Management admitted their primary AI growth engine is decelerating because customers are getting too good at testing.
Thesis: The pivot is real—Compute is now 34% of SOC revenue—but the easy money has been made. Management explicitly identified an 'efficiency anchor': as AI devices mature, test intensity drops. They are guiding for 15% growth in a Compute market growing only 5%, relying entirely on VIP socket wins to bridge the gap. With HBM in digestion and Robotics in perpetual restructuring, the structural tailwinds are weaker than the narrative suggests.
Verdict: HOLD — Conviction: MEDIUM
Catalyst: HBM4 transition timing in late 2025. If it pulls forward, it breaks the digestion cycle; if it stays in 2026, the year is flat.
Key Risk: The 5% Compute TAM growth assumption is fragile. If hyperscaler capex pauses even slightly, that low growth turns negative, and the 'share gain' thesis collapses.
The Tell: When asked about the $40B vs $140B ASIC TAM disconnect, CEO Greg Smith bluntly stated they align with the 'low end' because chipmaker margins don't equal test units. He explicitly rejected the hype numbers.
Friction Level: MODERATE_FRICTION — Management guides to 15% revenue growth while admitting Compute TAM grows only 5% due to efficiency gains. The gap is filled by 'share gains' which are finite.
Report not found
The report data is no longer available. Please return to the archive.