ACIE grew 31% sequentially to $37B while hyperscale grew 12% to $38B. The second engine is now equal in size and growing 3x faster.
Thesis: Nvidia's second demand engine (ACIE) at $37B growing 31% sequentially is now the same size as hyperscale and growing at nearly 3x the rate. The Street is still modeling concentration risk into 5-6 hyperscalers while the addressable buyer base has expanded to hundreds of sovereign, AI cloud, and enterprise accounts across 40 countries. The $145B supply commitment and $80B buyback are real capital deployed against this diversification. The capacity ceiling is real but management is pricing through it with 75% gross margins held flat.
Verdict: LONG — Conviction: MEDIUM
Catalyst: VeraRubin production shipments starting Q3 with $20B standalone CPU revenue visibility this year, expanding into a $200B TAM Nvidia has never sold into.
Key Risk: Hyperscaler CapEx guide-down. $38B quarterly revenue flows through 5-6 balance sheets. Analyst Reitzes already flagged hyperscaler CapEx growing 90-100% this year, which is unsustainable at scale.
The Tell: Colette corrected Jensen's dividend number unprompted: Jensen said $0.25 while Colette had said $0.20 in prepared remarks, and Jensen caught it in real time saying 'Colette meant to say $0.25.' The CFO and CEO were not aligned on a headline capital return number before the call, which is unusual for a $119B cash return commitment.
Detected Patterns
Beat Above Buy-Side Whispers: $82B revenue vs buy-side whisper expectations. $13.5B sequential increase was the largest in company history with third consecutive quarter of YoY acceleration. The beat was structural, not timing.
Pricing Power Signal: H100 cloud rental rates up 20% YTD and A100 up 15% while Blackwell ramps. Prices rising on older generation hardware during a new product cycle means demand has no substitute.
Capital Conviction: $145B supply commitments plus purchase commitments and prepaids. $80B new buyback authorization plus $39B remaining. Dividend raised from $0.01 to $0.25. Backed by $49B quarterly FCF.
Structural Demand Shift: ACIE segment at $37B growing 31% QoQ with AI cloud revenue more than tripling YoY, sovereign revenue up 80% YoY across 40 countries, and 80 partner data centers over 10MW. This is not cyclical hyperscaler CapEx.
Capacity Ceiling: Jensen expects to be 'supply constrained throughout the entire life of VeraRubin.' $145B supply commitment against $91B quarterly guide. Revenue capped by CoWoS capacity and HBM stacking, not demand.
Export Control Revenue Loss: H200 licenses approved for China but revenue is zero. Management excludes China data center compute from outlook for second consecutive quarter. Entire growth market severed.
Narrative Shift to Inference: Heavy emphasis on 'tokens are now profitable,' agentic AI harnesses, and billions of future agents. Shift to inference narrative coincides with training buildout maturing and hyperscale at 50% of data center.
Backlog Quality: Jensen cites 'we have got POs' for VeraRubin but Colette says 'little early to say' on ramp speed. POs are stated but not characterized as non-cancellable firm commitments.
Friction Level: MODERATE_FRICTION — Both sides agree the numbers are clean. Bull sees ACIE at $37B growing 31% QoQ as an unpriced diversification catalyst. Bear sees sequential deceleration from 20% to 11% as a ceiling effect from CoWoS and HBM constraints. Same facts, opposite read on rate of change.
Report not found
The report data is no longer available. Please return to the archive.
Two Demand Engines, One Supply Ceiling | Silicon Signal