Nvidia is worth $5.5 trillion. Annualise the current guidance run rate and you get north of $360 billion in revenue. That implies a forward price-to-sales ratio in the mid-teens for a hardware company, however dominant. The stock has fallen after each of its last four earnings reports despite beating estimates every time.
The pattern tells you something important. The market has already priced in the growth trajectory. A beat-and-raise quarter is table stakes. What investors are looking for now is evidence that Nvidia can widen the business fast enough to justify a valuation that assumes near-perfection in the core GPU franchise.
This quarter offered the clearest signal yet of how Nvidia intends to answer that question.
A business being reshaped in plain sight
Nvidia restructured its earnings breakdown into two segments: data centre and edge computing. Inside data centre, it split revenue between hyperscalers at $38 billion and a new category called ACIE, covering AI clouds, industrial and enterprise customers, at $37 billion. AI cloud revenue within that segment tripled year on year.
The restructuring is deliberate. Gabelli Funds portfolio manager John Belton flagged before the print that five customers account for roughly half of Nvidia's revenue. The new reporting is Nvidia's response: look at the breadth, not the dependency.
The bigger shift is CPUs. CFO Colette Kress said Nvidia expects $20 billion in CPU revenue this year and aims to become the world's leading CPU supplier, a market currently held by Intel and AMD.
The Vera CPU is positioned as the chip that runs agentic AI workloads, the orchestration, sandboxing and retrieval tasks that sit alongside GPU-powered inference. Jensen Huang said he expects Nvidia to be supply-constrained on Vera Rubin, the next-generation system, for its entire production life.
The $200 billion addressable market claim for CPUs deserves scrutiny. But the strategic intent is clear. Nvidia is building a second revenue pillar that does not depend on the training cycle.
Nvidia names the competitive threat
For the first time, Nvidia used its 10-Q filing to state plainly that its own customers are becoming semiconductor competitors. The company noted that hyperscalers are developing custom ASICs optimised for workloads that may not require the full feature set of Nvidia's systems.
Google is spinning out its TPU business into a standalone company backed by $5 billion from Blackstone. Amazon's Trainium chips are running Anthropic's Claude on more than a million units. Cerebras debuted on Nasdaq last week and hit a $100 billion market cap on day one. Nvidia's own Groq acquisition was described by Huang on the call as a niche product for some time.
The filing also warned that customers could constrain Nvidia's access to foundry capacity and scarce materials. When your biggest buyers are also competing for TSMC wafer starts, the commercial relationship acquires a new edge.
China is zeroed out but not gone
Nvidia excluded all China data centre revenue from its guidance. Huang attended President Trump's Beijing summit last week but the visit produced no clarity on H200 sales. Reuters reported that Alibaba, Tencent and ByteDance have received individual approvals to purchase H200s, but a US trade representative said export controls were not discussed.
China once accounted for at least a fifth of data centre revenue. That market is now absent from the forward estimates, which means any reopening represents pure upside but also that the current growth rate is being achieved without Nvidia's second-largest historical customer base.
The sprint ahead
Nvidia's response to the valuation question is to widen the aperture: new segments, new reporting, a $200 billion CPU opportunity, an $80 billion buyback, a dividend increase from one cent to 25 cents per share. Huang closed the call by declaring that agentic AI has arrived and that Nvidia sits at the centre of the transition.
On the numbers, that is hard to dispute. Revenue of $82 billion, free cash flow of $48.6 billion and 75% gross margins represent a level of execution that has no precedent in semiconductor history.
The question is whether the CPU push, the edge computing expansion and the enterprise AI build can collectively grow fast enough to support a company that needs to add the equivalent of an entire AMD to its revenue base every year to hold its current multiple. The earnings say yes. The stock price says show me.
