TrumpGrift Capital
Equity Investing Commentary
Navitas Semiconductor's Nvidia deal is worth far less than the hype suggests: even with the 40% share people expect, it might add only about $30M in total revenue. Wolfspeed already has almost double that in annualized data-center revenue today, before any 800V data centers or power-grid upgrades. x.comÂ
Navitas is exposed to a real thesis, power delivery as AI's next bottleneck, through a high-risk stock. Rack power is heading from 150kW to over 1,000kW, and Navitas has full-stack GaN and SiC, an 800V HVDC design partnership with Nvidia and a 96.5%-efficient 800V-to-6V board. But Q4 2025 revenue fell 59% to $7.3M, it trades near 211x trailing EV/revenue, loses money, revenue inflection is 2-3 years away, and Infineon, Vicor and Delta got similar Nvidia mentions. Valuation conviction is low. … Show more
Navitas is a one-customer concentration story facing heavy competition from tier-one suppliers. Its GaN portfolio winning Nvidia adoption played out, but STMicroelectronics and Infineon also supply Nvidia and are strong in GaN and SiC. GaN output may quadruple by 2030, yet SiC prices were hammered by Chinese capacity, unresolved, so revenue and margins could diverge sharply. Power and thermal efficiency are better played across these names than through Navitas alone. x.comÂ
Navitas Semiconductor is one of the favorite names of 2026, pitched about 15 times this year, and it has risen 47% since the April 8 thesis. x.comÂ
Navitas's licensing deal with Magnachip won't justify its elevated valuation any time soon. No financial terms were disclosed; typical IP licensing in this space brings $5M-$15M upfront plus 3-7% royalties, 100%-margin cash but small against Navitas's multiples. Magnachip needs 12-24 months to qualify and scale the silicon carbide technology in its Korean fab, making this a long-term build rather than a catalyst for next quarter. x.comÂ
Navitas remains a wait-and-see name because its high-voltage products are years from real revenue. The US Army selected it for ALATTIS, a program to build domestic 10 kV silicon carbide power devices, which validates its engineering and fits its roadmap toward solid-state transformers for AI data centers alongside its Nvidia 800 V work. But it is a prototype program with no disclosed value, and 10 kV devices are years from volume revenue. x.comÂ
Navitas and Wolfspeed look attractive after falling 63% and 61% from this year's peaks. The 800V theme was crushed by rumors of an Nvidia delay, but Nvidia's confirmation and ramping ASICs can change the narrative fast. x.comÂ
Silicon carbide is becoming the bottleneck material for the AI power stack, and Wolfspeed owns the only 200mm SiC wafer fab in the world. Capacity for SiC wafers is structurally constrained, and Wolfspeed is preparing for 300mm wafers in America, which supply-chain risk could make a strategic priority. x.comÂ
Wolfspeed has clear upside if it executes, given its 300mm platform and growth drivers in data centers and solid-state transformers. It is a more speculative play, carrying gross margin and execution risks. x.comÂ
Navitas is an option on Nvidia's 800V DC data-center architecture, where GaN handles the step-down from 800 volts to processor level inside the rack. The design halves power conversion stages so racks can reach the up to 8 megawatts Rubin Ultra needs, with full deployment alongside Kyber in 2027. Navitas, on Nvidia's partner list, is the only US-listed pure-play GaN company. Revenue is minimal until the Kyber ramp, and Nvidia's 29-company partner list means hard pricing pressure. x.comÂ
Wolfspeed is no longer that speculative and deserves a chance as a core position, with upside-skewed probabilities. Valuation isn't demanding at all after the stock fell from the $70s to the $50s, and it has a huge asset base as demand inflects. Catalysts include 800V data centers, TSMC's 300mm silicon carbide interposers and the grid. x.comÂ
