful for holders in the short run, but potentially supportive of the company’s longer-term capacity buildout, R&D roadmap, and dual-listing ambitions.

What paid readers get in the full PDF
Readers who upgrade get the full deconstruction below, including:

• A day-by-day timeline reconstruction of the AGM, placement, conversion, and stock reaction.

• A clearer taxonomy of discretionary versus contractual dilution, and why that distinction matters more than the headline share-count jump.

• A full breakdown of remaining issuance headroom, lock-up expiries, and what they imply for future supply.

• A balance-sheet walk-through showing the before-and-after debt structure and the real deleveraging effect of the Bootstrap conversion.

• A deeper section on governance, raise timing, employee incentive deferral, and use of proceeds, all tied back to what this means for Sivers’ next 12 months.
  

Read more

WHAT THIS REPORT ARGUES

• This note argues that silicon-28 is becoming one of the most strategically important isotope markets in advanced technology, and that ASP Isotopes may be moving into a near-monopoly position in Western supply.

• SEQH’s core point is that the market is still treating silicon-28 as an obscure specialty product, when it may actually be a critical quantum-enabling material with unusually strong pricing power, limited competition, and very high margin potential.

Core thesis

• The report frames silicon-28 as a chokepoint material for quantum computing and other next-generation applications, not just another isotope product inside the ASP portfolio.

• Unlike broader isotope baskets, this market appears unusually concentrated because the technical difficulty of enrichment is high, qualification pathways are narrow, and end users care more about purity and continuity than about broad supplier choice.

• SEQH’s conclusion is that if ASP can establish itself as the trusted commercial source of enriched silicon-28, the value of that position could be much larger than investors currently assign.

Why silicon-28 matters

• Silicon-28 matters because isotopically purified silicon reduces nuclear-spin noise, which makes it especially valuable for quantum computing architectures that rely on long coherence times.

• That gives the material an unusually attractive demand profile: the volumes are relatively small, but the strategic value per kilogram can be extremely high because performance matters more than commodity cost.

• The report appears to argue that this creates a market structure where a qualified supplier can capture economics that look closer to a technology monopoly than to a normal specialty-material business.

Why ASP could dominate

• SEQH seems to argue that ASP’s edge comes from a combination of technical capability, timing, and lack of credible Western alternatives.

• In this framing, the market is not asking whether silicon-28 demand exists, but whether anyone besides ASP can produce it at the right purity, in the right jurisdiction, and with the right commercial consistency.

• That is what gives the report its monopoly language: not necessarily monopoly in a legal sense, but a practical monopoly created by capability, scarcity, and customer qualification barriers.

Relationship to QLE

• The note appears to connect the silicon-28 thesis to the broader QLE and isotope sovereignty narrative, arguing that ASP is gradually assembling a platform built around hard-to-replicate isotope positions rather than one-off niche products.

• In that context, silicon-28 is important not only for its own economics, but because it strengthens the company’s identity as a Western strategic isotope supplier across both quantum and nuclear-linked markets.

• SEQH’s broader message seems to be that ASP should be understood less as a speculative collection of science projects and more as a portfolio of isotope chokepoints with asymmetric strategic value.

Economics and valuation read-through

• The report appears to imply that silicon-28 could become one of the highest-margin products in the company’s portfolio because of low-volume, high-value demand and limited competitive pressure.

• That matters because investors often undervalue businesses like this by using conventional materials logic, when the better framework may be scarcity pricing plus strategic dependency.

• In other words, the upside is not just more revenue, but the possibility that ASP controls a product category where customers have very few acceptable substitutes.

Why the market may still be missing it

• SEQH seems to think the market still underestimates silicon-28 because it sits at the intersection of quantum technology, isotope enrichment, and sovereignty, which means no single investor group fully owns the story.

• Quantum investors may overlook the enrichment bottleneck, while materials investors may not fully appreciate the end-market importance of isotopically pure silicon.

• That information gap is what allows the opportunity to remain mispriced even as the strategic logic becomes clearer.

Bottom line

• The cleanest way to read this note is that silicon-28 is not just an ASP product, but a potential monopoly-grade strategic asset.

• SEQH is arguing that if ASP proves commercial scale, purity, and supply continuity, the market may eventually treat silicon-28 as one of the company’s most important value drivers rather than as a small specialty side business.
  
  
  FULL 24-PAGE REPORT ON ASPI/QLE SILICON-28 CHOKEPOINT WITH EXTENSIVE PRICE MODELING AND FORECASTING ATTACHED BELOW:
  
  

Read more

WHAT THIS REPORT ARGUES

• This note breaks Sivers’ reported commercial pipeline into probability buckets and argues that the market is still mixing together very different kinds of opportunities, from near-term production programs to early-stage optical design-ins.

• SEQH’s central point is that Sivers is not one single bet but a collection of distinct revenue pathways, each with different timing, certainty, customer concentration, and margin structure.

Core thesis

• The report argues that investors have been treating the pipeline as a headline number when they should instead treat it as a stack of different conversion probabilities.

• Some parts of the pipeline are tied to named customers, manufacturing readiness, and visible qualification paths, while others are still closer to strategic options than revenue.

• The exercise is meant to improve underwriting discipline, not weaken the Sivers thesis. SEQH still appears constructive, but wants the market to separate credible 2027 conversion candidates from longer-dated optionality.

What is most real

• The most concrete pieces of the pipeline appear to be the programs with clearer production anchors, especially ALL.SPACE, selected wireless and SATCOM ramps, and the better-defined photonics relationships already tied to engineering and qualification activity.

• These are the opportunities that look closest to true commercial conversion because they have clearer customer identity, more visible unit economics, and more obvious manufacturing pathways.

• In SEQH’s framing, these are the parts of the story that can justify near-term revenue expectations rather than just strategic excitement.

What is promising but risk-weighted

• A second bucket includes programs that are strategically important but still need more evidence before being valued aggressively, especially parts of the CPO, ELS, LiDAR, and foundry-stack narrative.

• These may be real and high value, but they still depend on broader ecosystem timing, customer qualification, architecture choices, and production sequencing that Sivers does not fully control.

• The report seems to argue that these opportunities belong in the valuation as risk-adjusted optionality, not as straight-line revenue assumptions.

What is still too early

• The least mature parts of the pipeline are the ones where the technical relationship may be real, but where timing, customer disclosure, or production scope remain too vague to support confident near-term forecasts.

• SEQH appears especially focused on preventing investors from capitalizing every partnership announcement as if it were already a purchase order.

• The message is that early photonics design-ins can be very valuable, but the path from design relevance to volume revenue is long and uneven.

Why this matters

• The report’s broader purpose is to explain why Sivers can simultaneously be strategically stronger and harder to model cleanly.

• A large pipeline is useful, but only if investors understand which layers are likely to convert in 12 months, which are 2 to 4 year options, and which are still best viewed as technical footholds.

• This deconstruction matters because valuation error in Sivers is likely to come less from missing the total opportunity and more from mistiming the conversion curve.

Read-through for the stock

• SEQH appears to be telling readers that Sivers should still be viewed as a serious multi-vector photonics and wireless platform, but one where discipline around pipeline quality is now more important than enthusiasm around headline size.

• The strongest version of the thesis is not “everything converts,” but that enough of the high-quality buckets convert to justify the strategic premium while the longer-dated layers preserve upside.

• The clean takeaway is that Sivers’ pipeline is real, but not all dollars inside it deserve the same valuation multiple or the same calendar.
  
  
  FULL 26-PAGE PDF WITH EXTENSIVE FORECASTING, MODELING, AND COMPLETE BREAKDOWN OF SIVE PIPELINE ATTACHED BELOW:
  

Read more

WHAT THIS REPORT ARGUES

• SEQH’s core claim is that photonics is no longer just a growth sector but a sovereignty sector, because the West now faces concentrated foreign control across key layers of the optical stack, from substrates and epitaxy to laser modules and packaging.

• The report reframes the market from a normal supply-chain discussion into a strategic-control problem, arguing that “Western photon sovereignty” is becoming a standalone investment lens with direct implications for AI clusters, defense systems, telecom networks, and nuclear-adjacent photonics demand.

Core thesis

• SEQH says the next decade of photonics will be shaped by a simple reality: the West may fund AI factories and defense modernization, but still lacks secure control over several critical optical chokepoints needed to turn those systems on.

• The report’s framework separates photonics into sovereign, semi-sovereign, and exposed layers, then shows that the highest-value segments often still depend on non-Western manufacturing concentration, especially in substrate, compound-semiconductor, and certain module supply chains.

• The conclusion is that companies with genuine Western manufacturing control, auditability, and export-safe provenance could command valuation premiums similar to what happened in uranium, rare earths, and other strategic materials.

What sovereignty means here

• In SEQH’s usage, photon sovereignty does not just mean domestic demand or a Western headquarters. It means traceable control over production, trusted-jurisdiction manufacturing, resilience against export controls, and the ability to serve hyperscaler and defense buyers without hidden geopolitical dependencies.

• The report argues that this matters because optical systems now sit directly inside AI compute fabrics, military sensing, satellite links, and nuclear fuel-cycle technologies, so photonics is moving from component status toward critical infrastructure status.

• That shift is what allows sovereignty to become a pricing and multiple variable rather than just a background narrative.

Stack mapping

• SEQH maps the optical stack across substrates, epitaxy, device fabrication, packaging, engines, and systems, then identifies where Western exposure is weakest and where sovereign scarcity is highest.

• The report places particular emphasis on compound-semiconductor chokepoints, especially in indium phosphide and related laser layers, where a small number of suppliers and jurisdictional bottlenecks can affect the entire downstream AI-optics chain.

• It also connects this stack analysis to earlier SEQH work on Sivers, Coherent, Lumentum, and laser-enrichment photonics, arguing that the real strategic value is often not the end module but the bottleneck layer inside the module.

Company read-through

• The report appears to treat Sivers as one of the clearest examples of sovereign photonics leverage because it combines a Western InP manufacturing footprint with exposure to AI optical engines, CPO external light sources, and defense-linked photonics.

• Coherent is likely positioned as strategically relevant because of breadth across lasers and compound semiconductors, while Lumentum matters as a scale optical supplier but with a different sovereignty profile tied more to system and component leadership than unique jurisdictional exclusivity.

• SEQH’s broader point is that sovereign value accrues most strongly to firms that control a hard-to-replace layer, not merely to any Western-listed photonics company.

Strategic implications

• The report’s larger message is that Western governments and hyperscalers may increasingly pay for assured optical supply, just as they are now willing to pay for assured uranium conversion, enrichment, rare earth separation, and secure semiconductor capacity.

• If that happens, photonics names with sovereign manufacturing could benefit in three ways at once: higher strategic demand, stronger customer stickiness, and premium valuation treatment.

• In SEQH’s framing, photon sovereignty is therefore not a niche political overlay but a new analytical layer sitting on top of the AI optics supercycle.

Bottom line

• The cleanest way to read this note is that AI optics is still the revenue engine, but sovereignty is becoming the multiple engine.

• SEQH is arguing that the market has spent the last year pricing optical bandwidth growth, and may spend the next phase pricing which parts of the optical stack are actually Western, actually controllable, and actually irreplaceable.
  
  FULL 34-PAGE PDF WITH EXTENSIVE ORIGINAL RESEARCH, OUR CUSTOM MODEL OUTPUTS, AND FORECASTING ATTACHED BELOW:
  

Read more

WHAT THIS REPORT ARGUES

• The core thesis is that laser-based uranium enrichment is not mainly a nuclear niche, but an overlooked future demand engine for specialized photonics systems, created by the Western nuclear fuel shortage and the rise of HALEU-hungry advanced reactors.

• SEQH frames this as the largest undiscovered photonics TAM in its coverage universe, tying a $30.5 billion per year Western enrichment market by 2035 to companies across both photonics and nuclear fuel infrastructure.

Core setup

• SEQH says two structural forces are colliding: the U.S. ban on Russian uranium imports, which removes a major source of Western enriched fuel by 2028, and the buildout of advanced reactors that need HALEU, which requires far more separative work than standard LEU.

• In the report’s base math, HALEU demand reaches a point where Western supply is short by about 738 metric tons per year by 2035, while total Western enrichment TAM reaches about $30.5 billion annually at $200 per SWU.

• The key leap is that solving this gap may require commercial-scale laser enrichment, which turns the fuel crisis into a future procurement cycle for high-performance laser systems.

Why lasers matter

• SEQH explains that SILEX-type enrichment works by exploiting the tiny absorption difference between 235235UF66​ and 238238UF66​ molecules, using mid-infrared light near 16 microns to selectively excite the target isotopologue.

• That wavelength requirement is what makes this a real photonics problem: standard fiber lasers and telecom InP lasers do not fit, while the most credible paths are tunable CO2 systems with Raman conversion or potentially future QCL-based architectures.

• SEQH’s conclusion is that the first commercial laser enrichment plants could trigger one of the largest industrial procurements of specialized laser equipment ever, even though almost no photonics analyst models that today.

TAM and economics

• The report’s enrichment waterfall breaks 2035 demand into roughly $15.0 billion of replaced Western LEU demand, $8.0 billion of HALEU demand for advanced reactors, $2.5 billion of DOE strategic reserve and government use, and $5.0 billion of allied export demand, summing to $30.5 billion.

• SEQH argues laser economics have crossed from speculative to urgent because spot SWU pricing reached about $200, versus academic laser-enrichment cost estimates of roughly $30 to $60 per SWU, with first-generation commercial systems still attractive even at $80 to $100 per SWU.

• In that framework, a 6 million SWU per year laser facility would generate around $1.2 billion of annual enrichment revenue at current pricing, and the embedded photonics capex and maintenance demand is largely unmodeled by the Street.

Ecosystem map

• SEQH identifies four active laser-enrichment programs as of June 2026: GLE / SILEX, QLE / ASPI, LIS Technologies, and an AVLIS revival path tied to lithium rather than uranium.

• GLE is the sector’s anchor proof point because it reached TRL-6 in October 2025 and has a Paducah commercial license application in review, while QLE / ASPI is the only U.S.-listed public equity where laser enrichment is the core equity story.

• LIS Technologies is positioned as the key U.S.-origin patented alternative, with a planned $1.38 billion Oak Ridge investment and 5.5 million SWU per year target, while NNE matters because it is the only public advanced reactor developer with a disclosed DOE-linked relationship to a laser-enrichment program.

Company read-through

• Within photonics, Coherent is judged the most relevant public company because it has the broadest laser portfolio and the closest commercial adjacency through CO2, fiber, and DPSS systems, although its 10.6 micronCO2 base still needs engineering work to reach the ~16 micron enrichment requirement.

• ASPI / QLE is the direct integrated enrichment equity, BWXT is the downstream beneficiary regardless of which enrichment technology wins, and OKLO plus NNE are treated as HALEU demand catalysts rather than photonics beneficiaries.

• SEQH is explicit that Sivers and Lumentum do not directly fit the enrichment-laser physics: Sivers’ InP platform operates at 1270 to 1650 nm, and Lumentum’s relevance is only indirect through industrial lasers and pump-module adjacency.

Main conclusions

• SEQH’s five-part conclusion is that laser enrichment has now crossed the technical threshold to be commercially credible, SWU pricing has structurally improved the economics, photonics analysts still do not model the TAM, COHR is the most relevant photonics name, and BWXT benefits downstream no matter which enrichment platform wins.

• The broadest implication is that photonics and nuclear fuel will converge when the first commercial laser-enrichment procurement contracts are placed, because enrichment capacity then becomes a direct buyer of advanced laser systems rather than just a nuclear policy topic.

• The biggest caveat is also central to the note: exact SILEX laser specifications are classified, so SEQH’s photonics TAM is a first-principles framework rather than a disclosed procurement model.

FULL 33-PAGE PDF REPORT INCLUDING EXTENSIVE SECTOR MODELING, FORECASTING, VALUATION METRICS, AND MORE AVAILABLE BELOW:

Read more

WHAT THIS NOTE SAYS

• SEQH reaffirms OVERWEIGHT / High Risk on Sivers, but the framing has changed: the stock is no longer a cheap optionality story and is now an expensive but better-validated AI photonics and mmWave platform.

• The June update argues that the thesis improved fundamentally through a $799 million opportunity pipeline, GlobalFoundries reference-design validation, and an $8.2 million ALL.SPACE production order, even as Q1 results, cash burn, and PCAOB-related restatements made the accounting and risk profile less forgiving.

What changed

• Since the April note, the stock moved from SEK 28.36 to SEK 67.00, a gain of about 136 percent, despite also suffering a 39 percent drawdown from the June 3 peak of SEK 110.00.

• Q1 2026 revenue came in at SEK 61.9 million, down 22 percent year over year, and adjusted EBITDA was -SEK 13.8 million, which SEQH attributes mainly to U.S. government shutdown and defense timing plus FX, not to a thesis break.

• The more important change was commercial validation: the pipeline expanded from about $453 million to $799 million, GlobalFoundries adopted Sivers laser arrays for SCALE optical-engine reference designs, and ALL.SPACE placed a named $8.2 million production order for 2027.

Core thesis

• SEQH’s central view is that InP laser scarcity is now strategic, not thematic, with lasers, substrates, and yields increasingly acting as real bottlenecks in AI optical scaling.

• Sivers is still attractive because it has multiple independent revenue vectors rather than one product bet: automotive LiDAR, pluggables and LRO lasers, CPO and NPO external light sources, SATCOM terminals, Tier-1 FWA, and defense arrays.

• The report says 2027 is the real conversion year, when investors should expect the debate to shift from whether the technology is credible to whether programs actually turn into manufacturable volume.

Photonics

• Q1 photonics revenue was SEK 17.8 million, down 32 percent year over year, with segment EBITDA around -SEK 7.6 to -7.7 million, so near-term reported numbers were weak even as strategic positioning improved.

• SEQH sees the GlobalFoundries release as the highest-quality new photonics data point since April because it puts Sivers inside a silicon-photonics reference-design context, not just a standalone component evaluation.

• The Glasgow plus WIN manufacturing setup remains the core architecture: Glasgow provides owned InP process control and qualification credibility, while WIN gives the volume path needed for 2027 and beyond.

• In SEQH’s base case, photonics revenue rises from SEK 93 million in FY25 to SEK 140 million in FY26, SEK 335 million in FY27, and SEK 850 million in FY30, while the bull case reaches SEK 2.12 billion by FY30.

Wireless

• Wireless remains the near-term revenue engine, with Q1 wireless sales of SEK 44.1 million, though EBITDA was still -SEK 10.5 million and the mix stayed heavily weighted toward NRE revenue rather than recurring hardware.

• The most important new wireless proof point is the ALL.SPACE $8.2 million 2027 production order, which turns the SATCOM thesis from forecast dependency into a named production anchor.

• Other updates, including the Tachyon $1.5 million 60GHz development partnership, Year-2 Microelectronics Commons $6.6 million funding, and the Tier-1 telecom FWA track for end-2026, reinforce the idea that wireless could become the cleaner validator of the 2027 revenue bridge.

Financial reset

• The annual report restatement changed the quality of the accounting base more than the revenue base: FY25 revenue moved to SEK 306.6 million from SEK 304.1 million, but adjusted EBITDA reset to -SEK 50.3 million, EBIT to -SEK 177.8 million, and reported equity to SEK 949.8 million.

• Q1 operating cash flow was -SEK 49.2 million, and cash at March 31 was just SEK 26.6 million before the SEK 125 million directed issue, so cash burn remains one of the central risks.

• SEQH’s updated base case now models group revenue at SEK 390 million in FY26, SEK 720 million in FY27, SEK 1.05 billion in FY28, and SEK 1.75 billion in FY30, with a bull path to SEK 3.05 billion by FY30.

Valuation and targets

• At SEK 67.00 and 319.95 million registered shares, the headline equity value is about SEK 21.44 billion, equal to roughly 70x FY25 sales, 55x FY26E base sales, and 30x FY27E base sales before cash adjustments.

• SEQH’s updated 12-month framework is SEK 35 bear, SEK 82 base, SEK 145 bull, and a probability-weighted target of SEK 86, which implies about 28 percent upside from the June 25 close.

• The note is explicit that the market has already capitalized much of the strategic narrative, which is why the base case upside is now more moderate even though the underlying business validation improved.

Risks and what matters next

• The biggest risk is no longer that the thesis is obscure, but that the stock now discounts several successful ramps before the P&L has proved them.

• SEQH highlights valuation compression, cash burn, PCAOB and U.S. listing timing, execution across LiDAR, SATCOM, CPO, FWA and defense, customer concentration, Achilles / DDM overhang, short interest, and dilution as the core risk set.

• The next checkpoints are the August 6, 2026 H1 report, Tier-1 FWA product milestones in H2 2026, Q4 2026 LiDAR readiness, possible Nasdaq New York filing activity, and 2027 qualification progress across Jabil, GlobalFoundries, POET, and Ayar.

The Full PDF report with deeper insight, valuation metrics, and further outlook projection is available below:

Read more

WHAT THIS NOTE SAYS

• This report argues that Sivers has moved from being viewed as a standalone InP fab to a designed-in light-source layer inside the AI datacenter optics stack, with the June 2, 2026 GlobalFoundries collaboration as the key catalyst that changes the quality of the story.

• SEQH frames the new value not as a firm contract but as a capacity option, then adds that option layer plus a pipeline-conversion layer to prior Sivers work and arrives at a platform value around SEK 150, versus a spot price of SEK 86.45.

Core thesis

• The central idea is that being embedded in a tier-one foundry’s silicon-photonics reference designs can pre-qualify Sivers for future CPO and LPO volume without Sivers having to fund the foundry capex itself.

• SEQH stresses that the GlobalFoundries arrangement carries no disclosed volume, revenue, or exclusivity commitments, so the correct valuation frame is not backlog but a real option on future foundry-scale optical-engine volume.

• The note says this does not replace the earlier Sivers thesis around InP scarcity and sovereignty, but extends it by adding a Foundry-Stack Capacity Option and a pipeline-conversion uplift.

Why the foundry stack matters

• In silicon photonics, silicon can modulate and route light, but it still needs a III-V laser source, usually indium phosphide, supplied from outside the silicon die.

• As optics move from pluggables toward co-packaged optics, the external light source becomes a qualified part of the foundry reference design, and that is exactly the layer where Sivers sits.

• SEQH highlights GlobalFoundries as the most important anchor because it is presented as one of the largest pure-play silicon-photonics foundries, with photonics revenue moving from about 200 million dollars in 2025 toward 400 million dollars in 2026 and a 1 billion dollar run-rate exiting 2028.

Design-win sequence

• The rerating is framed as cumulative rather than single-event. SEQH’s ledger includes ESA SATCOM, U.S. defense mmWave, a LiDAR ramp, O-Net and Enablence 8-channel ELS, Jabil 1.6T LRO, the POET collaboration, the Ayar Labs 16-wavelength WDM path, and finally the GlobalFoundries SiPh / SCALE embed.

• The key pattern is that the wins now span defense, space, LiDAR, pluggable optics, and CPO, which reduces reliance on any one program and broadens the platform narrative.

• At the same time, SEQH is careful to note that these are still mostly design wins and qualification events, not committed-volume supply contracts, which is why the valuation uses risk-weighted conversion assumptions.

TAM and valuation

• SEQH sizes the light-source slice by starting from large optical TAMs, including 50 billion dollars plus for pluggable optics, 73 to 100 billion dollars for AI-cluster optics, 4.7 to 15 billion dollars for CPO hardware, and 1.5 billion dollars plus per year for ELSFP laser modules.

• Applying a 10 to 15 percent laser-content share to those module markets, the note estimates roughly 8 billion dollars of addressable laser and ELS content by 2030 that could sit within a tier-one foundry stack.

• Method 1, the sum-of-layers capacity-option build, starts from a rebased prior platform core of SEK 36.7 per share, then adds SEK 102.8 for the Foundry-Stack Capacity Option and SEK 14.0 for pipeline conversion, reaching SEK 153.

• Method 2, the forward EV/Sales peer cross-check, gives SEK 129 using a 2028 revenue anchor and SEK 150 using a 2030 revenue anchor, so the triangulated range comes out to roughly SEK 144 to 150, which is why SEQH frames the platform value at about SEK 150.

Key assumptions

• The dominant valuation layer assumes that of the 8 billion dollar 2030 light-source pool, Sivers can capture about 9.5 percent share with a 55 percent probability that designed-in positions convert into actual volume.

• That produces about 418 million dollars of option-case revenue, which SEQH capitalizes at a discounted forward 10x EV/Sales to derive the largest single piece of the valuation.

• The separate pipeline-conversion layer uses the 799 million dollar opportunity pipeline, an 18 percent blended win rate, and a 35 percent durable-revenue annualization, deliberately kept modest to avoid double counting the larger capacity option.

Financial position and risk

• On the financial side, FY2025 revenue was SEK 304.1 million, gross margin was 87.2 percent, operating income was SEK -141.3 million, and net income was SEK -186.5 million, showing that the company still sits firmly in an investment phase.

• Q1 2026 sales were SEK 61.9 million, down 22 percent year over year, while the opportunity pipeline expanded to about 799 million dollars, up 77 percent year to date.

• SEQH also flags material overhangs: the stock has risen roughly 28-fold from its February low, the valuation implies extreme future-scale assumptions, losses are widening, and the note explicitly highlights short-seller pressure, market-conduct scrutiny, governance risk, dilution risk, and competitive displacement as real threats to the thesis

• In SEQH’s own downside framing, if flagship platforms slip, GF ramps move out, financing becomes punitive, or governance confidence is impaired, the valuation could compress back toward the SEK 37 to 60 commercial core range.

Bottom line

• The conclusion is that Sivers now deserves to be analyzed less as a niche component maker and more as a levered option on foundry-scale AI optics volume, because being designed into the stack is strategically better than simply selling merchant laser arrays.

• But SEQH is explicit that SEK 150 is a contingent platform value, not a recommendation and not a contract-backed base case, so the upside case depends heavily on whether those designed-in positions actually convert into durable volume over 2027 to 2030.
  
  FULL 15 PAGE REPORT LOCATED BELOW:
  

Read more