Nvidia shares rise after revenue beats outlook

NVDA----SOL----

Nvidia’s data center business generated $89 billion in fiscal 2027 second-quarter revenue, accounting for more than 90% of the company’s $96.221 billion quarterly sales and reinforcing how a small group of cloud companies is shaping the AI infrastructure cycle. Nvidia said demand for accelerated computing remains ahead of available supply, with capacity constraints expected to limit growth through at least fiscal 2028.

The company reported revenue growth of 106% from a year earlier and 18% from the previous quarter for the three months ended July 26, 2026. Non-GAAP net income reached $53.954 billion, up 118% year over year, while adjusted earnings per share came in at $2.22, Nvidia said in its quarterly results.

Nvidia shares rose more than 4% in after-hours trading following the release, reaching $219.53, as the company also issued a fiscal third-quarter revenue forecast of $108 billion. The outlook points to another sizable sequential increase from the July quarter, even as Nvidia says it cannot yet fully meet demand for its systems.

Chief Financial Officer Colette Kress said Nvidia expects fiscal 2028 revenue to rise by roughly 70%. The forecast excludes any data center revenue from China, leaving the company’s projection dependent on demand in other markets and on its ability to obtain enough components, manufacturing capacity, power infrastructure, and data center space.

Supply, not demand, is the constraint

Nvidia’s results place the supply chain at the center of the company’s next stage of growth. Kress said supply limitations are expected to remain the company’s primary growth constraint through at least fiscal 2028, a rare position for a business already producing nearly $100 billion in quarterly revenue.

The bottleneck extends beyond Nvidia’s graphics processors. Large AI deployments require high-bandwidth memory, advanced packaging, networking equipment, data center buildings, cooling systems, and access to large volumes of reliable electricity. Expanding each layer takes time, particularly when cloud companies are seeking capacity at the same time.

Nvidia expects non-GAAP gross margin of about 74% in fiscal Q3, compared with 75% in the second quarter. Management linked the expected decline partly to higher memory costs, including pricing pressure around high-bandwidth memory, or HBM, a specialized type of memory used alongside AI processors to move large amounts of data quickly.

The one-percentage-point change is modest against Nvidia’s current profitability, but it illustrates a practical limit to how much of the AI spending surge can flow directly to chip earnings. As component costs rise and newer platforms enter production, Nvidia must balance pricing, supply availability and margins while its customers seek to build increasingly large clusters.

Cloud platforms drive the data center surge

Nvidia said hyperscale customers generated approximately $48.7 billion of its data center revenue during the quarter. Amazon, Microsoft, Google and Meta continue to expand AI infrastructure spending, according to the company, concentrating a substantial portion of demand among a handful of technology groups with the capital and existing cloud operations to deploy systems at scale.

Amazon Web Services has announced plans to deploy 2 million Nvidia GPUs, Nvidia said. Such commitments demonstrate the scale of compute being ordered, but they also show why hardware supply has become only one part of the deployment challenge. Installing millions of processors requires completed campuses, networking, cooling and utility connections, none of which can be added as quickly as a software service.

Nvidia’s data center revenue rose 117% from a year earlier and 18% from the prior quarter. The segment’s growth far exceeded the company’s smaller edge computing business, which produced $7.2 billion in revenue, up 27% year over year and 13% sequentially.

Nvidia has positioned edge computing as part of its “physical AI” strategy, a term it uses for systems that run AI workloads closer to machines, sensors, factories and other real-world operations rather than solely inside centralized cloud facilities. The business remains small compared with data center revenue, yet it gives Nvidia an avenue to sell hardware and software into robotics, industrial automation and autonomous systems.

Vera Rubin reaches production

Nvidia said its Vera Rubin platform entered full mass production and began shipping earlier in August. The company expects the platform to contribute about 20% of data center revenue in fiscal Q3, a fast initial ramp that would make it a meaningful part of the $108 billion quarterly revenue target.

The platform is part of Nvidia’s effort to maintain a rapid upgrade cycle for AI infrastructure. New architectures can prompt cloud operators to refresh or expand their fleets, although the value of each launch depends on the availability of supporting memory, networking and data center capacity.

Management estimated that every gigawatt of deployed compute capacity represents a potential $40 billion revenue opportunity for Nvidia. The figure underscores how AI infrastructure is increasingly measured in power as well as processor counts. A gigawatt-scale deployment can require dedicated generation or major grid upgrades, which brings utilities, landowners, construction firms and infrastructure financiers into the same buildout previously dominated by chipmakers and cloud providers.

Nvidia recruits infrastructure capital

Nvidia also disclosed partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR aimed at mobilizing more than $500 billion of third-party capital for AI infrastructure. The company is working with SB Energy at the PORTS-Pike Technology Campus in Ohio to secure land, power and building-shell capacity for compute deployments.

Those arrangements suggest Nvidia is seeking to reduce friction beyond its own supply chain. The company can manufacture and ship AI systems, but customers cannot recognize the full value of those systems until sites are powered, constructed and equipped. Bringing large asset managers into the process could give data center projects access to financing structures more commonly used for energy, transport and real estate assets.

Nvidia’s quarterly figures show that AI spending is moving from a chip-purchasing cycle into a larger contest for physical capacity. The company’s revenue forecast remains exceptional, yet management’s own guidance makes clear that the next constraint is increasingly likely to be the ability to build, power and cool the facilities needed to run its hardware.