Navitas Semiconductor
Navitas designs power chips made from gallium nitride and silicon carbide, and has bet the company on the power supplies inside AI data centres.
The scores
Growth1/4 ↑
Revenue fell 45% in 2025 and is still 27% below a year ago; two quarters of sequential growth come from a far smaller base.
Business quality2/4 →
Its process is good enough to license, but fourteen suppliers share NVIDIA's list and larger rivals own their fabs; margins show no pricing power.
Financial position3/4 ↑
About $557M of cash and no debt covers years of losses, but it came from selling shares, not from operations.
Execution record2/4 ↑
Revenue halved as the old market collapsed and the main distributor was lost in 2024; two clean guides under a new CEO are too few.
Per share capture2/4 ↓
Shares up 39% in eighteen months and stock pay near revenue; not lower because the stock was sold dear and no raise is needed soon.
Potential2/4 →
A real market shift, but nothing signed with volume: NVIDIA's timetable, socket wins against thirteen rivals and a new GaN foundry must all land.
Risk profile2/4 ↓
One distributor carries 71% of sales, and Wolfspeed's suit seeks an injunction on GaN and SiC products: a single ruling inside the horizon.
Valuation1/4 →
For the price to be ordinary for a chip company, revenue must reach 11 to 16 times today's, more than any analyst forecasts before 2030.
Evidence quality3/4 →
The price gap and the new shares are filed, but the growth case rests on unfiled management figures and one reprinted market estimate.
If it works 0.4× · if it fails you keep about 5%.
The Verdict
We rate Navitas unattractive. It designs power chips from newer materials that waste less electricity, and it has bet the company on AI data centres after leaving phone chargers, its main market. The business is weak: revenue halved, losses continue and the share count keeps rising. The price is stretched, valuing the company far above any rival it names for each dollar of sales. As a bet it is fragile: even if it works, today's money shrinks by more than half; if it fails, almost nothing is left. We think the shift to new data-centre power is real, but Navitas is one of fourteen suppliers with no large order signed.
The long caseAI data centres need far more power per rack and are moving to higher voltages, where Navitas's chips waste less energy than silicon; it still has to win production orders.
The evidence under this - 9 claims
- An analyst on Navitas' Q2 2026 call cited market reports that the Kyber rack may have been cancelled and replaced; the CEO declined to comment on NVIDIA's plans (S04).
- Yole puts first commercial 800-volt rollouts only 'around 2027' (S12).
- A fast grower can deserve far more than a mature one's multiple, and investors pay forward, not trailing: at consensus 2028 revenue the figure is 21.6 times (C27).
- The cohort's members sell mostly silicon into cars and industry; a pure AI-power name may carry a premium for years.
The company
Every computer, charger and server has a power supply that turns one voltage into another by switching current on and off very fast. Navitas designs the switches. Its gallium nitride chips switch faster than silicon, so power supplies can be smaller; its silicon carbide chips handle thousands of volts, which suits grid equipment and the input of data-centre power.
It owns no factory: TSMC, Powerchip, GlobalFoundries and X-Fab make its chips. Its buyers are power-supply makers, and behind them the data-centre operators NVIDIA is designing for.
The evidence under this - 3 claims
How it makes money
Navitas sells chips, mostly through electronics distributors, who resell them to power-supply makers. One distributor handled 71% of sales in the June 2026 quarter.
Each chip earns about 40 cents of gross profit per dollar on the company's adjusted measure, but at $10.5M of quarterly sales that does not cover the costs of running the business, so it loses money every quarter. It pays its bills from cash raised by selling new shares.
The evidence under this - 4 claims
Where the money comes from, stream by stream
| Stream | Share of sales (Q4 2026 (guide)) | How it earns it |
|---|---|---|
| AI infrastructure: data-centre and grid power | not disclosed; guided above a third of sales by Q4 2026 | GaN and SiC chips sold through distributors to power-supply makers for AI racks and grid and energy equipment. |
| Performance computing and industrial electrification | not disclosed; with AI infrastructure, the majority of Q2 2026 sales | Higher-power GaN chips for gaming and workstation power supplies, laptop chargers above 200 watts, and SiC for industrial equipment. |
| Mobile and low-end consumer chargers | not disclosed; guided to insignificant by the end of 2026 | GaN chips for phone and small laptop chargers sold mainly in China; the 2021-2024 growth market, now being exited. |
Who runs it
Chris Allexandre has run Navitas since September 2025, after running the power-chip division at Renesas, which is now suing him and the company. Tonya Stevens became finance chief in March 2026, and the founders have left. Insiders own about an eighth of the shares, mostly early investors and the founder of the silicon carbide business, who has been selling; the chief executive owns very little. His bonus and new share awards ride on sales, none on profit or the value of each share.
The evidence under this - 5 claims
Renesas, Allexandre's former employer, sued Navitas, the CEO and another former Renesas employee for trade-secret theft in July 2026 (C16).
Singh sold 3.72M shares in late May 2026, after the record date, with no Rule 10b5-1 plan flag (C15), so the insider stake is now smaller.
No long-term award pays on profit, cash flow or anything per share, so the share issuance in C11 costs the plan nothing, while revenue bought with Claros (C13) would count toward cumulative revenue unless the committee excludes it.
Why it wins, and who could beat it
Why customers come back
What Navitas has is know-how: a gallium nitride chip that puts the switch and its control circuits on one piece of silicon, and a silicon carbide design good enough that Magnachip, another chipmaker, is paying to license it.
What it does not yet have is pricing power. NVIDIA's list puts it beside thirteen other suppliers, and its gross profit per dollar of sales is well below Power Integrations', a larger rival selling to the same buyers.
The evidence under this - 4 claims
- Wolfspeed alleges the GaN and SiC products infringe its patents (C16), so the IP is contested rather than clear.
- Thirteen other suppliers are on NVIDIA's list (C06) and adjusted gross margin is 39.5% against Power Integrations' 53.6% (C32), which is not what a protected product earns.
The threat
The threat is not a new technology but bigger companies with the same one. Infineon and onsemi make both materials in their own factories; Power Integrations now sells gallium nitride switches rated to 1,700 volts aimed at data-centre power supplies. All three sit on NVIDIA's list, and a power-supply maker likes to buy the same part from two sources.
The evidence under this - 3 claims
The rivals: POWI, ON, STM, WOLF, and 2 more that are private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | How far along | Priced against it |
|---|---|---|---|
| Infineon CoolGaN and CoolSiCPriced against it: Listed abroad | On sale | Listed abroad | |
| POWI | Power Integrations PowiGaNPriced against it: Yes | On sale | Yes |
| ON | onsemi EliteSiC, SiC JFETs and vertical GaNPriced against it: Yes | On sale | Yes |
| Innoscience GaNPriced against it: Listed abroad | On sale | Listed abroad | |
| STM | STMicroelectronicsPriced against it: Yes | On sale | Yes |
| WOLF | WolfspeedPriced against it: Yes | On sale | Yes |
How big it could get
How big this could get
Data centres are growing fast: TrendForce puts their power capacity at 161 gigawatts in 2026, up about 31% in a year. The chips that serve them are a much smaller pond.
Yole, an industry research firm, expects gallium nitride chips sold into data centres and telecom to reach a little over $380M a year by 2030, across every supplier. Navitas claims a $3.5B market for its products, silicon carbide included; nobody outside the company has sized that half.
The evidence under this - 4 claims
- Yole's figure is GaN only; the SiC that Navitas sells into rack power supplies and grid equipment is not in it, so it bounds one half of the business.
- Industry forecasts of this kind are revised often and are sold to the companies they describe.
- Yole sizes the whole 2030 power GaN market at about $3bn and the telecom and infrastructure part of it, where data centres sit, at a little over $380m (C18).
- Navitas' own filings say its SAM rests on assumptions about 800-volt adoption that could be wrong in size or timing (S02 risk factors).
Who has signed
Nothing with a volume. Sales run through distributors on cancellable orders, and when an analyst asked how much of 2027 is committed, the finance chief would not say. The U.S. Army picked Navitas in September 2026 for a development programme with no stated value, and Magnachip's licence has no disclosed fee.
The evidence under this - 4 claims
The results
The numbers
Sales fell 46% over the twelve months to Jun 2026, about the same pace as the −45% in the year to Dec 2025.
Operating margin is −317%, down from −149% in the year to Dec 2023.
Running costs take 355¢ of each sales dollar, up from 334¢ in the year to Dec 2021.
Shareholders' money lost 53% a year, down from 33% in the year to Dec 2022.
It used $67.8M of cash over the last twelve months, so the $557M it holds lasts about 8.2 years at that pace.
Yearly sales growth has ranged from −45% to +109% over the last 5 years, with sales falling in 1 of them.
The share count rose 514% in 4.5 years.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-04. Fiscal years end in December.
| Dec 2021 | Dec 2022 | Dec 2023 | Dec 2024 | Dec 2025 | 12 mo to Jun 2026 | |
|---|---|---|---|---|---|---|
| Size and growth | ||||||
| Sales | $23.7M | $37.9M | $79.5M | $83.3M | $45.9M | $36.5M |
| Sales growth | +100% | +60% | +109% | +4.8% | −45% | −46% |
| Profit | ||||||
| Gross margin | 45% | 31% | 39% | 34% | 31% | 38% |
| Operating margin | −289% | −326% | −149% | −157% | −235% | −317% |
| Net margin | −643% | 192% | −184% | −102% | −255% | −857% |
| What growth costs, per sales dollar | ||||||
| Research and development | 116% | 133% | 87% | 91% | 109% | 146% |
| Selling, general and admin | 216% | 207% | 77% | 75% | 77% | 109% |
| Pay in shares | 174% | 167% | 68% | 52% | 32% | - |
| Capital spending | 8.7% | 12% | 6.0% | 8.1% | 3.2% | 3.8% |
| Cash | ||||||
| Cash from operations | −$41.7M | −$44.5M | −$41.4M | −$58.8M | −$42.9M | −$66.5M |
| Free cash flow | −$43.8M | −$49.1M | −$46.2M | −$65.6M | −$44.4M | −$67.8M |
| Balance sheet | ||||||
| Cash and short-term investments | $268M | $110M | $152M | $86.7M | $237M | $557M |
| Debt | $6.92M | $0 | - | - | - | - |
| Net cash | $261M | $110M | - | - | - | - |
| Returns and shares | ||||||
| Return on capital | - | −55% | - | - | - | - |
| Shares (diluted) | 39.2M | 146M | 169M | 182M | 206M | 241M |
What growth costs
Most of Navitas' costs are engineers and sales staff whose pay does not shrink when sales do, so its losses move with revenue. When sales doubled in 2023, each dollar carried far less cost; when they halved in 2025, losses widened again. Growth would make it more profitable, but sales need to be about 3 times today's before it stops losing money, even leaving out stock pay.
The evidence under this - 2 claims
SG&A fell in dollars from $78.4M in 2022 to $35.2M in 2025 through two restructurings, so part of the ratio's fall was cutting, not scale (C29).
- Adjusted figures leave out stock pay, $8.4M in Q2 2026 (C30), and amortisation, so GAAP break-even sits higher still.
- Management expects gross margin to keep rising with mix (C32), which would lower the bar.
The last report (27 Jul)
Quarter ended June 2026, filed figures.
The June 2026 quarter looked like a turn. Sales rose 22% from the March quarter to $10.5M, and management guided the September quarter to $13.5M, which would be the first rise on a year earlier since the pivot.
Two things temper it. Sales were still 27% below a year before, and the gross profit per dollar barely moved, so the new mix is not yet earning more per chip. The reported loss of $228.2M was mostly a one-off accounting charge from the company's 2021 stock-market listing deal.
What it did to the thesisIt hinted at a turn but proved nothing. Sales are still below last year, the profit kept on each chip has not improved, and the data-centre growth is unmeasured. That keeps the view negative: the price assumes a boom this report does not yet show.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $-0.04 against $-0.04 expected; $-0.95 as filed, the gap being what the company adds back.
The evidence under this - 5 claims
- 2025 revenue fell 45% as the old business collapsed faster than the new one grew (C02), and Q2 2026 revenue was still 27% below a year earlier.
- The CFO would not say how much of 2027 is committed backlog (C36), so the guide beyond Q4 rests on programmes the company will not size.
- The CFO said on the same call that Navitas does not break revenue down by its high-power markets, so the growth rate cannot be checked against any filed figure.
- If AI infrastructure only reaches a third of about $15M of Q4 revenue, it is about $5M a quarter: fast growth from a very small base.
What management did with the money
Management sold shares while the price was high: $380.7M of new stock in May and June 2026 at about $21.89 each, nearly twice today's price. It is now spending some of it, and new shares, on Claros, a two-year-old start-up. Directors sold too, and none bought.
Insiders in 2026Q2: 6 purchases and 8 sales, from their Form 4 filings.
The evidence under this - 4 claims
- 2026-10-07editedthe deal closed on 2026-10-06 (8-K Item 8.01), so it is no longer pending; terms not restated at closing
The price, and where we disagree
In short. Priced for perfection. The price assumes Navitas becomes a large, profitable supplier of power chips to AI data centres. It could, but the price leaves nothing for the likelier middle outcome.
The shares have fallen about two-thirds from a May 2026 peak, and that can look like a chance. It is the second time: in October 2025 they doubled on NVIDIA news and fell back within three months.
The next real evidence comes on 2 November, with the September-quarter results, and in 2027, when NVIDIA's 800-volt racks are due. Two lawsuits, from Wolfspeed and Renesas, also run through that period.
The evidence under this - 4 claims
- An analyst on Navitas' Q2 2026 call cited market reports that the Kyber rack may have been cancelled and replaced; the CEO declined to comment on NVIDIA's plans (S04).
- Yole puts first commercial 800-volt rollouts only 'around 2027' (S12).
- 2025 revenue fell 45% as the old business collapsed faster than the new one grew (C02), and Q2 2026 revenue was still 27% below a year earlier.
- The CFO would not say how much of 2027 is committed backlog (C36), so the guide beyond Q4 rests on programmes the company will not size.
How we read the multiple
Counting its cash, buyers pay about 69 times Navitas's yearly sales. The rivals its own annual report names sell for about 4 times, and its own usual is about 14.
For its price to look ordinary beside those rivals:
- Navitas would need about $400M to $580M of yearly sales.
- That is 11 to 16 times today's sales.
- The five analysts who forecast 2028 expect $116.2M.
- Even then, the price would be about 22 times sales, far above every rival.
Against its own history. Far dearer than its own past, while sales have shrunk, so the rise is enthusiasm about the future rather than a better business today.
The evidence under this - 4 claims
- A fast grower can deserve far more than a mature one's multiple, and investors pay forward, not trailing: at consensus 2028 revenue the figure is 21.6 times (C27).
- The cohort's members sell mostly silicon into cars and industry; a pure AI-power name may carry a premium for years.
- Forecasts for a company this early are wide: single analysts publish $200M for 2029 and $500M for 2030, at which the multiple would be 12.6 and 5.0.
- A buyer paying for 2030 rather than 2028 is paying roughly what the filed rivals trade at today, if that $500M arrives.
Priced against its rivals
The competitors its own annual report names, each confirmed against the rival's filing. Measured on company value for each $1 of sales, cheapest first:
| Ticker | Company | company value for each $1 of sales |
|---|---|---|
| STM | STMicroelectronics | 3.6× |
| WOLF | Wolfspeed | 3.8× |
| ON | ON Semiconductor | 4.9× |
| POWI | Power Integrations | 6.3× |
| NVTS | Navitas Semiconductor | 68.7× |
The middle of the other 4 is 4.3. Navitas Semiconductor Corp is dearest of 5.
Rivals' figures come from a data provider, so read the spread of the group rather than any single pair.
Far dearer than every rival its own annual report names. That position is its own, not the group's.
Also checked and set aside: Texas Instruments Incorporated (a substitute, not the same product); Qorvo (a substitute, not the same product).
The bet: what it pays if it works
Even if this works, the shares look expensive. The best case we can defend is one analyst's sales forecast for 2029, several times today's sales. That needs Navitas to win a large slice of a market every rival is chasing, and the company has not said how much of next year's demand is signed.
If it fails, shareholders are left mostly with cash, and that cash shrinks each year the losses and share awards continue.
| Revenue in two to three years | Points of that market needed | Valued like its peers | Like the cheapest |
|---|---|---|---|
| $116.20M | about 21 at 100% kept | 0.2× | 0.2× |
| $200.00M (the big win) | about 43 at 100% kept | 0.4× | 0.3× |
| $240.00M (market forecast) | about 54 at 100% kept | 0.4× | 0.3× |
| $400.00M | about 96 at 100% kept | 0.7× | 0.6× |
Yole and TrendForce put non-consumer power GaN near $1 billion and SiC outside cars near $2.6 billion in 2029; holding its 16.5% GaN share and taking 3% of that SiC gives Navitas about $240 million. On that forecast, $240.00M pays 0.4× from today's value: an Expensive bet on the market case.
To pay 2× (a Fair bet) it would need about $1.14B of revenue (about 290 points of that market); to pay 5× (a Cheap bet), about $2.85B of revenue (about 740 points of that market), at the same multiple.
The steps it has to take
- Done Named on NVIDIA's list of 800-volt power-chip suppliers
- Done Two quarters at or above guidance under the new CEO
- In progress Q3 2026 revenue at or above the $13.5M guide, adjusted gross margin at or above 39.7% due 2026-11-02
- Done Claros acquisition closes
- In progress A second GaN wafer source qualified before TSMC stops making GaN in July 2027 due 2027-03
- Not started A named production order or design win with volume for 800-volt racks due 2027-06
- Not started Quarterly revenue of $30M with adjusted gross margin at or above 45% due 2027-12
- Not started 800-volt power native to the GPU tray in production, where GaN is needed due 2028-06
The detail behind the label
The evidence under this - 6 claims
- Yole's figure is GaN only; the SiC that Navitas sells into rack power supplies and grid equipment is not in it, so it bounds one half of the business.
- Industry forecasts of this kind are revised often and are sold to the companies they describe.
- Management is raising operating costs by $1.0M to $1.5M a quarter from Q3 and talks of capacity reservations with foundries, so the burn will not stay at the first-half rate.
- Claros is a pre-revenue start-up whose own losses join the group at closing and are not disclosed.
Where we disagree with the market
Navitas is an early winner of NVIDIA's 800-volt shift, and 2027 is when the revenue arrives.
The shift is real and Navitas is in it, but as one of fourteen suppliers, with nothing signed and no pricing power yet.
The market reads NVIDIA's supplier list and management's timetable. The filings show one distributor carrying most sales, no disclosed volume, and a company selling stock into each rally.
What would prove us wrong: quarterly sales reaching $30M in 2027, with gross profit per dollar rising and a named data-centre customer placing a production order.
What analysts say, and the evidence
The evidence under this - 6 claims
What could go wrong, and how we would know
The risks
- NVIDIA's timetable slips. Most of the growth case is dated to 800-volt racks in 2027; an analyst asked on the call whether NVIDIA's Kyber rack had changed.
- Two lawsuits. Wolfspeed seeks to stop sales of chips it says copy its patents; Renesas accuses the chief executive of taking trade secrets.
- One distributor. A single distributor handled 71% of sales, and Navitas lost its previous main distributor in 2024.
- A forced change of factory. TSMC stops making gallium nitride in July 2027, so the chips must be requalified at new factories during the ramp.
- More shares. The share count keeps rising from stock pay and the Claros deal, so any future profit is split more ways.
- Rivals with factories. Infineon, onsemi and Power Integrations sell the same kind of chips to the same buyers.
The evidence under this - 6 claims
- An analyst on Navitas' Q2 2026 call cited market reports that the Kyber rack may have been cancelled and replaced; the CEO declined to comment on NVIDIA's plans (S04).
- Yole puts first commercial 800-volt rollouts only 'around 2027' (S12).
What we still do not know
- How much of 2027's sales are backed by signed orders, and from which customers.
- How much silicon carbide goes into each data-centre rack; no outside firm publishes it.
- What the Wolfspeed patents cover, and whether a court could stop any product line.
- What Claros sells today and how much cash it burns.
The evidence under this - 4 claims
- 2026-10-01editedthe 2026-10-01 market case found Yole's $200M SiC data-centre opportunity (C45), and TrendForce's combined SiC-and-GaN share (C48), so outside SiC figures exist; narrowed to what they lack
- 2026-10-07editedthe deal closed on 2026-10-06 (8-K Item 8.01), so it is no longer pending; terms not restated at closing
What would change our view
- Unattractive→AttractiveThe ramp shows up in revenue, at a price short of assuming it
It needs all of:
- ○Quarterly sales reach $30M, with gross profit per dollar of sales at 45 cents or more.checked at the next results, due by 2 Nov
- ○The company's value falls to 15 times its last year of sales.about ≈$2.56 today, 78% below the last close
- Unattractive→NeutralThe 2027 demand is signed
It needs:
- ○Navitas names a production order with a volume from a data-centre operator, GPU maker or power-supply maker for 800-volt racks.watched as it happens
We would also be wrong if
- A court orders Navitas to stop selling a gallium nitride or silicon carbide product line.
- Sales fall from one quarter to the next in 2027.
What is coming
- 2 November 2026 Q3 2026 results: Revenue against the $13.5M guide, adjusted gross margin against 39.7%, a Q4 outlook of $15M or more, and whether 2027 committed demand is finally quantified.
Every dated event ahead
- 2 November 2026 Q3 2026 results: Revenue against the $13.5M guide, adjusted gross margin against 39.7%, a Q4 outlook of $15M or more, and whether 2027 committed demand is finally quantified.
- March 2027 First qualified GaN product from Powerchip or GlobalFoundries: Whether Navitas can replace TSMC's wafers before TSMC stops GaN; a slip would starve the data-centre ramp of supply.
- June 2027 800-volt sidecar racks begin ramping: Whether the design wins Navitas claims turn into production orders against thirteen other listed suppliers.
- July 2027 TSMC stops making GaN: Whether the buffer wafers and the new foundries carry supply without a gap or a margin hit.
Since we wrote this
- 7 OctNewsClaros acquisition closes on the terms announcedThe deal the thesis already counted as pending is done. Cash and shares go out as planned; Claros revenue and cash burn are still undisclosed, so the view holds.
- 6 OctNewsNavitas and Microchip announce a shared 800-volt power design for AI data centers. View heldNavitas and Microchip announce an 800V reference design for AI data centers.
- 5 OctThesis reviewThe bet now names what it ridesThe owner asked that every bet name the macro trend behind it, after BWXT's bet was missing its chief executive's case for nuclear over gas.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the proxy and filings.
- 29 SepThesis writtenUnattractive, medium convictionWritten two months after Q2 2026 results and the day after the Army award, with the shares down two-thirds from May.
Also checked 3 times with nothing new to act on.
Sources 31
Sources - 31
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-02-27Item 1 business, the Navitas 2.0 pivot, fabless manufacturing (TSMC, Powerchip, GlobalFoundries for GaN; X-Fab for SiC), the competition section naming eleven rivals, 190 employees, TSMC's exit from GaN in July 2027, distributor dependence, the 2024 termination of Distributor A and its $7.5M bad debt, FY2025 revenue of $45.9M against $83.3MbacksC01C02C11C17C28C29C34C37C50C55C56
- S02primarycompany release2026-07-27Q2 2026 revenue, GAAP and non-GAAP gross margin and operating loss with the reconciliation, the $203.1M earnout charge, $557.4M of cash, the Q3 2026 guide, the claim that AI infrastructure will pass a third of sales by year end
Third quarter 2026 net revenues are expected to increase to $13.5 million, plus or minus $0.5 million, which at the midpoint represents 28% sequential growth and would mark a return to year-over-year growth.
backsC03C04C30C32C35 - S03primarycompany filing2026-07-27customer concentration (Distributor A 71% of Q2 revenue), revenue by region and by end-customer country, the two at-the-market programmes ($125.0M and $255.8M gross), the earnout settlement and $229.9M of shares issued for it, shares outstanding (261,093,488 at July 24, 2026; 230,525,464 at December 31, 2025), cash flows, the prepaid wafer build, the Wolfspeed and Renesas suits, the Magnachip partnership and its related-party statusbacksC03C09C10C11C12C16C17C22C24C37C39C50C55C56
- S04primarytranscript2026-07-27Q2 2026 earnings call transcriptdata/cache/transcripts/NVTS-2026Q2.json (FMP earning-call-transcript, quarter label 2026Q2, call held 2026-07-27)the four 800-volt inflection points and their dates, AI infrastructure growing over 50% a quarter, the $21.89 average price of the Q2 share sales, the TSMC buffer prepayment, the answer on committed 2027 demand, the Kyber question, the Wolfspeed and Renesas answer, pricing, and five analysts' questions (Needham, Rosenblatt, Morgan Stanley, Baird, Craig-Hallum)
We are breaking out what percent is committed in 2027 or what percent relates to our backlog.
backsC04C05C08C10C17C36C51 - S05primarycompany release2026-08-25deal value, consideration mix, the $12.97 reference price, the $28.9M of performance share awards, the claimed $8 billion 2030 SAM ($3.5B GaN and HV SiC, about $1B JFET, at least $3.5B VPD and IVR), the 2028/2029 contribution timing
Navitas will acquire Claros in a transaction valued up to approximately $232.8 million, comprised of approximately $216.0 million to be paid at closing in a combination of cash and shares of the Company’s Class A common stock
backsC13C14 - S06primarycompany filing2026-09-086,912,729 closing shares and up to 1,289,869 earnout shares issuable to Claros holdersbacksC13
- S07primarycompany release2026-02-24the $3.5 billion 2030 SAM, Q4 2025 revenue of $7.3M, the November 2025 private placement ($95.6M net), the GlobalFoundries partnership, the Q1 2026 guide, WT Microelectronics and Avnet as consolidated distributors
First quarter 2026 net revenues are expected to increase to between $8.0 million and $8.5 million.
backsC14C35C39 - S08primarycompany filing2026-09-21Navitas buying 1,461,988 Magnachip shares at $3.42, about $5.0MbacksC22
- S09primarycompany filing2026-08-2625 open-market sales of 7,925,392 shares, 20 of them (7,708,814 shares) with no Rule 10b5-1 flag; no open-market purchases; director Ranbir Singh's 3,060,118 shares at $29.29 and 664,058 at $28.72 on 27-28 May 2026backsC15
- S10primaryindustry data2025-10-13the fourteen silicon suppliers NVIDIA lists for its 800-volt architecture, and the Kyber rack's 64:1 converter stepping 800 VDC to 12 VDC beside the GPU
The NVIDIA Kyber rack architecture uses a 64:1 LLC converter to step 800 VDC down to 12 VDC adjacent to the GPU ... Silicon providers : AOS, Analog Devices, Efficient Power Conversion, Infineon Technologies, Innoscience, MPS, Navitas, onsemi, Power Integrations, Renesas, Richtek, ROHM, STMicroelectronics, Texas Instruments.
backsC06C07 - S11primaryindustry data2025-05-20NVIDIA Technical Blog - NVIDIA 800 VDC Architecture Will Power the Next Generation of AI Factories NVIDIAthe 2027 timetable for 800 VDC, the 13.8 kV to 800 VDC conversion at the perimeter, 45% less copper than 415 VAC
Full-scale production of 800 VDC data centers will coincide with NVIDIA Kyber rack-scale systems in 2027, ensuring seamless scalability for increasingly demanding AI models.
backsC06C07 - S12secondaryindustry data2025-10-29power GaN device market of $355m in 2024 rising at a 42% CAGR to about $3bn in 2030; telecom and infrastructure GaN above $380m by 2030; first 800-volt rollouts around 2027. yolegroup.com returned HTTP 403 to curl and to an agent fetch, so the text was read from the reprint, which quotes the release
The power gallium nitride (GaN) device market is rising at a compound annual growth rate (CAGR) of 42% from $355m in 2024 then growing sixfold from now to about $3bn by 2030 ... The aim is to integrate GaN devices into 800V HVDC power systems. These partnerships mark the beginning of large-scale GaN deployment, with Yole Group anticipating first commercial rollouts around 2027. The telecom & infrastructure segments collectively are hence expected to grow at a 53% CAGR to more than $380m in GaN revenues by 2030, making data centers one of the most promising growth pillars of the power GaN market.
backsC18 - S13primaryindustry data2026-09-16data-centre power capacity for 2025 and 2026 and the growth expected in 2027, and that operators are planning high-voltage DC architectures
Global data center power demand capacity reached 122.9 GW in 2025 and is estimated to rise to 161 GW in 2026. In 2027, continued CapEx expansion by CSPs and sustained overall server shipments are expected to keep annual growth in data center power demand capacity above 30%.
backsC19 - S14primarycompany filing2026-02-06Power Integrations Form 10-K, fiscal year ended December 31, 2025 Power IntegrationsPowiGaN transistors to 1700 volts, higher-power GaN products aimed at AI data-centre power supplies, the Odyssey vertical-GaN purchasebacksC20
- S15primarycompany filing2026-02-09silicon, silicon carbide and GaN designed and made in-house, a portfolio for the complete AI data-centre power tree, vertical GaN, and the January 2025 purchase of Qorvo's silicon carbide JFET businessbacksC21
- S16primarycompany filing2026-08-10Magnachip Semiconductor Form 10-Q, quarter ended June 30, 2026 Magnachip SemiconductorMagnachip's own account of licensing Navitas' SiC technology and porting it to its fabbacksC22
- S17secondaryvendor data2026-09-29Research pack, frozen at content/NVTS/research-pack.jsoncontent/NVTS/research-pack.jsonprice, market value and enterprise value, the filed-competitor cohort multiples, the price-move decomposition, daily bars, analyst estimates and ratings, FINRA short interest, the headcount seriesbacksC23C24C26C28C29C32C37C38C40
- S18secondarynews2026-09-28the award, its sponsor (Army Research Laboratory, JETX) and scope; no value. GlobeNewswire closes the connection without a response, so the reprint was read
The ALATTIS prototype project will develop, execute, and validate a novel domestic manufacturing process for ultra-high-voltage (≥10 kV) silicon SiC power semiconductor devices through iterative design, fabrication, and testing.
backsC41 - S19primarycompany filing2026-08-20Wolfspeed's own description of its silicon carbide MOSFETs, diodes and modulesread, but no claim rests on it
- S20primarycompany release2026-07-27the four 800-volt inflection points with dates, AI infrastructure as 80% of the $3.5B SAM, the quarterly non-GAAP gross-margin series (38.7%, 38.7%, 39.0%, 39.5%), the 10 kV award to be announced
Q4 2026 sales expected to be comprised of roughly one-third AI Infrastructure and an insignificant portion relating to Mobile
backsC04C32 - S21primarycompany filing2026-09-21the per-share price of $3.42 and 1,461,988 sharesbacksC22
- S22primaryindustry data2025-08-13a second analyst house's power GaN forecast to 2030, independent of Yole's; read through ir_fetch
According to TrendForce, the GaN power device market is projected to grow from USD 390 million in 2024 to USD 3.51 billion by 2030, representing a compound annual growth rate (CAGR) of 44%.
backsC43 - S23primaryindustry data2024-08-16the previous year's TrendForce GaN forecast, 25% higher for 2030 than the next year's, and the non-consumer share of the market in 2023 and 2030; read through ir_fetch
In 2023, the market size of global GaN power device was around USD 271 million, and it is projected to grow to USD 4.376 billion by 2030 at a compound annual growth rate (CAGR) of 49%. Notably, the proportion of non-consumer applications is expected to increase from 23% in 2023 to 48% by 2030, with automobile, data center, and motor drive being the core application scenarios.
backsC43C44 - S24secondaryindustry data2025-07-01Yole's SiC device market for 2030, the automotive share of it and its data-centre SiC opportunity; yolegroup.com answered HTTP 202 with no text to ir_fetch, so read from the reprint, which quotes the release
Despite a slowdown in EV sales which is delaying demand for SiC devices and wafers, Yole reckons the SiC market will have a 20% CAGR 2024-30 to reach $10.3 billion in 2030. Despite the slowdown, the automotive & mobility segment is expected to retain about 70% of SiC demand over the next 5 years. Yole also sees a: $200 million market opportunity for SiC in datacentres.
backsC45 - S25secondaryindustry data2025-12-18Semiconductor Today - Power SiC faces overcapacity downturn until 2027-2028 (Yole), 18 December 2025 Yole GroupYole's later SiC read: an overcapacity downturn to 2027-2028 and device revenue of nearly $10bn by 2030; read through ir_fetch
The downturn is expected to persist until 2027-2028. ... After 2027-2028, renewed growth will come from 8-inch production platforms and next-generation trench and superjunction MOSFETs, reckons Yole. Device revenue forecast to reach nearly $10bn by 2030.
backsC45 - S26primaryindustry data2025-11-27supplier shares of the 2024 power GaN market from TrendForce's own research; read through ir_fetch
According to TrendForce, Navitas holds a 16.5% share of the global GaN power device market in 2024, trailing only Innoscience (29.9%) and ahead of EPC (12.4%), Infineon (10.3%), and Power Integrations (9.8%).
backsC46 - S27primarycompany filing2026-03-27the GaN share leader's own revenue, data-centre GaN sales, wafer capacity and planned expansion, and its place in NVIDIA's 800-volt supply system; a Hong Kong exchange filing, fetched with ir_fetch's client and read with pdftotext
During the Reporting Period, the Company realized a sales revenue of RMB63.19 million for GaN chips in the AI and data center fields, representing a year-on-year increase of 50.20% ... By the end of 2025, the Group's wafer production capacity had steadily increased to 20,000 wafers per month ... Expand our production capacity of the 8-inch GaN wafer (from 12,500 wafers per month as of June 30, 2024 to 70,000 wafers per month over the next five years)
backsC47 - S28primaryindustry data2025-11-27TrendForce press release - AI to Reshape the Global Technology Landscape in 2026, 27 November 2025 TrendForcethe rate at which wide-bandgap chips are displacing silicon in data-centre power; read through ir_fetch
TrendForce predicts that the adoption of SiC and GaN in data center power systems will reach 17% by 2026 and exceed 30% by 2030.
backsC48 - S29primarycompany filing2026-05-11executive officers and their biographies, the 2025 summary compensation table, the CEO's employment terms and 2026-2028 revenue PSUs, the annual bonus metrics, the lapsed founder LTIP options, beneficial ownership at the April 28, 2026 record datebacksC52C53C54
- S30primarycompany filing2024-03-062022 and 2023 revenue, cost of revenue, R&D and SG&A from the income statement, read against the frozen numbers cardbacksC56
- S31primarycompany filing2026-10-06closing date 2026-10-06; management's claim that the deal doubles its identified 2030 SAM to over $8B; webinar on 2026-10-20
closed the transactions contemplated by the previously announced Agreement and Plan of Merger
backsC13