Nvidia Q2 FY27 Earnings August 26: $91B Guide and What Developers Watch

Nvidia will report second-quarter fiscal 2027 results on Wednesday, August 26, 2026, after the U.S. market close. Colette Kress confirmed the date on the May 20, 2026 Q1 call: "Our earnings call to discuss the results of our second quarter of fiscal 2027 is scheduled for August 26th." The event is listed on Nvidia investor relations.

For developers, this is not just a stock print. Nvidia earnings are the cleanest quarterly read on whether AI factory buildout — GPUs, networking, inference software — is accelerating or stalling. When data-center revenue misses or guidance softens, cloud GPU availability, reservation pricing, and enterprise inference budgets move within weeks.

Q1 FY27 Baseline: What Nvidia Already Reported

Nvidia's Q1 FY27 press release (quarter ended April 26, 2026, reported May 20, 2026) established the bar:

CEO Jensen Huang framed the quarter around "AI factories" — the largest infrastructure expansion in history — and agentic AI moving from demos to production inference loops.

Q2 FY27 Guidance: The Number Wall Street Anchors On

Nvidia guided Q2 FY27 revenue to $91.0 billion, plus or minus 2% ($89.2B–$92.8B range). That implies roughly +12% sequential growth from Q1's $81.6B.

Other Q2 guideposts from the May release:

• GAAP/non-GAAP gross margin: 74.9% / 75.0%, ±50 bps
• GAAP/non-GAAP opex: ~$8.5B / ~$8.3B
• China: Nvidia assumed zero Data Center compute revenue from China in the outlook

If Q2 reports near $91B with margins holding near 75%, the AI capex cycle narrative stays intact. A miss on data-center revenue or a margin guide-down would ripple through cloud GPU pricing and LLM API economics faster than most product announcements.

What Developers Should Watch on August 26

1. Data Center revenue and Blackwell mix

Q1 data-center networking grew 199% YoY — Spectrum-X, InfiniBand, and AI factory fabric are no longer a footnote. Watch whether management breaks out Blackwell shipment ramp, supply constraints, and inference vs training mix. Dynamo 1.0 — Nvidia's open-source inference stack claiming up to 7x generative and agentic inference gains on Blackwell — entered production in Q1. Developer adoption signals matter for whether inference efficiency reduces raw GPU unit demand or expands total workloads.

2. China and export-control commentary

Nvidia guided no China data-center compute revenue in Q2. Any change in H200/Blackwell export policy, Middle East allocation, or "assumed zero" language affects global GPU supply pools. Teams running multi-region inference should treat export-control remarks as capacity planning inputs — see our RASA export-control checklist.

3. Vera Rubin platform timing

Q1 announced the Vera Rubin platform: Vera CPU (positioned as purpose-built for agentic AI), Rubin GPU, and BlueField-4 STX for context-memory storage. August commentary on Rubin sampling, customer qualifications, and HBM supply informs 2027 cluster refresh cycles — detailed in our Hot Chips 2026 guide.

4. Networking as inference bottleneck

At $14.8B quarterly networking revenue, AI factory fabric is a first-class product line. Agentic workloads multiply east-west traffic (tool calls, retrieval, KV cache movement). If networking grows faster than compute again, platform teams should budget for Spectrum-X and DPU offload — not just more GPUs.

5. Hyperscaler capex correlation

Microsoft, Google, Amazon, and Meta collectively guide hundreds of billions in 2026 AI capex. Nvidia's Q2 data-center print is the receipt. Developer-facing impact: reserved instance pricing, new instance families (Google A5X Vera Rubin previews were announced in Q1), and regional capacity for fine-tuning jobs.

Our Analysis: Why This Earnings Call Matters for Infra Teams

Inference is the operating cost layer. Training capex gets headlines; inference bills hit every API call. Blackwell + Dynamo improvements directly affect per-token cost — the same variable that drove our $500M Claude bill coverage and GPT-5.6 FinOps guide.

Margin stability signals supply normalization. Gross margins near 75% suggest Nvidia is not discounting aggressively to clear inventory — demand still outstrips supply for top-tier accelerators. If margins compress while revenue grows, watch for competitive pressure from custom silicon and Google TPU v8.

The August 26 date clusters with Hot Chips (Aug 23–25). Hardware architecture details land days before financial metrics — read both together for a full picture of agentic AI infrastructure.

Developer Prep Checklist Before August 26

• [ ] Baseline your cloud GPU spend and reservation expiry dates
• [ ] Map workloads to Blackwell vs Hopper instance availability by region
• [ ] Track whether your provider passes through inference efficiency gains (Dynamo, TensorRT-LLM) or only raw GPU hours
• [ ] Set alerts for NVDA after-hours release (~4:20 p.m. ET typical) and 5:00 p.m. ET call
• [ ] Update FinOps models if data-center revenue growth decelerates below guided ~12% sequential

Key Takeaways

• Nvidia Q2 FY27 earnings: August 26, 2026, after U.S. market close — confirmed on May 20 Q1 call and IR calendar.
• Q1 FY27: $81.6B revenue, $75.2B data center (+92% YoY), 74.9% GAAP gross margin.
• Q2 guide: $91.0B revenue ±2%, margins ~75%, zero China data-center compute assumed.
• Developer watch: Blackwell/Dynamo inference ramp, networking growth, Rubin timeline, export-control language.
• For developers: Treat the print as a capacity and pricing signal for cloud GPU planning — not just equity noise.
• What to watch: Q2 actuals vs $91B guide, Q3 outlook on the call, any China revenue assumption changes.

Related Reading

• Hot Chips 2026 Rubin and BlueField-4 guide
• GPT-5.6 programmatic tool calling FinOps
• LLM API pricing tracker
• Nvidia Rubin HBM4 supply chain

Sources

• NVIDIA Q1 FY27 Earnings Press Release, May 20, 2026
• SEC filing Q1 FY27
• NVIDIA Q2 FY27 Earnings Event, August 26, 2026