Nvidia Posts $96.2 Billion Quarter as Data Center Revenue Doubles Year on Year (Aug 2026)

Nvidia reported results for its second quarter of fiscal 2027, ended 26 July 2026, on 26 August. Revenue reached $96.2 billion, up 18% from the prior quarter and 106% from a year earlier. Data center revenue was $89.0 billion, up 117% year on year, while the edge computing segment contributed $7.2 billion. Gross margin held at 75.0%, GAAP net income was $59.7 billion, and the company returned $26.0 billion to shareholders through buybacks and dividends during the quarter.

Guidance for the third quarter is $108.0 billion in revenue, plus or minus 2%, with gross margin around 74%. Notably, the outlook assumes no data center compute revenue from China. Operationally, Nvidia said its Vera Rubin platform is in full production across the major cloud providers, announced a Vera CPU designed for AI agent workloads, confirmed the Groq 3 LPX inference accelerator is in full production, introduced a DSX platform for AI factory infrastructure, and pointed to more than $500 billion in compute financing partnerships. CEO Jensen Huang summarised the thesis in one line: "Now, compute is revenue."

CNBC reported that Huang forecast roughly 70% revenue growth for fiscal 2028, well above analyst estimates.

Why it matters

The numbers confirm that AI infrastructure spending has not slowed despite July's security incidents and the growing policy debate about pacing frontier development. If anything, the emphasis has shifted from training to inference: a dedicated inference accelerator, a CPU built for agents, and financing structures designed to let customers reserve capacity years ahead all point to workloads that run continuously in production rather than in periodic training bursts.

For enterprise buyers, the demand picture means compute scarcity, long lead times and price power remain with suppliers. The explicit exclusion of China from guidance is a reminder that export controls are now embedded in the revenue model of the industry's largest supplier, which affects any multinational running inference in restricted jurisdictions. The $500 billion in financing partnerships also signals that hyperscalers and neoclouds are taking on long-dated obligations; concentration and counterparty risk in the AI supply chain are rising alongside capacity.

What it means for leaders

• Plan capacity as a multi-year commitment. Treat GPU and inference reservations like other long-lead procurement, with utilisation forecasts, exit terms and secondary sourcing.
• Assess concentration risk in your inference stack. A small number of chip, cloud and model providers now sit under most enterprise AI; NIST AI RMF's Govern function expects third-party dependency to be mapped and monitored.
• Budget for inference, not just pilots. Agent workloads consume tokens continuously; instrument cost per task and set guardrails before scaling.
• Check export-control exposure. Where you operate or serve customers in restricted markets, confirm which hardware and hosted models you can lawfully use.
• Tie infrastructure spend to governance maturity. Boards approving large AI capex should ask for the corresponding investment in evaluation, monitoring and incident response, which ISO/IEC 42001 frames as part of the management system, not an afterthought.