Sivers Semiconductors

The source thesis gives us a set of questions to investigate. It does not supply the answer to all of them. This is our analytical framework, rather than a claim about Serenity's exact language or current portfolio.

The central debate: does Sivers become an economically attractive volume supplier, or does it remain a technically relevant business that requires substantial capital before shareholders see cash? This is the bridge between a compelling supply-chain narrative and an investment conclusion.

Our burden of proof: recurring deliveries and product economics must substantiate the growth story; valuation must then leave adequate compensation for delay, financing and competition. An insightful source can generate an excellent research question even when the share price already discounts a favorable answer.

The issuer's Q2 results announcement describes approximately SEK 825 million of gross directed issues during the quarter, including the large issue resolved at quarter end, and subsequent conversion of the USD 12 million loan. A resolution, settlement, debt conversion and cash receipt are distinct events. Gross proceeds also differ from usable net cash.

The CEO's August letter says the remaining USD 5 million term loan was repaid in August. Treat that as a dated management statement; it is not a verified 20 October cash balance or proof that every lease or other obligation disappeared.

We do not compute a current runway or enterprise value from the June cash balance plus headline fundraising amounts. The missing bridge is: opening cash, settled net proceeds, cash spent, loan repayments, subsequent working capital and remaining obligations.

The 29 September EGM notice states 356,740,332 issued ordinary shares, including 12,872,916 held by the company. That implies 343,867,416 ordinary shares outside treasury on that date. This calculation is not a fully diluted forecast or confirmation of the count today.

The Q2 report presents FY2025 sales of SEK 302.8 million. The following is our illustrative arithmetic, using September shares outside treasury and that historical sales figure. Prices are hypothetical, not current quotations. Equity value / sales is not enterprise value / sales.

Formula: share price × 343,867,416 / 302,800,000. These examples show why a promising component story needs a demanding valuation examination. They do not identify fair value. The next model must connect current enterprise value to achievable revenue, operating margins, reinvestment, dilution and the time before shareholders receive cash. A comparison with a much larger semiconductor company is not a substitute for that bridge.

We have not reconstructed Serenity's full original post history. Therefore “most recommended,” exact mention counts, current conviction and current holdings remain unverified. The tracker describes SIVE as a leading conviction idea; that characterization is attributed to the tracker, not an independently established ranking. An original-post audit would need a complete collection window and a definition separating recommendations, mentions, replies and duplicates.

We also checked Nanalyze's August SIVE discussion as a counterpoint. Its conclusions are secondary commentary; we do not adopt unverified allegations or historical performance claims. The business case is tested against operating evidence rather than the popularity of either a bullish or bearish commentator.

• Independently verified current quote with exchange, currency and timestamp.
• Updated issued, treasury and fully diluted share counts after subsequent actions.
• A reconciled current cash/debt/capex bridge and valuation model.
• Counterparty confirmation of material programs where publicly available; company announcements are attributed to the issuer.
• Customer concentration, production margins, yields and firm order values where not disclosed in the reviewed material.
• A reproducible audit of the original Serenity claims.

Update 2026-10-22: the current quote, share count, cash position and a scenario valuation are now recorded in the dated valuation section below; customer concentration and the Serenity post audit remain open.

These gaps limit the investment conclusion. They do not prevent following a measurable business thesis. The next report update should compare actual shipments, financial results and financing with these checkpoints rather than merely repeating the AI-photonics narrative.

Added 2026-10-22 · Content updated 2026-10-22 · Evidence reviewed 2026-10-22 · Source 2026-04-15 · Review after 2026-11-26

Added: 2026-10-22. Content updated: 2026-10-22. Evidence reviewed: 2026-10-22. Review checkpoint: 2026-11-26. All dollar figures are industry estimates (est.), not Sivers disclosures; operators are speculated.

On the web page this section opens with a responsive HTML diagram; the static export is available as a cutaway poster (PNG).

Money flow for one campus (est.): $40B operator spend → $18 to 22B to racks (6,000 to 7,500 racks × $3M, 1 GW facility at PUE 1.1 to 1.3) → about $3B to transceivers (~500k GPUs × 3 to 4 modules × $1,500 to 2,000, both link ends plus fat-tree tiers) → $25 to 60M to Sivers lasers (2 dies × ~$15 per module × 1.5 to 2M modules, if Sivers supplied every laser; point estimate ~$45M). Assumptions: ~130 kW per rack, 3 to 4 scale-out modules per GPU, no co-packaged optics. One campus fully won is therefore worth one to two years of Sivers' FY2025 net sales of about SEK 307M (~$32M).

Sources: SemiAnalysis GB200 architecture (rack power, optics per GPU); NBER working paper 35865 and Bloomberg Law on the Meta Texas campus (campus cost with chips); Ascent Optics 1.6T guide and FS.com 1.6T listings (module prices); Credo Oz8 and OpenLight DR8 PIC (lasers per module); 800G module BOM discussion (die price).

Added 2026-10-22 · Content updated 2026-10-22 · Evidence reviewed 2026-10-22 · Source 2026-06-04 · Review after 2026-11-26

Added: 2026-10-22. Content updated: 2026-10-22. Evidence reviewed: 2026-10-22. Review checkpoint: 2026-11-26. Revenue figures per competitor are analyst estimates for 2026 datacom InP lasers unless marked otherwise.

On the web page this section opens with a responsive HTML diagram; the static export is available as a competitor map poster (PNG).

On the web page this section opens with a responsive HTML diagram; the static export is available as a market-share poster (PNG).

TrendForce (June 2026) puts Broadcom, Lumentum and Sumitomo at about 55% of EML plus CW-DFB supply, and Broadcom, Sumitomo, Coherent and LandMark at about 74% of CW-DFB capacity, with Lumentum's EML capacity booked through 2027. TrendForce reports capacity (50.7M dies a month) rather than dollars. At $8 to 10 per die that is roughly $3B a year, which would make Sivers' share about 0.3%, lower still for datacom alone.

The market size is the weakest input. The only published dollar series is Future Market Insights: $579M for CW-DFB chips (2025, all uses) growing to about $1B by 2030. The $3B figure is unit-derived from TrendForce. The scenarios below therefore show the share Sivers would need under both denominators.

Added 2026-10-22 · Content updated 2026-10-22 · Evidence reviewed 2026-10-22 · Source 2026-08-27 · Period end 2026-06-30 · Review after 2026-11-26

Added: 2026-10-22. Content updated: 2026-10-22. Evidence reviewed: 2026-10-22. Review checkpoint: 2026-11-26. This analysis, not company guidance. The published consensus is a single house (Redeye via MarketScreener): group sales SEK 273M (2026), 458M (2027), 621M (2028); Fintel shows 689M for 2028. Consensus implies photonics of roughly $30M in 2028.

On the web page this section opens with a responsive HTML diagram; the static export is available as a scenarios poster (PNG).

Implied growth from the SEK 93.4M (~$9.8M) 2025 base over five years: bust 1.8× (~12% a year), base 9× (~55% a year), boom 34× (~100% a year). The photonics run-rate in H1 2026 was SEK 33.7M, about SEK 67M annualised, so every scenario starts from a lower point than 2025.

What decides the case:

• Share, not market size. Going from 0.3% to 3% is the whole base case against suppliers that hold most CW-DFB capacity.
• CPO versus pluggables. In a pluggable the Sivers die is a $15 commodity. Analyst estimates put ultra-high-power DFB lasers at $50 to 100 each with about eight per external light source, or $300 to 600 of laser content per GPU under co-packaged optics versus $30 to 60 in pluggables (Mizuho via secondary, siliconmatter). No Ayar Labs SuperNova price is public. Ayar's volume qualification is targeted for H2 2027, so none of it appears before 2028. The boom case is a CPO case.
• Capacity ceiling. The Q2 2026 CEO letter says the expanded Glasgow facility is functional by end-2027 and targets a 1:2 ratio of internal to partner-foundry capacity. Revenue above a few tens of millions of dollars in 2027 depends on the partner line.
• LiDAR. The 20 October 2026 release confirms production orders: "Sivers will supply the continuous-wave distributed feedback (CW DFB) lasers and optical amplifiers", shipments from Q4 2026, production in 2027, "estimated cumulative revenue potential of USD 53 million to USD 138 million over the customer's product lifecycle" (programme first announced March 2026). This supersedes the USD 28 to 53 million cumulative 2026 to 2030 figure in the Q4 2025 report used in earlier drafts. The scenarios' LiDAR-and-sensing lines ($10 to 20M a year) sit inside that lifecycle range; the first order value is not disclosed.
• Pipeline. The $1.2B opportunity pipeline (July 2026) is non-binding, covers 2026 to 2030 and both divisions, and was $453M at end-2025 (Q4 2025 report).

Additional sources: LightCounting, May 2026; Photoncap on the light-source market; Lumentum Q2 FY26; SEQH Research; Ayar Labs and Jabil SuperNova sled.

Added 2026-10-22 · Content updated 2026-10-22 · Evidence reviewed 2026-10-22 · Source 2026-10-22 · Review after 2026-11-26

Added: 2026-10-22. Content updated: 2026-10-22. Evidence reviewed: 2026-10-22. Review checkpoint: 2026-11-26. This resolves the current-quote, share-count and cash items listed as open on 20 October, with the limits stated below. FX est. at EUR 11.0 and USD 9.5 SEK.

On the web page this section opens with a responsive HTML diagram; the static export is available as a valuation poster (PNG).

Scenario values. 2030 group revenue = photonics scenario + ~$45M wireless (SEK ~430M, extrapolated from consensus 2028); EV at 6× and 12× sales; plus $0.08B cash; divided by 355M shares; discounted four years at 12% (×0.64). No further dilution assumed.

Reading: at SEK ~23 the market already prices the base case at a Lumentum-like multiple. The consensus path alone does not support the price. The boom case offers two to four times, but needs CPO light sources at volume from 2028 and the partner foundry to deliver, and would likely need a larger raise or the photonics listing. Still open after 22 October: the first LiDAR order value and 2027 run-rate (lifecycle USD 53 to 138M is not a one-year forecast), customer concentration, Q3 cash, and a second analyst house. Carnegie, Pareto and Modular Finance estimates were not obtainable; "consensus" here is one house. The byNordic SPAC letter of intent (Aug 2024, paused Nov 2024) never disclosed a valuation.

Added 2026-10-22 · Content updated 2026-10-22 · Evidence reviewed 2026-10-22 · Source 2026-10-22 · Period end 2026-06-30 · Review after 2026-11-26

Added: 2026-10-22. Content updated: 2026-10-22. Evidence reviewed: 2026-10-22. Review checkpoint: 2026-11-26. Captured from IBKR Mobile on a read-only login between 12:48 and 12:55 Europe/Luxembourg (+02:00). Every row carries the as-of date IBKR showed and the capture timestamp of the screenshot it was read from. Third-party data (Morningstar, LSEG, TipRanks) via IBKR, not guaranteed by IBKR. Morningstar fields carry the "Q" marker: quantitative model output, not an analyst-written report.

How this squares with the rest of the page: the LSEG single analyst is consistent with the Redeye figures used in the valuation section (SEK 7.4 fair value, revenue 273/458/621M), so "consensus" on this page is one house seen through two data vendors. Morningstar's SEK 16.23 is a quantitative model value with very high uncertainty, not a research call. The one TipRanks Buy is the January 2026 Northland record already listed under analyst records.

Not captured: no live Stockholm quote (price, volume, 52-week range) on the read-only login; no short-interest or securities-lending data for this listing; no events or earnings-calendar tab in the Company layout; the Morningstar "Full Report" PDF was not downloaded; the 2DG (Xetra) line was not opened. The 52 capture screenshots are listed in the image manifest; four are embedded above.

Added 2026-10-22 · Content updated 2026-10-22 · Evidence reviewed 2026-10-22 · Source 2026-08-27 · Period end 2026-06-30 · Review after 2026-11-26

Added: 2026-10-22. Content updated: 2026-10-22. Evidence reviewed: 2026-10-22. Principal source published: 2026-08-27. Review checkpoint: 2026-11-26.

Who runs the company. Vickram Vathulya is not a founder; Sivers dates to 1951. He joined from outside as President and CEO on 19 August 2024 after a search by Hunt Scanlon (appointment). He replaced Anders Storm, CEO since 2016, who resigned on 26 March 2024 "for new challenges outside the Company" with no further reason. Chair: Bami Bastani. After the 15 June 2026 AGM the board is Bastani, Todd Thomson, Karin Raj, Joakim Nideborn and Helena Svancar; Erik Fällström (largest holder, 13.86% at end-2025), Tomas Duffy and Keith Halsey left. CFO Heine Thorsgaard since September 2025.

Track record. Philips/NXP (1999 to 2008), then twelve years at Maxim Integrated ending as VP and GM of Core Products, then President of Nuvotronics, a millimetre-wave subsidiary of Cubic, from November 2022 to August 2024. A divisional operator at a large analog house; no prior public-company CEO role, no photonics background, no controversies found. No independently verifiable outcome from Nuvotronics.

Promises versus delivery.

Pattern: hard milestones move one or two quarters right each letter while the non-binding pipeline figure ($453M → $799M → $1.2B) is the headline.

The NRE question, with the numbers (SEK m, restated).

All NRE is Wireless; Photonics carries no NRE after the restatement, so "shifting from NRE to product" is a Wireless story and says nothing about AI-optics demand. 2025 growth was NRE doubling on US defence programmes (three Wireless customers were 60% of revenue) while hardware fell. Contract liabilities (customer prepayments for new development work) fell from SEK 32.3M to 11.6M during 2025 while contract assets (NRE recognised but unbilled) rose from 54.7M to 100.4M: fewer new prepaid contracts and more unbilled work in progress, which is what a development pipeline running off looks like. The Q2 "product revenue +18%" headline is Wireless hardware; photonics hardware fell 8% in Q2 and 23% in H1. Sources: annual report 2025 Note 6, Q2 2026 report p. 14, Q1 2026 report p. 13.

Cash and dilution. Cash SEK 43.5M at end-2025 (operating cash flow −57M), 26.6M at end-Q1 2026, 62.7M at end-Q2 before the SEK 700M raise closed (Q2 operating cash flow −70M). Headcount 130 (June 2026).

Insider alignment. Vathulya holds about 4.54M shares (~1.3%) plus 3.7M options, committed at hiring to buy 2M shares, and bought 70,000 more after the lock-up on 21 July 2026 at about SEK 38 to 42: a net buyer above today's price. Chair Bastani sold 275,000 shares the day the lock-up expired (16 July 2026) and director Thomson's Headwaters sold 950,000 by 22 July (lock-up release). A circulated report that the CEO sold 1.23M shares was false (Boerse Express). 2025 pay SEK 5.1M, no bonus.

Red flags.

• May 2026 restatement: 2024 net sales cut from 243.7 to 219.2M (−10%), 2024 net loss widened from −116.3 to −183.9M, 2025 EBIT from −141.3 to −177.8M, equity cut by SEK 127M. Causes: NRE recognised too early, hardware timing, distributor returns, inventory write-ups, a loss-making NRE contract, understated amortisation and a SEK 28.4M impairment of capitalised Wireless development. Framed as a PCAOB audit uplift; these are IFRS errors in the core revenue model, under the current CEO for 2025.
• Share count up ~32% in 20 months; the 22 October 2026 EGM votes on a new option plan and Series C shares.
• Dagens Industri, 31 Aug 2026 called the Q2 report catastrophic and said sell; the stock fell 15% that day and 27% over seven sessions.
• Share price under his tenure: low SEK 1.76 (13 Nov 2024), high SEK 110 (3 Jun 2026), about SEK 23 to 24 now (TradingView).

Balanced verdict. For: real Wireless wins (ALL.SPACE $8.2M, Tachyon $3M, CHIPS Act programmes), genuine photonics partner names (Jabil, GlobalFoundries, POET, O-Net), a SEK 700M raise executed near the top, no debt, and the CEO buying with his own money while the chair and a director sold. Against: the group has not grown once restated; the AI division shrank 23% in H1 2026; the Ayar ramp promise vanished; the SPAC never closed; the restatement found NRE revenue overstated; new development bookings are drying up; guidance slides right every quarter. He reads as a credible divisional operator with aligned incentives, not a proven public-company CEO, and the letters run ahead of the numbers. The LiDAR production orders of 20 October 2026 are the first delivered milestone of the Q2 letter and count in his favour; the first hard numerical test is the Q3 report on 26 November 2026 and whether LiDAR shipments and Jabil orders show up in Q4.

Added 2026-10-22 · Content updated 2026-10-22 · Evidence reviewed 2026-10-22 · Source 2026-10-22 · Review after 2026-11-26

Added: 2026-10-22. An adversarial check of the diagram and scenario inputs found these errors in earlier drafts; the published figures above are the corrected ones.

Diagrams on the web page are responsive HTML built from the tables in this report; the PNG posters linked under each are static exports of the working design canvases and may clip a footer line.