Sivers Semiconductors AB (SIVE) — Investment Tree v1

🔴 SPECULATIVE / FRAGILE-THESIS — DEFAULT VERDICT: AVOID (not a 5-10 year hold). This is a pre-profit, cash-burning ~$32M-revenue Swedish photonics small-cap whose stock is up ~1,800% YTD on an AI-photonics narrative, now carrying a live Swedish insider-trading/market-abuse probe, a restatement that CUT revenue and WIDENED losses, a credible short report (Ningi, 2026-06-01) alleging ~31% of FY2025 revenue is dubious, US securities class-action investigations, and insiders (incl. the head of the wireless segment) and a directed-issue fund selling their ENTIRE stakes into the rally. Durability 8/25 with 3 fatal flags = UNOWNABLE for a long-term hold. The InP-laser efficiency edge and the GlobalFoundries/Jabil partnerships are real but are collaborations/reference designs, not committed volume supply contracts. At ~SEK 25bn (~$2.7bn) / ~80x trailing P/S the price embeds the bull tail as the base case. EV ~−47%; asymmetry UNFAVORABLE; the credible asymmetric trade here is the short, not the long. Source quality: MIXED Tier B/C throughout — small-cap, restated accounts, company site blocks automated fetch; figures triangulated from primary-filing summaries via PRNewswire/Cision/aggregators. Flagged † where Tier C. R2 (primary-filing verification of segment splits, gross margin, exact share count, ADR mechanics) is owed before any change of stance — though no plausible R2 finding makes this a long-term hold. R2-VERIFIED (2026-06-08): the allegation-grade claims are independently corroborated across multiple sources — the Ningi short report (ningiresearch.com, 2026-06-01; alleges ≥SEK 97M / ~31% of 2025 revenue is dubious), the Ekobrottsmyndigheten + Finansinspektionen insider-trading probe (state prosecutor Jonas Myrdal: pattern "striking… hallmarks of pump-and-dump"), board director Erik Fällström's Achilles Capital selling its entire 29M-share stake for ≥SEK 461M while calling the stock "significantly undervalued," and US class-action investigations by Rosen + Bronstein Gewirtz & Grossman. Q1 2026 net sales SEK 61.9M (−22% YoY). The AVOID verdict is REINFORCED, not changed. (The "~1,800% YTD" rally is ~1,200–1,800% depending on the start date/window — order-of-magnitude pump confirmed; ATL SEK 1.75 → ATH SEK 110.) See evidence_2026-06-08.jsonl (SIVE-R2-2026-06-08-018).

Stage 7 final essay. Bilingual companion: tree_v1_zh.md. Date: 2026-06-08 · Anchor price: SIVEF $7.77 (US OTC) / SIVE.ST SEK 78.80 (both 2026-06-05 close; ATH SEK 110 on 2026-06-03) · SEK/USD ~9.42 · Market cap ~SEK 25bn / ~$2.68bn · ~320M shares · Cash YE2025 SEK 43.5M; Q1 burn ~SEK 49M/qtr — runway ~2-3 quarters post-May raise · Dividend: none Archetype: pipeline-promise photonics cash-burner riding an AI-attached retail pump — closest analogs PLTR (speculative/fragile) + POET (pre-scale photonics)

I. One-sentence verdict

Sivers is a genuine InP-laser foundry with a real (but narrow) efficiency edge and two real (but non-binding) AI-optics partnerships — wrapped in a ~$32M-revenue, perpetually-loss-making, serially-diluting shell that has been bid up ~1,800% YTD into a ~$2.7bn / ~80x-sales valuation, and is now simultaneously fighting a live insider-trading probe, a revenue-cutting restatement, a credible 31%-dubious-revenue short report, US class-action investigations, and visible insider exit-selling; with durability 8/25 (3 fatal flags) and an expected value of roughly −47%, the only defensible action for a 5-10 year holder is AVOID — the asymmetry runs against the long, and the credible trade is the short.

II. Company snapshot

Sivers Semiconductors AB (publ) is a Swedish photonics + RF semiconductor company (HQ Kista; roots to 1951) organized in two segments: Photonics (Sivers Photonics, Glasgow III-V/InP foundry from the 2017 CST Global acquisition — InP DFB lasers, gain chips, PICs, silicon-photonics enablement; the AI-datacenter CPO/LPO/LRO + LiDAR + SATCOM story) and Wireless (Sivers Wireless incl. 2022 MixComm — fabless RF/mmWave beamforming ICs; 5G/FWA/SATCOM). Primary listing Nasdaq Stockholm (SIVE.ST, SEK); US OTC ordinary (SIVEF, USD).

The company is pre-profit and cash-burning. Restated FY2025 (published ~2026-05-15 after a PCAOB audit uplift for a potential US dual listing): net sales SEK 306.6M (~$32M), EBIT SEK -177.8M, net loss SEK -222.6M. The restatement CUT revenue and WIDENED losses versus original figures (and cut FY2024 revenue -10% while widening FY2024 loss +58%). Cash at YE2025 was just SEK 43.5M against a Q1 2026 operating cash burn of SEK -49.2M — one quarter of burn exceeded the entire cash balance. Q1 2026 net sales were SEK 61.9M (-22% YoY), hit by the US government shutdown / defense-budget delays and unfavorable FX.

The dominant market view (the retail momentum bid) is that SIVE is an InP-laser "chokepoint" for the AI-optics supercycle, validated by GlobalFoundries (2026-06-02) and Jabil (OFC 2026) partnerships and a $799M "opportunity pipeline" (+77% YTD) — and that today's price is early entry into a multi-bagger. The opposing view (credible short sellers; short interest 1.6%→17%) is that this is "a retail-driven pump built on speculative hyperscaler relationships, a fabricated bottleneck narrative, a rumored volume ramp-up, a delayed US listing, and a vague opportunity pipeline."

The H-0 thesis names what the price implicitly embeds: that the pipeline + partnerships convert to a transformational 2027+ volume ramp at >50% gross margin, and that the company survives the financing gap and clears the integrity overhang — all at once. This tree tests that conjunction and finds it leaning falsified.

III. The Tree

H-0: SIVE is a merchant InP-laser chokepoint for the AI-optics supercycle (+ LiDAR/SATCOM);
     the $799M pipeline + GF/Jabil partnerships convert to a transformational 2027+ ramp at
     >50% GM, and the company survives without value-destroying dilution → today's price is
     early entry into a 5-10 year compounder.
│
├── L1A — Photonics / AI-optics ramp  ⚠️C (real tech, UNCONVERTED)
│   ├── 1.1 $799M pipeline converts to committed revenue 2026-28   ⚠️C revenue actually -22% Q1; 25x gap
│   ├── 1.2 GF/Jabil partnerships → committed VOLUME purchase order ⊗ reference designs only, no PO
│   └── 1.3 CPO/LPO/LRO adoption timing supports a 2027 ramp        ⚠️C wave real, Sivers' capture unproven
│
├── L1B — Wireless / SATCOM / LiDAR  ⚠️C
│   ├── 2.1 LiDAR customer ramps Q4 2026 at material scale          ⚠️C $53-138M LIFETIME — small vs mcap
│   ├── 2.2 SATCOM/wireless 2027 ramp materializes                  ⊗ 2027-dated, no orders
│   └── 2.3 Defense exposure is high-quality recurring revenue      ✗B budget delays CUT Q1; quality questioned
│
├── L1C — Financing / runway / dilution  ✗B (BINDING — negative)
│   ├── 3.1 Adequate cash runway                                    ✗B SEK 43.5M vs SEK 49M/qtr burn
│   ├── 3.2 Dilution is bounded / non-destructive                   ✗B serial diluter; 53.8M more authorized
│   └── 3.3 Credible path to self-funding before per-share crush    ⚠️C requires the unproven 2027 ramp
│
├── L1D — Accounting integrity & governance  ✗B (BINDING — negative)
│   ├── 4.1 Revenue recognition is clean                            ✗B own restatement cut revenue/widened loss
│   ├── 4.2 Insiders aligned with long-term holders                 ✗B segment head + fund sold ENTIRE stakes
│   └── 4.3 Short thesis (31% dubious revenue) is wrong             ⚠️B partly corroborated; probe LIVE
│
├── L1E — Moat / IP / competitive  ⚠️C
│   ├── 5.1 InP pJ/bit efficiency edge is real (~3-4 vs 4-6)        ✅C technically credible (vendor figures)
│   ├── 5.2 Not designed out by SiPh integrated lasers/incumbents   ⊗ real risk; thin moat
│   └── 5.3 Merchant-supplier bargaining power vs LITE/COHR         ✗C sub-scale; weak power
│
└── L1F — Valuation / structural  ✗B (negative)
    ├── 6.1 ~80x P/S justified by implied expectations              ✗B requires bull tail as base case
    ├── 6.2 Thin-float / short-squeeze structure is benign          ✗B €10.23→€6.70 in 2 days
    └── 6.3 SIVEF OTC line investable for a US holder               ✗C illiquid; stale (~10.14 vs 9.42 FX)

Total: 1 ✅ / 8 ⚠️ / 8 ✗ / 3 ⊗ across 21 leaves (tier-weighted B/C — low evidential strength is itself a finding)
H-0 verdict: LEANING FALSIFIED, ~28% confidence. The lone ✅ (InP efficiency) is the weakest-leverage
leaf — a real technical edge does not save the equity when financing (L1C) and integrity (L1D) fail first.

IV. Key findings

Finding 1 — Financing is the binding constraint, and it is failing the test

Cash at YE2025 was SEK 43.5M against a Q1 2026 operating burn of SEK -49.2M — i.e., less than one quarter of runway before the May raise. The ~SEK 125M directed issue (8.62M shares @ SEK 14.5, 2026-05-11) buys roughly 2-3 quarters. The June 15 AGM authorizes up to 53.8M more shares (~15-17% dilution). The company has never been FCF-positive. For a 5-10 year holder this is decisive: you are buying a serial diluter that funds its losses by issuing equity, and the share count grows toward whatever ramp eventually (if ever) arrives. (L1C 3.1/3.2 ✗B.)

Finding 2 — The integrity cluster is, on its own, disqualifying for a long-term hold

Four independent integrity signals fired inside ~6 weeks: (a) the company's own restatement cut FY2024/FY2025 revenue and widened losses during a US PCAOB uplift; (b) the Ningi short report (2026-06-01) alleges ~SEK 97M (~31% of FY2025 revenue) is dubious — products not yet manufactured or government subsidies booked as commercial sales — and "broken promises of an imminent volume ramp since 2018"; (c) a live Swedish Economic Crime Authority insider-trading/market-abuse probe into a leak ~48h before the Nasdaq-US listing announcement, which the prosecutor called "reminiscent of pump-and-dump cases"; (d) insiders and a directed-issue fund sold their ENTIRE stakes into the rally (head of wireless ~SEK 99.5M; Cicero Fonder ~SEK 450-470M). The short source is adversarial, but its central claim rhymes with the company's own restatement — and management has not refuted it with order/contract data. Trust is a precondition for a long-term hold, and it is absent. (L1D 4.1/4.2/4.3.)

Finding 3 — The technology is real; the conversion to revenue is not

The InP efficiency edge (~3-4 pJ/bit vs ~4-6 for silicon photonics) is technically credible, and the GlobalFoundries (silicon-photonics / SCALE reference designs, 2026-06-02) and Jabil (1.6T LRO, OFC 2026) partnerships are real. But all of these are collaborations / reference designs / demos — not committed volume supply contracts with disclosed revenue. Meanwhile revenue FELL -22% in Q1 and the "$799M opportunity pipeline" is a non-binding sales funnel ~25x larger than actual revenue (~$32M). The single most important missing fact in the entire bull case is a committed volume purchase order — and it does not exist. (L1A 1.1 ⚠️C, 1.2 ⊗.)

Finding 4 — The valuation embeds the bull tail as the base case

At ~SEK 25bn / ~$2.7bn on ~$32M revenue, SIVE trades at ~80x trailing P/S while pre-profit and FCF-negative. A reverse-DCF (see implied_prob.md) shows the price requires revenue to compound 12-20x to ~$400-700M by ~2032 at the aspirational >50% GM / >30% EBITDA-margin targets, with that outcome treated as roughly the modal case. Back-solving the scenario targets, the market is pricing ~50% on the bull ramp+listing and only ~10% on the integrity/financing break — almost the inverse of the analyst vector (15/50/35). The mispricing is to the downside. (L1F 6.1 ✗B.)

Finding 5 — The asymmetric trade is the short, not the long

Credible funds have taken the other side: short interest went 1.6% → 17% of float in three months. Analyst EV ≈ SEK 42 (−47%); the base case alone (50% weight) is a ~−58% outcome; the bull payoff (~+97%) is option-like (low probability, with a fraud/probe left tail attached) rather than a margin-of-safety setup. For a long-only 5-10 year holder there is no version of this that is a buy at SEK 79.

Finding 6 — Durability 8/25 with 3 fatal flags = UNOWNABLE for a long-term hold

The durability test (durability_test.md) scores 8/25 and fires three fatal flags (capital allocation 1/5; disruption-survival 1/5 — driven by endogenous failure risk; balance-sheet survivability — sub-1-quarter cash). Per methodology, 2+ fatal flags = unownable for a 5-10 year hold. The InP edge and the partnerships justify watching, not owning at this price.

V. Valuation regimes

(Anchor SEK 78.80 / SIVEF $7.77, 2026-06-05. Bimodal distribution — scenario sketches, not precision. Must match scenarios.md.)

Adopted probability vector: 15 / 50 / 35. Market-implied (back-solved): ~50 / 40 / 10 — the inverse; the market over-weights the ramp ~35pp and under-weights the break ~25pp. Expected value (analyst): ≈ SEK 42 = −47%. Asymmetry: UNFAVORABLE — even the base case (50% weight) is a ~−58% outcome; bull payoff is option-like with a left tail.

Key tension point: the price requires the bull case to be modal; the evidence makes it a tail. The mispricing runs against the long-only buyer.

VI. Long-term durability test

Full execution at reports/SIVE/durability_test.md. Applied with the speculative caveat (clinical-stage-biotech analogue: great science can still be a zero-equity if the company dilutes/mis-reports before commercialization). Summary:

Fatal flags fired: 3 (capital allocation; disruption/endogenous-failure survival; balance-sheet survivability). 2+ = UNOWNABLE for a 5-10 year hold.

⚠️ LONG-TERM-UNSUITABILITY WARNING (escalated to Ming). SIVE scores 8/25 with 3 fatal flags — well inside the "Avoid for long-term holds" band. This is NOT a 5-10 year hold candidate. If exposure to the InP-optics theme is desired, scaled profitable names (Lumentum, Coherent, MACOM) carry the theme without the financing/integrity left tail. SIVE is, at most, a tiny written-off-able speculative option (<0.5%) — and even that is unattractive at ~80x sales with a live insider-trading probe. Default: AVOID.

VII. Triggers and Red Flags

🟢 Triggers (would genuinely strengthen the bull case)

• T1 — A committed VOLUME purchase order (disclosed $, multi-quarter) from GF, Jabil, the LiDAR customer, or a hyperscaler → L1A 1.2 ⊗→✅. The single most important missing fact.
• T2 — Gross margin inflects toward 50% with rising revenue → proves foundry economics.
• T3 — Clean US Nasdaq listing + insider-trading probe closes with no company finding → removes the integrity overhang.
• T4 — FY2026 revenue back above SEK 500M (real H2 recovery).

🔴 Red Flags (several ALREADY FIRING)

• RF1 (FIRING) — live insider-trading/market-abuse probe (escalates to fatal on charges).
• RF2 (FIRING) — restatement + credible 31%-dubious-revenue short report (escalates on restatement #2 / auditor qualification).
• RF3 (FIRING) — insiders + a fund selling entire stakes into the rally; founder leaving board.
• RF4 — another deeply-dilutive raise (53.8M shares authorized June 15 = the loaded gun).
• RF5 — pipeline inflates while revenue stagnates/falls (Q1 -22% already supports).
• RF6 — US class action filed / US listing blocked or withdrawn.

应对 (Response) playbook

• Default: AVOID / do not initiate. Not a "wait for a pullback" name — the left tail (probe, restatement) can impair the equity at any price.
• If legacy-held: treat as a written-off option; do NOT average down; exit into strength.
• Re-engage only if T1 (real volume PO) AND T3 (clean listing + probe closed) BOTH fire — and then only as a tiny speculative option at a far lower sales multiple. Two genuine de-risking events are the minimum bar.
• If RF1 escalates (charges) OR RF2 escalates (restatement #2): uninvestable; thesis over.

Continuous tracking: see dashboard.md.

VIII. Investment logic conclusion

The H-0 — that SIVE is early entry into a 5-10 year InP-optics compounder — requires the joint resolution of seven low-base-rate assumptions (pipeline conversion, partnerships→POs, clean accounting, survival without value-destroying dilution, InP winning the architecture war, >50% gross margin, clean integrity overhang). Each is individually uncertain; jointly the conjunction probability is low. The evidence base (21 leaves: 1 ✅ / 8 ⚠️ / 8 ✗ / 3 ⊗, with four ✗ on the two binding branches L1C+L1D) leaves H-0 leaning falsified at ~28% confidence.

For the 12-24 month horizon: at SEK 78.80 the EV is ≈ −47% with an UNFAVORABLE asymmetry; the market prices ~50% on the bull tail vs the analyst's 15%. The credible asymmetric position is the short (short interest 1.6%→17%), not the long.

For the 5-10 year horizon: durability 8/25 with 3 fatal flags = unownable. The binding constraints are financing (serial dilution; <1 quarter cash pre-raise) and integrity (restatement + live insider-trading probe + short report + insider selling), either of which can impair the equity regardless of how good the InP technology is.

Recommended action for Ming: AVOID. This is a textbook AI-attached small-cap pump with a fragile thesis and a left tail. The technology is interesting; the equity, at this price and with this integrity profile, is not a long-term hold.

IX. System warnings — what could falsify (or, here, what could RESCUE) H-0

Because the default verdict is negative, the falsification logic inverts: the bull case would need to be rescued. It would require, in combination:

1. A committed multi-quarter VOLUME purchase order at disclosed economics (the missing keystone).
2. Gross margin inflecting toward 50% with rising revenue across 2+ quarters.
3. The insider-trading probe closing with no company finding AND a clean US Nasdaq listing.
4. No further restatement / auditor qualification.
5. Financing reaching self-funding without further deeply-dilutive raises.

Conversely, the thesis is confirmed dead by any 1 of: insider-trading charges; a second restatement; the US listing blocked/withdrawn; a dilutive raise below ~SEK 40-50. Several red flags (RF1/RF2/RF3) are already firing.

X. Reading guide / cross-reference

For general stock concepts, read MANUAL_en.md / MANUAL_zh.md at the repo root.

XI. Position conclusion (verdict for Ming)

Verdict: AVOID. SIVE is not a 5-10 year hold. It is a pre-profit photonics small-cap on an AI-attached ~1,800% retail pump, priced at ~80x sales, with three fatal durability flags, a live insider-trading probe, a revenue-cutting restatement, a credible short report, US class actions, and insiders selling out — an expected value of roughly −47% and an asymmetry that favors the short.

Why AVOID (not "buy the dip"):

• Financing: never FCF+; <1 quarter cash pre-raise; serial diluter (53.8M more shares authorized). You are diluted toward a ramp that may not come.
• Integrity: restated (worse) accounts; live Swedish insider-trading probe; Ningi alleges 31% of revenue dubious; US class actions; segment head + a fund sold entire stakes. Trust is absent.
• Conversion: zero committed volume orders; revenue -22% in Q1; the "$799M pipeline" is a ~25x-revenue marketing funnel.
• Valuation: ~80x P/S embeds the bull tail as the base case; EV ~−47%; left tail attached.

What's genuinely real (and why this is a WATCH, not a buy): the InP efficiency edge (~3-4 pJ/bit), the GlobalFoundries/Jabil partnerships, and the LiDAR/SATCOM 2027 optionality. These could, if they convert to committed volume revenue and the integrity overhang clears, justify revisiting the name — at a far lower multiple and after at least two genuine de-risking events (T1 + T3).

Position management: Do not initiate. If exposure to the InP-optics theme is wanted, use scaled profitable names (Lumentum, Coherent, MACOM). If an investor insists on SIVE, size it as a written-off-able speculative option (<0.5%) — and even that is hard to justify at SEK 79 with a live probe. Trade the SEK primary (SIVE.ST), never the illiquid/stale SIVEF OTC line.

Stories lie, structure doesn't. The SIVE story is a thrilling AI-photonics chokepoint. The structure is a ~$32M-revenue serial diluter with a single ✅ leaf, eight ✗ leaves, three fatal flags, and a live criminal probe. Don't read the rally; read the tree.

Don't read news; update your tree.

XII. Investment Scorecard (15-question long-term-hold reference)

Per MANUAL_en.md Part K.6. Bottom-line decision artifact. NO raw buy/sell rating.

Verdict tally: 1 ✅ · 3 ⚠️ · 11 ✗ (Q1✅C; Q2/Q3/Q13/Q14 ⚠️; rest ✗).

K.3.5 Weighted-score derivation

Applying the 4-tier weighting from MANUAL §K.3.5 (verdict values: ✅ = 1.0, ⚠️ = 0.5, ✗ = 0.0). Theoretical max = 39.

Raw K.3.5 = 9.5/39 = 24.4%. A small downward reliability haircut for the pervasive Tier-C / restated / adversarial source base lands the reported xii_score at 22% (the structural quality is so low that the haircut is within rounding of the raw figure; both confirm "avoid"). Bands: ≥85% strong · ≥65% partial · ≥45% weak · <45% avoid ← SIVE. The K.3.5 number, the durability 8/25, and the H-0 28% all agree: avoid. INDEX_META carries fatal_flags: 3.

Scorecard summary

Final verdict: AVOID

For Ming specifically:

• ❌ Do NOT buy / do NOT initiate. Durability 8/25, 3 fatal flags, EV ~−47%, live insider-trading probe, ~80x sales pump.
• ❌ Do NOT average down if legacy-held — treat as a written-off option; exit into strength.
• 📅 Re-engage only if T1 (real committed volume PO) AND T3 (clean US listing + probe closed) BOTH fire, at a far lower multiple. Two de-risking events minimum.
• 📅 Watch the June 15 AGM (board + 53.8M-share dilution) and the ~Aug Q2 interim (revenue recovery? any volume PO?).
• ✅ If the InP-optics THEME is wanted, own scaled profitable names (Lumentum, Coherent, MACOM) instead.

The 2-minute pitch (per MANUAL Part K.1 #4):

"Sivers makes real InP lasers with a real efficiency edge, and just signed AI-optics partnerships with GlobalFoundries and Jabil — which is why a ~$32M-revenue, money-losing Swedish small-cap is up ~1,800% to a ~$2.7bn / ~80x-sales valuation. But the partnerships are reference designs, not volume contracts; there are zero committed orders; revenue actually FELL 22% last quarter; the company has under one quarter of cash and survives by serial dilution; it just restated its accounts DOWN; a short seller alleges a third of its revenue is fake; insiders and a fund dumped their entire stakes; and Swedish prosecutors are running a pump-and-dump-flavored insider-trading probe. Durability is 8/25 with three fatal flags. The asymmetric trade here is the short, not the long. For a 5-10 year hold: avoid — and if you want the photonics theme, buy Lumentum or Coherent."

Risk types most relevant (per MANUAL Part K.4): financing/dilution risk; accounting/integrity (fraud-tail) risk; execution/conversion risk; competition (design-out) risk; valuation/momentum-reversal risk; liquidity risk (SIVEF OTC); regulatory/legal risk (insider probe, class actions).

"When NOT to buy" anti-patterns (per MANUAL Part K.5) — DO ANY APPLY?

• ✅ Online hype / retail-driven: FIRES — explicitly characterized as a "retail-driven pump."
• ✅ Stock went up fast: FIRES — ~1,800% YTD; ATH SEK 110 then -34% in two days.
• ✅ Has "AI" attached: FIRES — the entire re-rating is AI-datacenter-optics narrative.
• ✅ Looks exciting / cutting-edge story over numbers: FIRES — "chokepoint/supercycle" vs falling revenue + zero orders.
• ⚠️ "Next [Nvidia]"-style framing: present in bull commentary ("chokepoint for the supercycle").
• ✅ Added integrity tells: insider selling + live insider-trading probe + short report + restatement.

Net: every major "do NOT buy" anti-pattern fires. This is the canonical configuration the discipline is designed to refuse. AVOID.

Tree v1 written 2026-06-08. Next refresh: ~Aug 2026 (Q2 interim) + June 15 AGM outcome. Refresh cadence: quarterly, OR immediately on any RF1/RF2/RF6 escalation (insider-trading charges, restatement #2, US listing blocked).