Vicor's patent fight is setting the price of the $233 million power layer Navitas just bought
Navitas closed its Claros purchase on October 6, 2026; three days later the US International Trade Commission opened Vicor's case against vertical power delivery. Navitas is not a respondent, but the layer it bought is the layer Vicor wants licensed, at royalties that climb as enforcement advances.
Richard Tang · 3 min read
Navitas Semiconductor−5.58% — Navitas Semiconductor, down 5.58 percent today, a maker of gallium-nitride power chips, closed its purchase of Claros on October 6, 2026, paying about $232.8 million for a developer of integrated voltage regulators, the converters that sit closest to an AI processor 1. Three days later the US International Trade Commission voted to investigate vertical power delivery, the hardware that pushes thousands of amps through the last millimetres of a rack, on a complaint from Vicor−8.35% — Vicor, down 8.35 percent today, a Massachusetts power-module maker 2. A day before Navitas closed, German rival Infineon bought into the same layer 1.
The bottleneck in AI power has moved beneath the processor
The rebuild runs from the grid to the chip. Racks are moving to 800-volt DC distribution, and Navitas already builds the NVIDIA+0.35% — NVIDIA, up 0.35 percent today co-developed board we covered earlier this week, converting 800 volts to 12 in a single stage on the compute motherboard 5. The stage after that is the hard one: conventional regulator modules push power sideways across the board, while Claros puts the converter beneath or inside the processor package, cutting the distance current travels from inches to millimetres 1.
Infineon's target, C2i Semiconductors, makes the multiphase controllers for the same final conversion 1. Two incumbents buying the same link within a week is the industry naming its next constraint.
Vicor claims that link, and it has collected before
Vicor filed its complaint on September 9, and the Commission's vote named a directory of AI rack manufacturing: Foxconn+1.88% — Foxconn, up 1.88 percent today, Quanta, Delta, Celestica−2.62% — Celestica, down 2.62 percent today and Flex−5.46% — Flex, down 5.46 percent today, plus Infineon and Monolithic Power Systems−0.26% — Monolithic Power Systems, down 0.26 percent today 2. Navitas is not a respondent, but its route into AI racks runs through companies that are. Vicor says any OEM or hyperscaler importing systems built on respondents' parts needs a license 3.
Vicor's published royalty escalator climbs from 1X toward 7X as enforcement reaches an import ban enforced by US Customs 4. An earlier Vicor case ended in exactly such a ban, and in September 2026 Vicor signed its first vertical-power license with a leading AI OEM 4. Navitas has bought into a layer whose price someone else is setting.
Two earnings calls will show whether the layer pays
Vicor reports third-quarter results on October 20, 2026, its first since the case began 7. Navitas follows on November 3 with a harder job: turning summit showcases into order backlog 69. Short interest sits near a fifth of its tradable float 6. The stock still commands more than 70 times sales while quarterly revenue keeps shrinking as the consumer business it abandoned winds down 68.
Revenue has halved while Navitas rebuilt itself for AI racks
Data
| Revenue | |
|---|---|
| Q3 '24 | $21.68M |
| Q4 '24 | $17.98M |
| Q1 '25 | $14.02M |
| Q2 '25 | $14.49M |
| Q3 '25 | $10.11M |
| Q4 '25 | $7.3M |
| Q1 '26 | $8.6M |
| Q2 '26 | $10.53M |
Cash above half a billion dollars covers the wait 8. The $8 billion market Claros is meant to unlock by 2030 is a company estimate, with no customers or production dates attached 1. The patent fight sets part of the price of that rebuild: builders who use the respondents' parts face Vicor's escalating royalties, and that bill lands on the layer Navitas just bought.
Cash rebuilt to half a billion dollars while revenue halved
Data
| Cash | |
|---|---|
| Q3 '24 | $98.61M |
| Q4 '24 | $88.24M |
| Q1 '25 | $75.61M |
| Q2 '25 | $161.34M |
| Q3 '25 | $151.22M |
| Q4 '25 | $238.6M |
| Q1 '26 | $223.37M |
| Q2 '26 | $558.27M |
Deepdive
AI-generated from this story and its cited sources. Not investment advice.



