Navitas Semiconductor is trying to finish a hard pivot from phone chargers to the power train inside AI factories and the grid that feeds them. Chris Allexandre, who took the chief executive seat last September after running the power franchise at Renesas, has recast the company as Navitas 2.0 and is walking away from mobile and low-end consumer volume that once defined the print. The investment debate is whether that mix shift converts named design work around eight-hundred-volt data-center architectures into a franchise large enough to justify a multi-billion equity, or whether the market is still paying for an option on NVIDIA's power roadmap while the operating company remains sub-scale and cash-consuming.
The June quarter is the first clean sequential test of that story. Revenue rose twenty-two percent from the March period to just over ten million, and management says high-power end markets grew more than fifty percent from a year earlier. The year-ago comparison still declined because the company is deliberately shrinking the old mobile book. Cash ended the quarter at $557 million after an opportunistic at-the-market raise that brought in roughly $373 million at an average print well above today's share price. That war chest funds the foundry transition off TSMC and the inventory buffer for AI wafers, but it also reset the share count and reminded holders that the equity is the standing financing vehicle.
The next several prints decide whether this is a mix-shift that compounds or a narrative that outran the P&L. Management guided the September quarter to about $14 million at the midpoint and talked about a return to year-over-year growth, with AI-related infrastructure approaching one third of sales by year-end. The counterargument is already on the tape. TSMC is leaving gallium nitride production in the middle of next year, Wolfspeed sued across the gallium nitride and silicon carbide catalog in early July, and GeneSiC founder Ranbir Singh left the board in June after activist-style thirteen-D filings. The open question is whether the high-power book keeps compounding fast enough that dilution and foundry risk recede, or whether the multiple compresses back toward a small, still-unprofitable analog name.