H100 revenue exceeded A100 in its second quarter of shipping, yet total data center revenue fell 6% sequentially.
Thesis: The street is misinterpreting a supply-chain transition as a demand signal. The -6% sequential drop in Data Center is an air pocket caused by the shift from Ampere to Hopper, not a lack of appetite. H100 overtaking A100 in two quarters is the only metric that matters—it confirms the 'Structural Demand Shift'. Pricing power is intact (GM expanding to 66.5%), and the 'Capacity Ceiling' ensures those margins stick. This isn't a cyclical recovery; it's a platform monopoly re-pricing the datacenter.
Verdict: LONG — Conviction: HIGH
Catalyst: GTC on March 21st. Management promised 'new chips, systems, and software', which will likely confirm the H100 backlog velocity.
Key Risk: China revenue decline (20-25% of DC) is permanent and masked by the H100 ramp. If the non-China ramp decelerates, the hole becomes visible.
The Tell: Management attributed the Data Center decline to a hyperscaler 'pause to recalibrate' while simultaneously admitting China revenue fell 'largely in line with expectations.' They lumped a permanent geopolitical loss (China) with a temporary tactical delay (CSP retrofit) to soften the headline miss.
Friction Level: HIGH_FRICTION — The 'Pause'. Bulls see a temporary recalibration for Hopper; Bears see structural demand destruction and a China hole that isn't coming back.
Report not found
The report data is no longer available. Please return to the archive.