Navitas’ 2,300V SiC Module: What Samples Prove Before Grid Revenue

Navitas’ new 2,300V silicon-carbide module broadens the technical case for its grid and energy business, but sample availability does not yet establish customer adoption or incremental revenue. The investment question is whether a stronger device offering can pass qualification and design selection quickly enough to become profitable volume shipments. The October 21 announcement supplies product evidence; it does not supply that commercial bridge.

Research checked October 22, 2026, Korea time. The announcement was published October 21 at 8:30 a.m. Eastern Daylight Time. Financial figures below are historical consolidated U.S. dollar amounts, with periods stated; valuation scenarios remain dated AHNIM STUDIO assumptions.

What the 2,300V announcement actually establishes

In its issuer release, Navitas announced a 2,300V full-bridge SiC module family, G4H08MT23GB4, with samples immediately available. It advertises 8 milliohms of on-resistance, standard and high-current pin options, and an optional thermal-interface material. The release says the family will be qualified to its internal AEC-Plus benchmark. That wording describes a future qualification step, not completed third-party certification.

A Navitas-authored technical presentation hosted by Mouser labels the resistance per switch and also shows the part as sampling. October 21 should therefore not be treated as proof of its first-ever availability. We could not retrieve the device-level datasheet in this review, so we do not assign unverified gate-drive, junction-temperature or current test conditions to the advertised resistance.

The release does not identify a customer, order value, selling price, volume-ramp date or quantified revenue uplift. Packaging claims remain company claims. A component improvement cannot be translated directly into an equal improvement in complete-system efficiency or lifetime.

Grid conversion and rack conversion are different opportunities

The earlier 24kW 800V-to-12V board analysis addressed power conversion inside an AI infrastructure architecture. This announcement emphasizes a higher-voltage SiC module for energy and grid applications. The devices can participate at different conversion stages, with different design cycles, qualification requirements and competitive alternatives. A shared AI-power narrative does not make the revenue pools interchangeable.

NVIDIA’s October 13, 2025 architecture discussion describes a wider 800V DC ecosystem. It is useful technical background, rather than an order announcement for this module. Utility-side conversion, facility distribution and conversion near computing loads must each earn their own design selection.

Lower on-state resistance can reduce conduction loss, which scales approximately with current squared multiplied by resistance. Yet switching losses, gate drive, operating temperature and the thermal path also affect system performance. A credible comparison needs matching topology, operating conditions and cooling assumptions. The advertised resistance alone cannot establish lowest total cost.

Competition already exists at this voltage class. Infineon announced 2,300V XHP 2 modules in May 2026, while Wolfspeed lists a 2,300V WolfPACK half-bridge. Different bridge configurations and packages prevent a simple resistance ranking. Navitas must win on the buyer’s complete power-stage requirements, availability and economics.

The commercial evidence ladder

Stage	Evidence as of this review	What would change the investment case
Samples	Availability established by the issuer	Evaluation access; not a production order
Qualification	AEC-Plus described as a future step	Completed tests, conditions and customer acceptance
Design selection	No named win disclosed in this release	Selected platform, scope and expected ramp
Volume shipments	No ramp or unit economics disclosed	Recognized sales, margin and repeat shipments
Cash generation	No product-specific cash contribution disclosed	Collections after development and operating costs

These stages can overlap in practice, but none automatically completes the next. Samples may improve the probability of a design win. They do not tell us how many buyers will qualify the module, how much content Navitas will capture, or how long evaluation will take.

Why the historical cash balance needs a dilution footnote

The June 30, 2026 Form 10-Q reported Q2 revenue of $10.529 million: up 22.5% from $8.598 million in Q1, but down 27.3% from $14.490 million a year earlier. GAAP gross margin was 0.4%; adjusted gross margin was 39.5%, using the company’s non-GAAP reconciliation. Those measures should not be mixed.

At June 30, cash and equivalents were $557.409 million. During the first half, operating cash outflow was $48.348 million and property-and-equipment purchases were $0.615 million. Net proceeds from at-the-market share sales were $373.165 million. That financing strengthened liquidity while diluting shareholders; the cash balance is neither operating cash generation nor an October cash estimate. It predates the subsequent Claros closing and further spending.

The reported $228.218 million Q2 GAAP net loss included a $203.068 million noncash earnout remeasurement. Treating that loss as the quarter’s cash burn would be misleading. The cash-flow statement provides the better starting point for funding analysis.

Management’s July 27 outlook called for Q3 revenue of $13 million–$14 million and adjusted gross margin of 38.7%–40.7%. Those are company forecasts, not achieved results. The module announcement does not itself revise them.

Valuation: preserve the horizon and avoid counting the same opportunity twice

The October 11 fundamental analysis uses model AHNIM-NVTS-LT-20261011-R1. Its conditional December 2035 success scenarios are $4.80 for Partial success, $16.87 for Scaled success and $36.35 for Strong leadership. Failure could produce values below all three success scenarios. Discounted at 14% over 9.221081 years, the corresponding October 11, 2026 equivalents are $1.43, $5.04 and $10.86. These are not 12-month price targets. The dated scenarios and their original assumptions remain unchanged.

The earlier 2027 revenue bridge illustrates the sensitivity: 35 million kW × 100% eligibility × 15% adoption × $25 per kW × 20% Navitas capture = $26.25 million of modeled inside-800V revenue. Adding $55 million for conventional AI and other markets gives $81.25 million. These are AHNIM assumptions, not management guidance. Raising adoption alone from 15% to 20% adds $8.75 million, illustrating how much a commercialization input matters. The new announcement does not justify that change.

Grid products must be mapped to the existing model’s revenue categories before any increment is added. Reusing the same infrastructure spending in a second product bucket would overstate the opportunity. A verified customer program, realistic content per system, achievable share and a delivery schedule would support a numerical update; a product label does not.

MarketBeat’s October 21 snapshot showed a $15.92 average 12-month target. Its six ratings included five visible numeric targets averaging that amount. The latest visible Rosenblatt entry was a September 29 Neutral/$13 reiteration, preceding this launch. We did not verify a launch-related broker target change. No comparable December 2035 broker average was available, so a same-horizon 20% discrepancy test cannot be established.

What to check next

The next useful evidence is completed qualification with defined conditions, a selected customer platform, shipment timing and a revenue-and-margin bridge. Quarterly disclosures should then distinguish growth from legacy products, acquired operations and new organic grid wins, while updating cash use and the share count. The product strengthens the technical opportunity; the commercialization timetable remains the decisive uncertainty.