Navitas Semiconductor (Nasdaq: NVTS) designs power semiconductors, not processors. Its gallium nitride and silicon carbide chips sit in the conversion stage that turns grid voltage into something a server rack can use.
Anyone researching NVTS stock should start there. The company sells into the AI buildout without ever touching AI compute itself.
What the company actually sells
The Q2 2026 10-Q describes the product line as GaN power ICs, high-voltage SiC devices, high-speed silicon system controllers and digital isolators, all used in power conversion and charging.
GaNFast power ICs put drive, control and protection on one chip. The FY2025 10-K claims 100x faster switching and up to 40% energy savings against incumbent silicon. GeneSiC, acquired in August 2022, adds trench-assisted planar MOSFETs rated to 6.5kV.
Navitas is fabless. It owns no wafer plant, which holds fixed costs down and puts supply in someone else’s hands. That is the same trade running through the AI component layer, from power chips to optical interconnect suppliers.
Navitas 2.0 and what the pivot costs
In the fourth quarter of 2025 the company announced a restructuring it calls Navitas 2.0. It pairs a 19% workforce reduction with a deliberate retreat from mobile charging, low-end consumer electronics and China-based segments.
The Q2 2026 filing shows both sides of that trade. Net revenues fell to $10.5 million from $14.5 million a year earlier, a 27% decline the company attributes mainly to lower mobile sales in Asia and consumer markets.
Margin moved the other way. Cost of revenues dropped to 61% of revenue from 84%, which the filing credits to sales mix shifting toward high-power products.
Research and development ran at 125% of revenue for the quarter and 145% across the six months. Navitas spends more on engineering than it collects in sales, and the filing expects operating losses and negative cash flow to continue.
The foundry problem underneath the thesis
Navitas historically used TSMC as its sole supplier of GaN wafers. On 1 July 2025 TSMC announced it will stop GaN production in July 2027, which reaches far past one customer and fits the pattern in TSMC’s foundry positioning.
The stated mitigation is buffer inventory bought from TSMC, expanded work with Powerchip, and a long-term GlobalFoundries partnership announced in November 2025. SiC wafers are already made in the United States by X-Fab.
Navitas frames that domestic footprint as an advantage in markets with national security exposure. The 10-K also says there is no assurance it can source enough wafers on the cost and timeline it needs.
Concentration is the number most people miss
One distributor accounted for 71% of net revenues in Q2 2026. Hong Kong accounted for 76% of revenue by region.
The channel has not caught up with the strategy yet. The 800VDC architecture Navitas targets rests on solid-state transformers converting 11 to 13kV AC, the same grid-side buildout visible in GE Vernova’s power equipment orders.
Cash is not the near-term constraint. Cash and equivalents stood at $557.4 million on 30 June 2026 against $236.9 million at year end, after $255.8 million raised under a $500 million at-the-market program. Operating activities used $48.3 million over six months.
Is Navitas profitable?
No. The company reported a $27.2 million loss from operations for Q2 2026 and $55.0 million for the first six months, and states it expects net operating losses to continue.
Why does the TSMC GaN exit matter?
GaN wafers came from a single supplier that is leaving the business in July 2027. Qualifying GlobalFoundries and Powerchip takes time, and the 10-K lists delays there as a risk to orders, costs and market share.
How does Navitas fit into AI data center power?
Its GaN and SiC parts go into 800VDC delivery and solid-state transformers that take 11 to 13kV AC down to 800 or 1500VDC. Adoption depends on 800V systems themselves being adopted, which the company lists as a forecasting risk.
Daniel Reyes is a markets writer for S4Tips covering the AI infrastructure and semiconductor supply chain. He focuses on the companies that build and power the AI compute stack. His articles are for information only and are not financial advice.