NVIDIA splits hyperscaler sales reporting

Why it matters

Nvidia’s reporting change matters because it separates two parts of the AI market that had been bundled together in the company’s data-center story. In its first-quarter fiscal 2027 results on May 20, Nvidia reported record revenue of $81.6 billion and data-center revenue of $75.2 billion, then told investors it would distinguish hyperscaler cloud sales from the rest of the market in future disclosures. Jensen Huang paired that accounting change with a broader argument about AI economics. On the earnings call and in follow-up comments, Huang said AI had crossed a “critical threshold” and that “tokens are now profitable,” framing AI output itself as a revenue-generating unit rather than just a cost center. Benzinga reported that Nvidia tied that view to the idea that AI-factory spending could eventually reach trillions of dollars annually. (nvidianews.nvidia.com) ### Why did Nvidia split out hyperscaler sales now? May 20 is the key date because that is when Nvidia delivered the new framing alongside another quarter of surging growth. CNBC reported Huang told analysts the reporting change was meant to explain the business more clearly, as the company sells into giant cloud providers on one side and a wider mix of enterprises, sovereign buyers and other customers on the other. (benzinga.com) Stratechery said the distinction highlights different competitive dynamics inside Nvidia’s customer base. In that reading, hyperscalers are the buyers most capable of pushing suppliers toward commoditized hardware economics over time, while non-hyperscaler customers often buy more of Nvidia’s broader stack of systems, networking, software and support. (nvidianews.nvidia.com) ### What does Huang mean by “tokens are now profitable”? Huang’s phrase points to inference economics. A token is the unit many AI systems use to meter text, code and other model output, and Nvidia has increasingly described AI infrastructure in terms of token throughput, token cost and revenue per megawatt. Nvidia’s own developer and blog materials have argued that lower cost per token is the central operating metric for AI factories. (stratechery.com) Nvidia has been building that argument for months. In March and April, company materials described AI factories as infrastructure that turns data into “intelligence” and emphasized token-metered services as a business model for cloud and telecom operators. Huang’s “critical threshold” comment suggested he believes that model has moved from theory into commercial reality. (benzinga.com) ### Why are analysts focusing on cloud relationships and data plumbing? The cloud providers sit between Nvidia and many end customers. If hyperscalers control the customer relationship, the rented compute environment and much of the networking and data movement around AI workloads, they may capture more of the long-term value even if Nvidia remains the core chip supplier. That is the line of argument Stratechery advanced after the earnings report. (blogs.nvidia.com) The Motley Fool made a related point from the market side, arguing that companies tied to moving data around AI data centers could be among the next beneficiaries of AI spending. CNBC also cited Needham analysis showing consensus estimates for hyperscaler capital expenditure reaching about $1.03 trillion in 2028, well below Huang’s own longer-range spending vision. (stratechery.com) ### Does this change Nvidia’s near-term story? Nvidia’s near-term numbers still show a company dominated by AI infrastructure demand. The company said first-quarter revenue rose 85% from a year earlier to $81.6 billion, while data-center revenue rose 92% to $75.2 billion. Huang said in the earnings release that the buildout of AI factories was accelerating at “extraordinary speed.” (cnbc.com) The new disclosure does not change those results. It gives investors a cleaner way to track whether growth is being driven mainly by a handful of cloud giants or by a broader base of enterprises, sovereign projects and other buyers that Nvidia says are adopting its full-stack platform. ### What comes next for investors to watch? May 28 is Nvidia’s next scheduled investor event, when the company is due to appear at the BofA Securities Global Technology Conference, according to its investor relations calendar. (nvidianews.nvidia.com) Future quarterly reports will show whether Nvidia formalizes the hyperscaler breakout in a way that lets investors compare cloud concentration against the rest of the AI market over time. (investor.nvidia.com) (cnbc.com)

Key numbers

• Nvidia said on May 20 it would break out hyperscaler cloud revenue from other sales as part of its first-quarter fiscal 2027 reporting.
• Jensen Huang said AI had crossed a “critical threshold” and that “tokens are now profitable” as Nvidia posted $81.6 billion in quarterly revenue.
• Nvidia’s next scheduled investor event is its May 28 appearance at the BofA Securities Global Technology Conference.
• In its first-quarter fiscal 2027 results on May 20, Nvidia reported record revenue of $81.6 billion and data-center revenue of $75.2 billion, then told investors it would distinguish hyperscaler cloud sales from the rest of the market in future disclosures.

What happens next

• In its first-quarter fiscal 2027 results on May 20, Nvidia reported record revenue of $81.6 billion and data-center revenue of $75.2 billion, then told investors it would distinguish hyperscaler cloud sales from the rest of the market in future disclosures.
• Benzinga reported that Nvidia tied that view to the idea that AI-factory spending could eventually reach trillions of dollars annually.
• May 20 is the key date because that is when Nvidia delivered the new framing alongside another quarter of surging growth.

Sources

• nvidianews.nvidia.com
• benzinga.com
• stratechery.com
• blogs.nvidia.com
• cnbc.com
• investor.nvidia.com
• cnbc.com

Quick answers