NVIDIA posts $96.2bn Q2 revenue as data centre sales double
NVIDIA reported second-quarter fiscal 2027 revenue of $96.2 billion on Wednesday, up 106% from a year earlier and 18% sequentially, as demand for AI infrastructure continued to outpace analyst expectations. Data centre revenue, which now accounts for roughly 93% of total group sales, reached $89.0 billion, up 117% year-on-year. GAAP gross margin held at 75.0%, up 2.6 percentage points from the same period last year. GAAP diluted earnings per share came in at $2.46, compared with $1.08 in Q2 fiscal 2026.
For the third quarter, the company guided revenue of $108.0 billion, plus or minus 2%, with gross margins expected at 74.0%. Notably, NVIDIA stated it is not assuming any data centre compute revenue from China in that outlook, a significant caveat given the US export-control environment.
The data centre and product picture
The headline product driving data centre growth is the Vera Rubin platform, which NVIDIA said is now in full production and ramping with hyperscale partners including CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius. The company also disclosed that its Blackwell architecture led across every category in the MLPerf Training 6.0 benchmarks, as well as in AgentPerf, described as the industry's first agentic AI infrastructure benchmark.
Beyond compute silicon, NVIDIA announced Spectrum-6 switch systems supporting pluggable and co-packaged optics, a new BlueField-4 STX security processor targeting agentic AI workloads, and the DSX platform, positioned as an end-to-end playbook for building and operating AI factories at scale. The company also revealed that its GPUs with Confidential Computing are now used inside Apple's Private Cloud Compute infrastructure, a commercially significant deployment that had not previously been disclosed.
On the financing side, NVIDIA announced strategic partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish independent compute financing platforms, with the stated aim of mobilising over $500 billion of third-party capital for AI infrastructure buildout over time, subject to definitive agreements.
Edge computing revenue for the quarter was $7.2 billion, up 27% year-on-year, supported by new products including the RTX Spark PC chip developed with Microsoft and the DGX Station for Windows.
Market context and regulatory backdrop
NVIDIA's sequential growth trajectory, from $46.7bn a year ago to $81.6bn last quarter and now $96.2bn, underscores how dominant the company has become in the AI accelerator market. AMD and Intel continue to pursue the data centre GPU opportunity, and custom silicon programmes at Google (TPUs), Amazon (Trainium) and Microsoft (Maia) represent the most credible long-term threat to NVIDIA's margin profile. However, the MLPerf results and the breadth of named hyperscaler partnerships in the Vera Rubin ramp suggest NVIDIA retains a meaningful performance and ecosystem lead heading into the second half of calendar 2026.
The China exclusion in Q3 guidance is the most significant regulatory overhang. US Bureau of Industry and Security controls have progressively restricted NVIDIA's ability to sell high-performance compute into China, removing what was historically a large and high-margin market. The company's forward guidance implicitly prices in that constraint remaining in place, though any easing or tightening of export rules could materially shift the revenue picture.
Founder and chief executive Jensen Huang said the AI infrastructure build-out is "at full steam", and described compute as having become a direct revenue generator for customers rather than a cost centre. NVIDIA returned approximately $26.0 billion to shareholders during the quarter through buybacks and dividends, and disclosed approximately $99.0 billion remaining under its repurchase authorisation, signalling continued confidence in cash generation capacity.