Nvidia (NVDA) has given investors the number they wanted and a new problem to worry about. The company expects revenue in the fiscal year ending January 2028 to grow roughly 70%, comfortably above the growth rate Wall Street had been modelling before the results. That outlook was enough to send Nvidia shares sharply higher and pull a broad group of semiconductor and AI infrastructure names with it.
The reaction matters because the market had spent much of August debating whether hyperscaler capital expenditure was approaching a ceiling. Nvidia’s forecast says the company is not seeing that ceiling yet. The immediate constraint is increasingly physical: memory supply, advanced packaging, power and the cost of assembling complete AI systems.
The rally broadened beyond one stock
The semiconductor index rose after the report, with gains in Micron (MU), Broadcom (AVGO) and Intel (INTC). CoreWeave (CRWV), which sells cloud access to accelerated computing, also rallied. Salesforce (CRM) and CrowdStrike (CRWD) jumped after their own results, reinforcing the idea that AI-related spending is showing up in software revenue as well as chip orders.
That breadth is important. A durable AI trade cannot rely indefinitely on one company’s earnings multiple. Investors need evidence that the capital spending produces revenue for cloud operators, software vendors and enterprise customers. This week supplied more of that evidence than the market had seen earlier in the summer.
Rubin moves the debate to execution
Nvidia said the Vera Rubin generation is moving into production and is expected to contribute meaningfully to future revenue. The transition is strategically important because the company has to keep customers upgrading while Blackwell-era systems are still being deployed. A smooth product cadence preserves pricing power; a delayed transition creates an opening for competitors and in-house chips.
The company also expanded its relationship with Amazon (AMZN), with plans involving large-scale GPU deployment through AWS. That makes the next phase of the cycle less about whether hyperscalers want AI capacity and more about whether the supply chain can deliver the systems on time.
Memory is becoming the swing component
High-bandwidth memory has already been one of the tightest parts of the AI supply chain. Nvidia’s commentary puts even more attention on that market because Rubin systems require enormous memory throughput. Micron and Asian memory suppliers therefore sit closer to the centre of the AI capex story than traditional PC-cycle analysis would suggest.
Rising memory prices can be positive for suppliers while pressuring Nvidia’s gross margin. The company expects margins to soften before stabilising, which is a reminder that extraordinary demand does not eliminate input-cost risk. At nearly $100 billion of quarterly revenue, a one-point change in gross margin is material.
The valuation question did not disappear
Nvidia’s results reduce the probability of an abrupt AI spending slowdown. They do not settle the question of how much future growth is already reflected in AI-linked valuations. Several companies in the chain still trade on assumptions that require years of exceptional demand and high utilisation.
The better framing after this report is therefore not whether the AI boom is over. It clearly is not. The question is which part of the value chain can convert rising infrastructure spending into durable free cash flow as supply constraints move from GPUs to memory, power and networking.