Nvidia earnings show the AI chip boom is still accelerating

Nvidia has now supplied the clearest fresh answer to the question hanging over this earnings week: is the AI spending boom still turning into real revenue, or is Wall Street merely trading on anticipation? Revenue for the first quarter of fiscal 2027 hit $US81.6 billion (about $125 billion). Net income jumped 211 per cent to $US58.3 billion (about $89 billion). Between the two figures, the debate shifts back towards acceleration rather than exhaustion.

Nvidia’s second-quarter guidance mattered almost as much as the headline numbers. The company forecast about $US91.0 billion (about $139 billion) in revenue for the current quarter. Hyperscalers and model labs buying its chips are spending at a pace that would look extraordinary for almost any other segment of the tech industry, and that is what makes this quarter read less like a routine beat-and-raise and more like a live check on whether the infrastructure phase of AI is still broadening.

But the sceptical case did not vanish when the numbers landed. It simply moved. After days of preview coverage that asked whether Nvidia had become too big to excite, the pressure point is no longer whether demand exists. The question now is whether customers keep funnelling so much of that demand through one supplier as in-house silicon, export controls and market concentration start to bite.

The bullish reading still has the stronger evidence

Straight up, the analyst view this quarter is not complicated: the AI capex cycle is still climbing and Nvidia remains its cleanest revenue proxy. Data centre revenue keeps cashing out the story because the company is selling into a market where the largest buyers are still building capacity ahead of clear near-term returns. That dynamic matters more than any one-day share-price move.

Context around the results makes that reading harder to dismiss as simple optimism. On Monday, CNBC reported that Blackstone will invest $US5 billion (about $7.7 billion) in an AI infrastructure venture with Google built around TPU chips. Even Nvidia’s competitors are responding to the same signal: demand for compute is still rising fast enough to justify cheques of that size. The Register reported the same week that OpenAI was floating upfront payments for guaranteed AI availability — another sign that scarcity has not left the system.

In Nvidia’s own earnings release, chief executive Jensen Huang argued that the next phase is about whole AI factories, not just another upgrade cycle for chips.

“The buildout of AI factories — the largest infrastructure expansion in human history — is accelerating at extraordinary speed.”
— Jensen Huang, Nvidia

The language is promotional. But the underlying point is harder to wave away after a quarter like this one. Revenue, profit and guidance all kept breaking higher. At this point the burden shifts to the bears to show not just that AI spending is expensive, but that it is actually slowing down.

The next argument is about bargaining power, not collapse

The more credible sceptical reading is not that Nvidia’s boom ends abruptly. It is that customers spend the next year trying to negotiate around it. Substitution risk is the most important challenge. Google has TPUs. Amazon has Trainium and Inferentia. Microsoft has Maia. Smaller rivals such as Cerebras are using inference speed and power claims to argue that the GPU stack will not own every workload forever.

None of this means Nvidia loses the centre of gravity soon. The market is simply starting to test where its pricing power ends. The Financial Times captured that tension in a broader piece on whether Nvidia had become systemically important to the AI trade, quoting Morgan Stanley analyst Joe Moore on the company’s centrality to nearly every major AI build-out.

“You’re clearly at the centre of everything.”
— Joe Moore, Morgan Stanley, via the Financial Times

That sentence is both the bullish case and the risk case. Being at the centre of everything means Nvidia keeps collecting the fattest part of the current spending cycle. Every large customer also has a reason to reduce dependency over time, especially if AI margins come under pressure and cheaper inference starts to matter more than bragging rights on training clusters.

Policy is still shaping the ceiling on demand

The policy watcher’s question is slightly different. It is not whether Nvidia still has buyers but how large its addressable market would be without geopolitics cutting away parts of it. CNBC reported in February that Nvidia had still not sold its US-approved China AI chips and was worried local rivals could fill the gap. Ars Technica reported this week that Beijing had banned an Nvidia gaming chip variant during Huang’s China visit. Those are different markets, but they point to the same reality: export controls and industrial policy are no longer side stories to the AI hardware business. They are part of the revenue model.

Nvidia’s quarter looks like a reality check rather than just another stock-market event for that reason. The company’s latest numbers say the AI build-out is still producing extraordinary commercial returns right now. The harder questions sit one layer below — around concentration, substitution and China. For Australian readers, especially enterprise buyers watching cloud pricing and AI-service availability, the practical read-through is simple: the global AI stack still runs on constrained, expensive infrastructure, and Nvidia just showed that constraint is not easing yet.