Nvidia’s quarterly revenue reached $96.2 billion, up 106% from a year earlier, as the company at the center of the artificial-intelligence infrastructure buildout delivered another result that would once have seemed implausible for a chipmaker. The Aug. 26 filing showed data-center revenue of $89 billion, a 117% increase, while net income rose to $59.7 billion for the fiscal second quarter ended July 26.

The results exceeded Wall Street’s already elevated expectations and prompted Nvidia to forecast approximately $108 billion in revenue for the current quarter. That outlook would represent another sharp increase and reinforces a central fact of the technology economy: spending on the computing systems used to train and run AI models remains one of the largest corporate-investment cycles underway anywhere in the world.

Yet the report also clarified the risks beneath that expansion. Nvidia faces shortages of advanced memory, questions about the returns its customers will earn on immense data-center budgets, limits on sales to China and scrutiny of financing arrangements that help buyers fund AI infrastructure. The company’s numbers show that demand remains exceptionally strong; they do not settle whether every participant in the broader investment cycle will ultimately earn an adequate return.

Data Centers Now Define the Company

Nvidia’s transformation is visible in the composition of its revenue. Data-center sales accounted for roughly 93% of quarterly revenue, leaving gaming and other businesses as comparatively small contributors. Just one year earlier, the company’s total quarterly revenue was $46.7 billion, according to its prior-year results. In twelve months, Nvidia added nearly $50 billion of quarterly sales.

The growth is being driven by large cloud providers, AI laboratories, enterprises and governments buying accelerated-computing systems. These purchases encompass more than individual graphics processors. Nvidia sells integrated racks, networking equipment, central processors, software and support, allowing customers to deploy large AI clusters as coordinated systems. That broader platform strategy helps explain why competing chips do not automatically displace Nvidia even when customers design specialized processors of their own.

The quarter also generated $2.46 in diluted earnings per share under generally accepted accounting principles and $2.22 on an adjusted basis. Independent estimates had placed revenue near $92.3 billion and adjusted earnings at about $2.09 per share, meaning Nvidia cleared expectations that had already assumed extraordinary growth.

Amazon Expands the Next Wave of Demand

The most concrete sign that the buildout is continuing came from Amazon Web Services. Nvidia and Amazon announced that AWS plans to deploy two million additional Nvidia GPUs across its global infrastructure during 2027 and 2028. The expanded partnership spans Blackwell Ultra, Rubin and Rubin Ultra systems and extends into Amazon’s warehouse-robotics operations.

The scale matters because cloud companies are the primary channel through which many businesses rent AI computing. Adding two million processors does not mean all will be installed at once or immediately run at full utilization, but it represents a substantial multi-year commitment to capacity. It also signals that Amazon expects customer demand to support continued investment even as it develops its own Trainium chips.

Nvidia said it expects revenue for the fiscal year ending in January 2028 to grow by about 70%, far above the 44% expansion analysts had anticipated, according to Reuters. That forecast depends on continued demand for AI inference—the computing required when deployed models answer questions or take actions—as well as training increasingly capable models. The shift from experimentation to routine use is important because inference can create recurring demand rather than a one-time development expense.

Supply Constraints Are Becoming Economic Constraints

Nvidia’s limiting factor is increasingly the availability of the components surrounding its processors. Advanced AI systems require high-bandwidth memory, networking equipment, power systems and specialized packaging in addition to GPUs. Shortages in any one layer can delay an entire installation. Nvidia warned that memory constraints would continue to restrict how quickly it can increase shipments even when customers are ready to buy.

Those constraints are also pressuring margins. Nvidia reported a 75% gross margin for the quarter, meaning about three-quarters of revenue remained after the direct cost of products sold. Management expects that measure to decline as memory and other input costs rise before recovering later. The tension is straightforward: scarcity supports demand and pricing power, but it also makes each finished system more expensive to produce.

This is why Nvidia’s earnings have implications beyond one company. The AI boom is pulling investment toward memory producers, semiconductor foundries, electrical-equipment manufacturers, utilities and construction firms. It is simultaneously competing with consumer electronics, conventional servers and industrial equipment for scarce components. The result can be higher prices and longer lead times in markets that receive none of the direct revenue reported by Nvidia.

The Financing Debate Follows the Spending Boom

The magnitude of AI capital spending has produced questions about how the infrastructure is financed and whether transactions are truly independent. Nvidia has invested in customers and projects that may later buy its equipment, while financial institutions are assembling large pools of capital for data centers. Critics describe some arrangements as circular because money can move from suppliers or investors to buyers and then return as equipment revenue.

The concern does not mean the reported sales are fictitious. It asks whether demand would remain as strong without supplier-backed financing, guarantees or equity investments, and who absorbs losses if a project fails to generate enough cash. Nvidia says demand exceeds supply and that its financing exposure is supported by valuable infrastructure and committed users. The Financial Times reported that the company is helping mobilize a $500 billion capital pool while defending its investment strategy against those criticisms.

Cash flow provides another useful measure. Free cash flow declined from $48.6 billion to $21.3 billion sequentially, according to the same reporting, even as accounting profit increased. Timing of customer payments and working-capital needs can cause large quarterly swings, so one period does not establish deterioration. Still, the difference is a reminder that profit, cash collection and financed infrastructure commitments must be evaluated separately during a capital-intensive expansion.

China and Competition Remain Structural Limits

Nvidia’s strongest growth is occurring while its China data-center business remains constrained by U.S. export controls and Chinese regulatory barriers. The company assumed no meaningful data-center compute revenue from China in its outlook. That removes a historically important market and creates space for domestic Chinese chipmakers to develop alternatives, potentially weakening Nvidia’s long-term position even if restrictions were later eased.

Competition is also coming from customers. Amazon, Google and other major cloud providers are developing custom AI accelerators designed for specific workloads. Those chips can lower costs and reduce dependence on Nvidia, particularly for internally controlled applications. Nvidia’s defense is its software ecosystem, rapid product cadence and ability to sell complete systems that work across multiple clouds and use cases.

The market’s initial response reflected both optimism and the unusually high standard Nvidia must meet. Shares first moved modestly, then strengthened after management’s forecast and the Amazon expansion became clear. Broader market coverage described the reaction as evidence that investors still view Nvidia as the clearest public-market proxy for AI demand.

Revenue Growth Is Proven; Customer Returns Are Not

Nvidia has demonstrated that the AI infrastructure boom is producing real revenue, cash and profit at unprecedented scale. Its quarterly sales now exceed the annual revenue of most large corporations, and its current forecast implies that growth will continue despite a much larger base. The Amazon agreement adds a tangible commitment extending through 2028.

The unresolved issue sits one step downstream. Cloud providers and AI companies must convert expensive processors, data centers and electricity into services customers will buy at prices that justify the investment. Some will succeed through productivity gains, new products and automation. Others may discover that usage, pricing or competitive advantage falls short of their projections.

That distinction makes Nvidia’s report both powerful and incomplete as a measure of the broader economy. The company has proved that suppliers can earn extraordinary returns during the buildout. The next phase will test whether the buyers of those systems can do the same. Until that evidence arrives, $96.2 billion in quarterly revenue is the clearest sign yet that the investment boom is accelerating—and the largest reminder of how much future value the market is already assuming.