A major theme on the call was the company’s view that AI economics increasingly depends on throughput and efficiency. Jensen Huang, CEO of Nvidia, described the relationship directly, saying in “this new world of AI, compute equals revenues. Without compute, there’s no way to generate tokens, and without tokens there’s no way to grow revenues.” He added that real-time inference is becoming the primary driver of cloud monetization, raising the importance of performance per watt as operators scale production workloads.
While addressing uncertainty around future hyperscaler spending. Jensen said stronger monetization from agentic workloads supports continued investment, noting “We’ve reached the inflection point. The usefulness of agents is generating incredible compute demand, and I am confident in their cash flow growing.” While hyperscalers represent a little over half of data center revenue, the company said it is seeing even faster growth from enterprises, AI model developers and sovereign customers.
The company emphasized the advantages of its ecosystem, particularly CUDA, which remains compatible across generations of GPUs. Jensen said, “We’re the only accelerated computing platform in every cloud, and CUDA makes the entire installed base better with every generation.” Nvidia pointed to improvements in inference performance and cost efficiency across Blackwell, Hopper and Ampere as evidence of how software optimization extends the life of older hardware.
Sovereign AI demand also accelerated, with revenue surpassing $30 billion for the year as several countries began building national AI infrastructure.
The call also explored longer-term themes, including early GPU deployments in space. Jensen noted that imaging and on-board processing are promising use cases. He also returned to his broader view that AI represents a break from prewritten software, saying “The amount of computation necessary for the modern way of doing software is a thousand times higher.”
Looking ahead, Nvidia said the Rubin platform remains on schedule for second-half availability. Rubin is designed to train models with roughly one-fourth the number of GPUs and to reduce inference costs materially. CFO Colette Kress said the company has secured long-dated supply commitments, stating “We have strategically secured inventory and capacity to meet demand beyond the next several quarters.”
What Else Stood Out on the Call
- Jensen said Nvidia is close to finalizing a new partnership agreement with OpenAI, noting that GPT-5.3 Codex is already trained and deployed on Grace Blackwell and heavily used inside Nvidia.
- Growth outside hyperscalers accelerated, with enterprises, sovereigns, AI model builders, telecom and robotics customers expanding faster than cloud providers and becoming a larger share of data center demand.
- Nvidia described the integration of the Grok inference team as an architectural extension similar to Mellanox, with new low-latency capabilities expected to become part of Nvidia’s full-stack platform.
- The company reiterated its capital allocation priorities, saying long-term supply commitments, ecosystem investments and strategic stakes in frontier model developers remain higher priority than accelerating buybacks despite strong cash generation.
Topline Numbers
Nvidia closed the quarter with revenue of $68 billion, up 73% from a year earlier, driven by persistent strength in its data center business. Data center revenue reached $62 billion, rising 75% year over year and 22% sequentially, bringing full-year data center revenue to $194 billion, an increase of 68%.
Networking revenue reached $11 billion, reflecting continued adoption of NVLink and Spectrum X, while gaming revenue came in at $3.7 billion, professional visualization reached $1.3 billion and automotive revenue totaled $604 million.
Gross margin for the quarter was 75% on a GAAP basis. Free cash flow was $35 billion for the quarter and $97 billion for the fiscal year, and the company returned $41 billion to shareholders through buybacks and dividends.
Nvidia expects first-quarter fiscal 2027 revenue to reach approximately $78 billion.