The CFO was asked directly how much 2027 growth comes from committed programs versus qualification-stage. She declined to answer.
Thesis: Navitas is building inventory and prepaid wafer capacity at 3x its quarterly revenue while the CFO refuses to separate committed orders from qualification-stage pipeline. The sequential trajectory is real but the YoY base shrank 26% because mobile collapsed before AI materialized. You cannot size the risk when the one number that matters, backlog quality, was explicitly withheld.
Verdict: AVOID — Conviction: MEDIUM
Catalyst: Q3 prints the $13.5M midpoint guide with AI infrastructure at one-third of mix, confirming the pre-800V SiC-in-AC/DC-PSU thesis is booking real revenue, not just pipeline. Management described multiple programs moving through DVT, EVT, PVT stages with qualification before year-end.
Key Risk: Inventory write-down if the AI ramp slips. They built $19.5M inventory plus $15M prepaid TSMC wafers, totaling $34.5M against $10.5M quarterly revenue. Buffers against uncertain demand become write-offs when demand doesn't arrive on schedule.
The Tell: CEO Chris Allexandre: 'We are surprised, and I'm sure you are surprised, by the momentum that we have in the business.' An unprompted admission that even management did not model this sequential acceleration from a base that is still shrinking year-over-year. If they didn't model it, nobody did.
Friction Level: HIGH_FRICTION — The bull sees a pre-800V structural demand shift across tens of programs at multiple hyperscalers driving real bookings. The bear sees a company with 26% YoY revenue decline, unquantified backlog, and inventory building 3x ahead of quarterly revenue against demand not yet contracted.
2026Q2 · Jul 27, 2026
4 of 5 readings agreed4 AVOID · 1 SHORTRead the full call
Four Sequential Beats, Inventory Building Ahead of Unbooked Demand
Navitas is building inventory and prepaid wafer capacity at 3x its quarterly revenue while the CFO refuses to separate committed orders from qualification-stage pipeline. The sequential trajectory is real but the YoY base shrank 26% because mobile collapsed before AI materialized. You cannot size the risk when the one number that matters, backlog quality, was explicitly withheld.
CEO Chris Allexandre: 'We are surprised, and I'm sure you are surprised, by the momentum that we have in the business.' An unprompted admission that even management did not model this sequential acceleration from a base that is still shrinking year-over-year. If they didn't model it, nobody did.
CFO Tonya Stevens was asked directly what percentage of 2027 growth is committed versus qualification-stage. She said 'We aren't breaking out what % is committed in 2027 or what % relates to our backlog.' Record book-to-bill was called but never quantified.
Revenue grew 22% sequentially to $10.5M while inventory rose 31% from $14.9M to $19.5M, plus $15M prepaid TSMC wafers. Total inventory-related working capital jumped from $14.9M to approximately $34.5M, more than doubling, ahead of demand not yet booked.
AI infrastructure will be one-third of revenue by year-end, but the entire 2027-2028 growth thesis maps to inflection points (native 800V, solid state transformers) that have not materialized. Current-period AI revenue is the pre-800V side: partial SiC replacement in AC/DC PSUs at modest scale.
Four consecutive quarters of sequential revenue gains. Q2 beat high end of guidance. Q3 guide of $13.5M represents 28% sequential growth with CEO stating guide came above Street expectations. Record book-to-bill and backlog extending beyond 2026.
CEO volunteered he is 'surprised' by momentum, admitting management did not model this acceleration. All confidence language points forward to unbooked inflection points. Zero hedging in a quarter where operating losses widened and YoY revenue declined ~26%.
$557M cash after raising $373M at $21.89 average, zero debt. $15M prepaid TSMC wafers and buffer capacity through 2029 secured. Magnachip foundry deal creates second SiC source. Management is buying capacity ahead of demand with cash.
TSMC buffer capacity secured through 2029, GlobalFoundries 8-inch GaN sampling by year-end, Magnachip second foundry for GeneSiC, OSAT consolidation. Management is systematically removing supply-side constraints before the AI ramp.
+70
CONFIDENT"CEO used 'inevitable', 'explosive growth', 'highly confident', volunteered system-level testing status and wattage specifics unprompted. Zero hedging language. Deflected only on Kyber specifics and backlog quality."
Full transcript with specific financials and guidance. However, CFO withheld committed vs qualification-stage backlog split, CEO declined Kyber specifics, and book-to-bill magnitude never quantified despite being called 'record'.
Q3 prints the $13.5M midpoint guide with AI infrastructure at one-third of mix, confirming the pre-800V SiC-in-AC/DC-PSU thesis is booking real revenue, not just pipeline. Management described multiple programs moving through DVT, EVT, PVT stages with qualification before year-end.
Inventory write-down if the AI ramp slips. They built $19.5M inventory plus $15M prepaid TSMC wafers, totaling $34.5M against $10.5M quarterly revenue. Buffers against uncertain demand become write-offs when demand doesn't arrive on schedule.
“Revenue in the second quarter of 2026 was at the high end of guidance, increasing 22% sequentially to $10.5 million.”
Changed less than 63% of calls
Wording compared with the May 5, 2026 call, ranked among 98 calls reporting on Q2 2026. A measure of how much the story moved, not a signal about the stock.
Figures are extracted from the transcripts and checked in code: every quote appears word for word. Language change is computed from word frequencies, with no model involved. Descriptive only: none of this feeds the verdict.
Four sequential beats with Q3 guided above Street. AI infrastructure compounding >50% QoQ in both Q1 and Q2, set to hit one-third of revenue by year-end. High-power markets up 50% YoY. Gross margin expanding 50bps sequentially on mix. The growth is pre-800V and spans tens of programs across multiple hyperscalers, multiple merchant power vendors, both GaN and SiC in the same systems. The street priced this on a single 2027 native-800V socket from one GPU vendor; management just showed the revenue is arriving from AC/DC PSU silicon-to-SiC replacement happening now. OpEx flat for nine months then up only 10% against 28% revenue guide. $557M cash, zero debt, capacity secured through 2029 at TSMC with GlobalFoundries and Magnachip as second sources. This is beat-and-raise with structural demand behind it.
Revenue of $10.5M is down approximately 26% YoY from roughly $14.3M in Q2 2025. Operating loss widened from $10.6M to $11.4M. The company burns $10-12M per quarter at the operating line on a $42M annual revenue base. Inventory and prepaids totaled $34.5M, 3.3x quarterly revenue, built ahead of demand the CFO won't quantify. The CFO explicitly refused to separate committed from qualification-stage backlog. Two patent lawsuits from Wolfspeed and Renesas target core IP at a fabless company whose value proposition is IP differentiation. The entire growth narrative maps to 2027-2028 architecture transitions with no contracted revenue attached. Cash of $557M bought at $21.89 average provides runway but the company needs 5x revenue scale to reach breakeven.
Backlog Quality
CFO Tonya Stevens declined to break out committed versus qualification-stage programs when asked directly, citing 'inflection points coming on top of each other.' Record book-to-bill was asserted but never quantified. Cannot verify demand quality.
Inventory Divergence
Inventory up 31% sequentially to $19.5M plus $15M prepaid TSMC wafers = $34.5M total, against $10.5M quarterly revenue. Pre-build ahead of unbooked ramp. Buffers against uncertain demand become write-downs when demand slips.
AI Label Without Substance
AI infrastructure anchors the valuation but the 2027-2028 inflection points (native 800V, SST) have zero booked revenue. Current AI revenue is partial SiC replacement in AC/DC PSUs at modest scale from a $10.5M base.
Unsustainable Trend Confidence
CEO used 'inevitable', 'explosive growth', 'highly confident' with zero hedging while operating losses widened YoY. Volunteered that management is 'surprised' by momentum they did not model. Maximum confidence on unbooked future growth.
AVOID. The CFO literally refused to answer the one question that determines whether this backlog is real orders or qualification pipeline. You cannot size a position when you cannot size the risk. Add two patent lawsuits on core IP at a fabless company and inventory at 3.3x quarterly revenue built ahead of uncommitted demand. Four sequential beats are real but the YoY base shrank 26%. This is binary. Wait for Q3 to confirm committed backlog quality or watch the inventory build. No edge without information the company won't give you.