NVIDIA on November 19 posted record third-quarter fiscal 2026 revenue of $57 billion, up 22% sequentially and 62% year over year, driven by continued demand for accelerated computing and AI infrastructure. Data centre revenue rose to $51.2 billion, up 66% year over year.
CEO Jensen Huang said demand for its Blackwell architecture continues to exceed supply. “Blackwell sales are off the charts and cloud GPUs are sold out,” he said in the earnings release, adding that compute demand is accelerating and compounding across training and inference—each growing exponentially. He described the industry as having “entered the virtuous cycle of AI.”
On the earnings call, CFO Colette Kress said, “Record Q3 data centre revenue of $51 billion increased 66% year over year, a significant feat at our scale.” She attributed the growth mainly to the GB300 GPU ramp, strong networking demand and expanding AI deployment across hyperscalers and model builders.
“The clouds are sold out, and our GPU installed base, both new and previous generations, including Blackwell, Hopper and Ampere, is fully utilised,” Kress added.
Addressing concerns about an AI bubble, Huang pushed back on the narrative. “There’s been a lot of talk about an AI bubble. From our vantage point, we see something very different,” he said.
Blackwell’s GB300 accounted for “roughly two-thirds of total Blackwell revenue”, overtaking GB200 as customers transitioned to the new platform.
Huang highlighted performance gains, saying Blackwell Ultra delivers “5x faster time to train vs Hopper”, while on DeepSeek R1 benchmarks, it offers “10x higher performance per watt and 10x lower cost per token versus H200.”
The next-generation Rubin platform remains scheduled for a 2026 ramp, with first silicon already received. “Our ecosystem will be ready for a fast Rubin ramp,” Kress said.
More Customers
NVIDIA said it is now involved in AI factory projects totalling five million GPUs, spanning cloud providers, sovereign initiatives, enterprises and supercomputing centres.
The company highlighted several large-scale deployments, including xAI’s Colossus 2, a gigawatt-scale data centre, an expanded collaboration with AWS and Humane, which plan to deploy up to 1,50,000 NVIDIA AI accelerators.
A new strategic agreement with Anthropic, which is adopting NVIDIA’s architecture for the first time and has committed up to one gigawatt of compute for its future systems.
“We run every AI model—OpenAI, Anthropic, xAI, Gemini, science models, biology models, robotics models,” Huang said.
Your reaction
Discussion
What's your take on this story?
No comments yet — be the first to share your view.