A newer SEC filing has been made since this research was written — check the primary sources before acting on a number here.
A real ~$40M-revenue GaN/SiC design house that the market has repriced as a ~$4.2B NVIDIA-800V options contract — the technology and the design-in are genuine, but at ~105x sales, a distributor-of-record 76% in Hong Kong, a TSMC-GaN exit in 2027 and management/insiders diluting and selling into the spike, the valuation is discounting a data-center ramp that has not yet shown up in a single quarter's numbers.
Price
Weekly closes
No Friday close is on the record for NVTS yet. The weekly job prices the covered universe; a name it cannot price is listed as missing rather than estimated.
Research
The Navitas Semiconductor dossier
Researched July 1, 2026
The verdict
A real ~$40M-revenue GaN/SiC design house that the market has repriced as a ~$4.2B NVIDIA-800V options contract — the technology and the design-in are genuine, but at ~105x sales, a distributor-of-record 76% in Hong Kong, a TSMC-GaN exit in 2027 and management/insiders diluting and selling into the spike, the valuation is discounting a data-center ramp that has not yet shown up in a single quarter's numbers.
Navitas is a fabless wide-bandgap (WBG) power-semiconductor design house. It sells (1) GaNFast™ gallium-nitride power ICs (GaN-on-Si, with integrated drive/control/protection — "GaNSense"), (2) GeneSiC™ high-voltage silicon-carbide MOSFETs and diodes (trench-assisted planar, up to 6.5 kV), and (3) associated high-speed silicon controllers and digital isolators, all used in power conversion and charging. Founded as Legacy Navitas in 2014, it went public via a Live Oak Acquisition Corp II SPAC on 2021-10-19; it acquired GeneSiC on 2022-08-15, which is the origin of the SiC line. HQ Torrance, CA; operations across the US, Ireland, Germany, Italy, Belgium, China, Taiwan, Thailand, South Korea, the Philippines; 300+ patents; CarbonNeutral-certified; Nasdaq: NVTS since 2021-10-20.
The defining fact of 2025 is the "Navitas 2.0" pivot announced late 2025: a hard turn away from the historical mobile/consumer/China charger business (which collapsed) and toward four high-power markets — AI data centers, energy & grid infrastructure, performance computing, industrial electrification — deliberately excluding EVs and low-voltage SiC. The commercial engine of the new story is the NVIDIA collaboration announced 2025-05-21: Navitas was selected for NVIDIA's next-generation 800 V HVDC data-center architecture, supplying GaNFast + GeneSiC content into the "Kyber" rack-scale systems that power Rubin Ultra-class GPUs.
Contract structure — a red flag hiding in plain sight. Navitas sells primarily via purchase orders against forecast, not long-term supply contracts; backlog "is subject to cancellation, reduction or delay". Revenue is booked overwhelmingly through distributors (distributor-of-record model). So the "customers" of record are electronic-component distributors, not the hyperscalers in the press releases — a critical distinction for revenue durability.
Supply Chain
Upstream → Navitas → end-customer, every named stakeholder:
GaN epi/substrate → wafer fab. GaN-on-Si. Historical GaN foundry: TSMC (Taiwan). Chokepoint / single-source risk — the headline of this whole dossier: on 2025-07-01 TSMC announced it will cease GaN production in July 2027. Mitigation, per the Form 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes. and confirmed on the wire: (i) buffer-inventory purchase commitment from TSMC; (ii) expanded collaboration with Powerchip Semiconductor Manufacturing Corp (PSMC, Taiwan); (iii) a Nov 2025 long-term strategic partnership with GlobalFoundries (GFS) to build GaN at GF's Burlington, Vermont fab — development early 2026, production later 2026. The Form 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes. names GF as the US GaN partner and X-Fab (US) as the SiC manufacturer; two further Asia fabs support GaN/CMOS.
Assembly / test / packaging → outsourced to third-party OSAT partners (fabless model, low Capital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs.).
Named end-ecosystem partner:NVIDIA (800V HVDC / Kyber).
The company frames its US-foundry footprint (GlobalFoundries GaN + X-Fab SiC) as a friend-shoring competitive advantage — that foreign-supply-chain competitors will be disadvantaged in AI-data-center and grid markets. Note the tension: it is simultaneously a dual-domesticated Ireland/US "inverted domestic corporation," which the 10-K flags could bar it from US federal contracts under FAR / the Homeland Security Act.
Competitive Advantages (moats)
Dual GaN + high-voltage-SiC full-stack. Navitas is one of very few suppliers offering both GaN (medium-V, high-frequency, high-density) and high-voltage SiC (up to 6.5 kV) — most rivals specialize in one. It claims it can cover "the entire power path from the utility grid to the GPU". This system-level breadth is the genuine differentiator into 800V HVDC.
Shipped volume + IP. 300M+ GaN devices and ~30M SiC devices shipped as of 2025-12-31; "one of the broadest IP portfolios" with 300+ patents spanning GaN and SiC fundamentals — a real barrier vs. new entrants.
Capital structure as moat. Fabless (low fixed cost) + no debt + a large cash pile — the 10-K explicitly cites "capital efficiency and financial strength" and "optionality".
The NVIDIA design-in. Being named into NVIDIA's reference 800V architecture is a powerful third-party validation and a potential switching-cost/qualification moat if it converts to volume.
Bargaining power — weak on both sides today. Navitas needs its distributors more than they need it (56% then 46%/59% single-distributor concentration), and it needs its foundries badly (TSMC exit forces a scramble). Its bargaining power rests almost entirely on the promise of design-win-driven demand, not on current scale. Moat verdict: real technology moat, not yet a commercial moat — a company with a genuine edge and almost no pricing power until the high-power ramp materializes.
Segments
Navitas reports a single operating segment (the CEO/CODM reviews consolidated results) — there is no product-line P&L split in the filings. What is disclosed:
By geography (billing/distributor location, not end-demand):
Region
FY2025
FY2024
Q1-2026
Q1-2025
Hong Kong
57%
64%
76%
59%
Rest of Asia
21%
18%
8%
18%
US
10%
9%
13%
9%
China
9%
4%
2%
7%
Europe
3%
4%
1%
4%
Read: despite the "US high-power" pivot narrative, revenue is still routed 76% through Hong Kong in Q1-2026 — more concentrated than a year ago, because the surviving base is still the legacy Asia charger channel while high-power design wins have not yet shipped in scale. Management says "high-power markets grew ~35% YoY and now represent the majority of revenue", but the majority-of-a-shrinking-pie is on ~$8.6M/qtr. The geographic mix is the tell that Navitas 2.0 is a promise, not yet a print.
Phase B — Measure performance
Earnings Result
FY2025 (year ended 2025-12-31):
Net revenue $45.9M, −45% YoY (from $83.3M) — driven by the collapse of mobile/consumer in China.
Operating loss −$107.8M; net loss −$117.0M; EPS −$0.57.
Net loss included a −$12.4M non-cash earnout re-mark (stock rose, liability grew) — swinging from a +$36.6M gain in 2024, a $49M YoY swing that is pure fair-value noise.
Cash from ops −$42.9M; capex only $1.5M (fabless).
Q1-2026 (quarter ended 2026-03-31) — the latest print:
Revenue $8.6M vs $14.0M Q1-2025 (−39% YoY) but +18% sequential vs Q4-2025.
Balance sheet. Cash $221.0M at 2026-03-31 (down from $236.9M at YE2025; YE2025 up from $86.7M after raising ~$200M PIPE+ATM in 2025). No debt. Goodwill $163.2M (mostly GeneSiC); intangibles $53.3M; inventory $14.9M; accumulated deficit −$501.7M. AR only $3.6M.
Market reaction / what's priced in: the tape is completely decoupled from the P&L — the stock ran from a $1.52 low (2025-04-04) to a $34.17 high (2026-06-03), ~22x, on the NVIDIA/800V narrative, not on a $45M-revenue, $117M-loss year. The Q1 "beat" was a beat on a de-minimis base.
Earnings Calls (sentiment trend)
No transcripts on the shelf (transcripts/ empty) — ``. Management tone across 2025→Q1-2026 has shifted decisively from apology (2024/early-2025: China/mobile collapse, distributor disengagement, restructuring) to evangelism (late-2025 onward: "Navitas 2.0," 800V, NVIDIA, "AI factory power," sequential-growth-through-2026). Recurring new phrases: "high-power markets," "800 VDC," "GaN + SiC full power path," "Electrify Our World." Things they stopped saying: mobile-fast-charging leadership, consumer TAM. The Q1-2026 call framed the story as inflection ahead — guiding sequential growth ($8.6M→$10.0M) and leaning on the NVIDIA/GF partnerships rather than current dollars. Sentiment is bullish-forward but the substance is still guidance, not delivered high-power revenue.
Comps
Peer set = named competitors in the 10-K (GaN: Infineon, Power Integrations, TI, Innoscience, Renesas, EPC; SiC: Infineon, Wolfspeed, onsemi, ROHM, Qorvo, STMicro). Multiples are `` with date, or n/a. No multiple is fabricated.
Company
Ticker
Mkt cap
Revenue basis
P/S
P/E
Note
Navitas
NVTS
$4.24B
TTM $40.5M
~105x
n/m (loss)
~$40M rev, −45% YoY
Power Integrations
POWI
~$4.85B
—
n/a
~301x
closest GaN/power-IC peer; also richly valued
onsemi
ON
~$44.2B
—
n/a
~35x fwd
SiC scale leader; buying Synaptics $6.7B
Wolfspeed
WOLF
~$2.5B
—
~0.25–3.3x
n/m (loss)
distressed SiC pure-play post-restructuring
Infineon
IFX
n/a
—
n/a
n/a
largest power-semi incumbent
STMicro
STM
n/a
—
n/a
n/a
GaN+SiC incumbent
5-yr avg ROE: n/a for peers; NVTS ROE is deeply negative (net loss on $443.7M equity).
Takeaway: the entire GaN/power-semi group is expensive in 2026 (POWI ~301x P/E is a warning about the sector's AI-power exuberance), but NVTS at ~105x sales is in a class of its own — it is priced not as a semiconductor company but as a levered call on the 800V/AI-power TAM. onsemi (~35x P/E, profitable, $44B) and even distressed Wolfspeed (~0.25–3.3x P/S) show how far NVTS trades from any earnings- or sales-anchored reality.
Stock-Price Catalysts (moves >5%)
Pattern is unambiguous — this stock trades on data-center/partnership headlines and dilution, essentially not at all on its own P&L:
2025-04-04: all-time low $1.52 — capitulation on the China/mobile collapse.
2025-05-21:NVIDIA 800V HVDC selection — the launch of the entire re-rating.
2025-08-22 / effective 2025-09-01: founder CEO Sheridan out, Chris Allexandre (ex-Renesas) in.
2025-11-20:GlobalFoundries GaN partnership (US GaN, Vermont).
2026-07-01: ~$17.19, off ~50% from the high but still +265% YTD, mkt cap $4.24B.
The market reacts to (a) NVIDIA/hyperscaler design-in news, (b) foundry-supply resolution, and (c) equity issuance/insider selling. Fundamentals (revenue, margin) move it only marginally. This is a story/positioning stock, not a numbers stock — for now.
Phase C — Judge people & books
Management
CEO — Chris Allexandre (since 2025-09-01). 25+ yrs semis; came from Renesas, running power management for cloud/industrial — a credible, operator profile deliberately matched to the high-power pivot. Pay: $520k base, 120% target bonus, 800k RSU recruitment grant (vesting 2027–2029) + a 2026 PSU worth ~$2.5M. New, unproven at Navitas; skin-in-the-game is being granted, not bought.
Founder — Gene Sheridan (co-founder, CEO 2014–2025) stepped down 2025-08-31; departure "not due to any disagreement". A founder handing off mid-pivot to a professional manager — a reasonable move to bring in scaling discipline, but it removes the founder-driver at the exact moment the story is being sold hardest.
Ranbir Singh — GeneSiC founder / SiCPower principal, on the board; entered a new agreement April 2025 — and sold ~3.7M shares for $108.7M on 2026-05-27 (retains ~14.9M). Legal, disclosed, and enormous — the person who knows the SiC asset best took nine figures off the table into the spike. A material insider-distribution signal.
Capital allocation. Fabless, disciplined on capex ($1.5M FY2025). But the financing pattern is aggressive dilution: shares 188M (2024) → 230.5M (YE2025); ~$200M PIPE+ATM in 2025; $122M ATM completed May 2026; a fresh $500M ATM launched June 2026. They are funding the company off a rich share price — rational financial engineering, but it caps upside via relentless share creation.
Red flags (governance): a director exit accompanied the $500M ATM; two auditor changes in as many years (see Lens 10); heavy SBC; a still-live earnout complex. Archetype: transitioned from founder-led to professional-manager-led, financially opportunistic, technology-credible.
Forensic Red Flags
Grounded in the filings; every figure labeled.
Earnout liability volatility. Up to 10,000,000 earnout shares (9.71M vested-share tranche liability-classified) valued by Monte Carlo; fair value $22.6M at YE2025 vs $10.2M prior; re-marks drove a −$12.4M hit in FY2025 and a −$7.9M hit in Q1-2026 — a KPMG-flagged critical audit matter. This makes GAAP net loss swing on the stock price, not operations. Milestones ($12.50/$17.00/$20.00 VWAP hurdles) expire 2026-10-19 — with the stock ~$17–34 in 2026, some hurdles have plausibly triggered, crystallizing dilution.
Auditor change.KPMG (engaged 2025) replaced Baker Tilly (2023–2025) — KPMG signed the FY2025 opinion, Baker Tilly the FY2024. An auditor transition concurrent with a CEO change and a strategic pivot warrants attention, though the 10-K discloses no disagreement and the opinion is clean (no going-concern).
Stock-based comp distorts non-GAAP. SBC $14.5M FY2025 and $10.3M in Q1-2026 alone (>the entire quarter's revenue) — the 39% "non-GAAP" gross margin and −$0.04 "non-GAAP" EPS strip out real economic dilution. FY2024 R&D/SG&A swung on ~$28M of SBC reversals from executive departures — earnings quality is low.
Inventory flagged by KPMG as the other critical audit matter ($13.3M YE2025, judgment-heavy demand forecasts amid the pivot).
Goodwill $163.2M (36% of assets) from GeneSiC — not yet impaired, but a candidate if the SiC/high-power thesis stalls.
Cash vs. earnings:Cash burnHow much more cash goes out than comes in, per period. The clock on a company with no profits. (−$42.9M FY2025) is smaller than net loss (−$117.0M) because the loss is stuffed with non-cash items (amort $18.9M, SBC $14.5M, earnout $12.4M, impairment $18.0M). Working capital is a source of cash as revenue shrinks (AR fell $9.5M) — a melting-ice-cube tailwind that reverses when growth returns.
Tax structure: dual-domesticated Ireland/US "inverted domestic corporation" — flagged as a potential bar to US federal contracts and a source of double-taxation risk.
Regulatory findings (required).
SEC LR / AAER:none. "No LR found" and "No AAER found" for Navitas Semiconductor, 2021-07-01 → 2026-07-01, per SEC EDGAR EFTS.
10-K Item 3 (Legal Proceedings): incorporated by reference to Note 15 "Commitments and Contingencies"; the filing discloses no material pending litigation — contractual obligations are lease + an equipment-purchase agreement, not lawsuits.
Non-SEC (FTC/DOJ/FDA/etc.): web search surfaced no material enforcement action, consent decree, fine or penalty against Navitas.
Conclusion: No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER), web search, and 10-K Item 3 as of 2026-07-01. The forensic risk here is earnings-quality / dilution / valuation, not fraud or litigation.
Phase D — Project & stress-test
Forward Projection
No our model create in this watchlist run (per skill: log a Brier forecast only on genuine commitment). Company is loss-making, so the scoreable question is a revenue trajectory + path-to-breakeven, not an EPS beat. Built bottom-up from the guide.
Base inputs: Q1-2026 $8.6M actual; Q2-2026 guide ~$10.0M; management guides sequential growth through 2026; non-GAAP GM ~39% and rising with high-power mix; opex ~$30M/qtr; cash $221M pre the new $500M ATM.
Bear (design-ins slip, mobile keeps bleeding): FY2026 rev ~$38–42M; FY2027 ~$50–65M; EPS stays negative (~−$0.45 to −$0.60) through FY2028; high-power ramp is a 2028+ event; dilution continues.
Base (NVIDIA/800V converts on schedule): FY2026 rev ~$42–48M (H2 acceleration), FY2027 ~$80–120M as 800V racks ship, FY2028 ~$150–220M; non-GAAP EPS approaches breakeven in FY2028, GAAP later; GM to ~40–45%.
Bull (800V standard adoption + share gains): FY2027 rev $150M+, FY2028 $300M+; operating leverage flips the model to non-GAAP profit in FY2027, GAAP FY2028; GM 45%+.
Scoreable base call (for a future our position log commit, not logged here):NVTS FY2027 revenue ≥ $100M, p≈0.40 — i.e. the base/bull path (800V ramping) is roughly a coin-flip-to-underdog, and today's ~$4.2B cap requires the bull. The binary that actually matters: does 800V HVDC content ship in production volume across FY2027, and does it land as Navitas revenue (not just design-win PR)?
Bull vs Bear
Bull case. Navitas is a rare dual-GaN+SiC full-power-path supplier with a named NVIDIA 800V design-in at the exact moment AI data centers are forced from 12V/48V toward 800 VDC — a genuine architectural inflection where WBG is required, not optional. It has 300M+ GaN devices shipped (real reliability data), 300+ patents, a debt-free ~$220M+ (soon $700M+) balance sheet, US foundry supply (GlobalFoundries GaN, X-Fab SiC) that friend-shores it into critical-infrastructure demand, and a professional operator (ex-Renesas) now running the pivot. If 800V racks ramp on the Rubin-Ultra timeline and Navitas captures even single-digit content share of the AI-power TAM, revenue could 5–10x off a tiny base and the model flips to profit on fabless operating leverage. The market is paying for optionality on a structural power-architecture shift.
Bear case (permanent-impairment risks).
Design-win ≠ revenue. The 10-K itself warns design wins "typically do not result [in revenue] for one year or more, if ever." NVIDIA can dual-source, re-architect, or delay; a reference-design slot is not a contract. If 800V volume lands in FY2028+ or goes largely to Infineon/onsemi, the thesis is right on the trend, wrong on the beneficiary.
Foundry cliff. TSMC exits GaN in July 2027; production at GlobalFoundries/Powerchip must qualify and ramp into the demand window. Any slip risks missing the 800V ramp it is being valued for.
Valuation. ~105x TTM sales on a company that just shrank 45% and loses $100M+/yr. Even the base case (FY2027 ~$100M rev) leaves it at ~40x forward sales — a multiple that assumes flawless execution and offers no margin of safety.
Pre-mortem (18 months out, thesis broke): it's end-2027. 800V racks shipped, but the power-stage sockets went disproportionately to Infineon/onsemi/Power Integrations, who out-shipped Navitas on price and qualification. Navitas revenue is ~$70M, still loss-making; TSMC's GaN exit forced a costly re-qual that dented margins; the earnout and two $500M-class ATMs pushed share count past 260M; the NVIDIA halo faded from "supplier" to "one of several." The stock re-rated from ~105x sales to ~10x → down 70–80%.
Contrarian view (what the market refuses to see): the market is treating "NVIDIA selected Navitas" as if Navitas is NVIDIA's power supplier — but Navitas is a ~$40M-revenue components vendor selling through Hong Kong distributors, and the incumbents it must beat (Infineon, onsemi) are 10–1000x its size with their own GaN/SiC and deeper OEM relationships. The bet is not "is 800V real?" (it is) — it's "does a sub-scale specialist win and keep the socket against giants who want it too?" That is far from priced.
Devil's Advocate (short-seller)
Revenue concentration is extreme and fragile. One distributor = 56% (2024) → 46%/59% (Q1-2026) of revenue and 58% of receivables; 76% of revenue routes through Hong Kong. Lose or shift one distributor (as happened with "Distributor A" at end-2024, costing $7.5M bad debt + $5.0M write-down + $1.7M R&D abandonment) and the P&L craters.
The moat may be thinner than bulls think. GaN-on-Si and SiC are being pursued by every large power-semi (Infineon and Innoscience and onsemi and Power Integrations); Navitas' edge is integration + first-mover shipping, not a defensible process node. Its foundry (GF/X-Fab/Powerchip) is contract capacity anyone can rent.
Most dangerous competitor bulls underestimate: Infineon — it plays in both GaN and SiC, has the OEM relationships, the balance sheet, and is explicitly named as a competitor on every product line. If Infineon prioritizes 800V, Navitas' "unique full-stack" claim erodes fast.
Capital-allocation / incentive misalignment: a $500M ATM into a spike + a director's $108.7M insider sale + heavy SBC = insiders and the company monetizing the share price while public holders absorb dilution. When the people closest to the asset are net sellers/diluters, weight it.
What must hold for today's price: 800V converts to Navitas revenue by FY2027; GF/Powerchip qualify before the TSMC cliff; margins scale to 40%+; NVIDIA stays loyal; no incumbent takes the socket. All of it, roughly on time.
If growth disappoints 20–30%: at ~105x sales there is no earnings floor — a miss re-rates the multiple, not just the estimate; a halving-or-worse from a growth stumble is the base outcome, not a tail.
Single scenario that permanently impairs: NVIDIA (and/or the 800V ecosystem) standardizes on a competitor's power stage → Navitas is left as a sub-scale, cash-burning charger-legacy company with a $163M goodwill impairment risk. Plausibility: moderate — real, given the incumbents' size and the design-win-≠-revenue gap.
Management Questions (ordered by information value)
Of the NVIDIA 800V/Kyber collaboration, what is contracted vs. reference-design — do you have binding purchase commitments or forecasts, and what quarter does 800V content first appear as recognized Navitas revenue?
What is your expected FY2027 revenue from AI-data-center high-power specifically (not total), and what content-per-rack / share assumption underlies it?
Walk through the TSMC-GaN-exit bridge: what % of GaN volume is qualified at GlobalFoundries and Powerchip today, and is production certain to ramp before July 2027 with no margin hit?
Who is "Distributor A/B" at ~46–59% of revenue, and what is the concrete plan to reduce single-distributor and Hong-Kong concentration as high-power scales?
Why launch a $500M ATM on top of $220M cash and $122M just raised — is this war-chest optionality, M&A, or a signal about the burn/ramp timeline?
Which incumbents (Infineon, onsemi, Power Integrations) are you designed against at the hyperscalers, and where do you win/lose the socket on price vs. performance?
What non-GAAP operating-profit revenue level do you need, and in which fiscal year do you expect GAAP operating breakeven?
Post the 2026-10-19 earnout expiry, what is the final dilution, and are any hurdles now met?
What did the CEO transition from the founder change strategically — what would Chris Allexandre do differently from Gene Sheridan?
How durable is 39%+ non-GAAP gross margin as you scale, given automotive-grade/data-center qualification cost and foundry re-sourcing?
What is the design-win pipeline by dollar value and expected-revenue-timing, not logo count?
How much of R&D is now high-power vs. maintaining the legacy mobile/consumer base, and when does legacy revenue stop declining?
What happens to the model if 800V HVDC adoption slips 12–18 months industry-wide?
Given SBC ran >100% of Q1-2026 revenue, what is the multi-year SBC and share-count trajectory?
What single competitive or supply development would most threaten the thesis over the next 24 months, and how are you hedged?