Nvidia's Earnings Reveal a New Buyer Class Outgrowing Microsoft, Google and Amazon

Nvidia's (NASDAQ: NVDA) second-quarter 2027 (ended July 26, 2026) earnings report made the usual point in unusually large numbers. Data center revenue reached $89 billion during the quarter, up 18% from the prior quarter and 117% year over year. What I think matters more than the headline figure is the bifurcation inside that number.

Hyperscale customers generated $48.7 billion in revenue, up 13% sequentially and 102% from a year ago. The AI Clouds, Industrial, and Enterprise group (ACIE) generated $40.3 billion, an increase of 25% from last quarter and 138% year over year.

While big tech is still the largest buyer of Nvidia's systems, it is no longer the fastest growing. That shift changes how the AI story should be read.

Image source: Nvidia.

The hyperscalers form Nvidia's foundation

The hyperscale bucket primarily revolves around the public cloud oligopoly: Amazon Web Services (AWS), Microsoft Azure, and Google Cloud Platform (GCP). These companies aren't buying Nvidia's products as a hobby. They buy them because GPU clusters have become a scarce resource supporting training runs, large-scale inference deployments, and the rented capacity that developers consume.

Despite designing its own Trainium, Inferentia, and Graviton silicon, AWS remains one of Nvidia's largest customers. The reason is simple: Nvidia's GPUs and CPUs play a critical role in how start-ups, labs, and enterprises rent Blackwell and Rubin chips from cloud providers without owning their own facilities packed with liquid-cooled racks.

Azure has bound itself to this same stack. Microsoft's Copilot suite, OpenAI-related training, and Azure-OpenAI services all rely on Nvidia systems. This structure is unique, as it makes Microsoft both a customer of and a distribution channel for Nvidia.

Google Cloud also designs its own custom silicon, called Tensor Processing Units (TPUs). Even so, Alphabet still buys enormous quantities of Nvidia hardware. One reason why is that many customers want access to CUDA, Nvidia's software ecosystem that runs on top of its GPUs. This makes the ability to move AI workloads across clouds much more efficient. For now, Google's TPUs cannot serve that type of demand on its own.

Taken together, these three cloud hyperscalers form the floor supporting Nvidia's data center operation. By signing multiyear capacity plans they essentially turn Nvidia's racks into an infrastructure-as-a-service empire. When revenue from the hyperscalers more than doubles in a year, it is evidence that the largest buyers see returns on procuring more accelerators. It is not evidence, however, that the GPU market only has three buyers.

How new adoption is becoming a big market for Nvidia

Analyzing the results from the ACIE customers is where Nvidia's story gets more interesting. Nvidia describes this category as AI natives, enterprises, sovereign customers, and the specialist clouds -- neoscalers like CoreWeave and Nebius Group -- that sit between a hyperscaler and a data center. During the second quarter, this mix of customers grew almost twice as fast sequentially as hyperscale. Over the last year, it grew even faster.

This is unique because it shows that AI budgets are not recycled through the same three offices. An AI cloud is an emerging kind of intermediary, one that buys Nvidia's chips, packs them into clusters, and leases capacity to companies that never need to negotiate with a chip designer directly. Industrial and enterprise buyers are even different. Manufacturers running digital twins on the factory floor, a bank assessing risk, or a government designing a sovereign cluster is far more complex than a corporation simply increasing its operating budget to rent incremental storage on AWS.

This distinction is important because skeptics seem to think that the AI revolution is confined to Amazon, Microsoft, and Alphabet. This gives bears an excuse to call AI a circular trade. Cloud giants spend on GPUs so they can sell AI services whose customers are none other than frontier model developers. But when ACIE customers outgrow the hyperscalers, Nvidia's roster of buyers becomes larger. This proves that demand is moving beyond a small cohort of platform businesses to a broader class that also needs generative models for production use cases, not just for platform differentiators.

Most importantly, rising sales from ACIE dampens the AI bubble argument. A bubble story has to explain why Nvidia's non-hyperscale book is accelerating. A new class of buyers means the capex supercycle is developing a second demand curve, and second demand curves are how infrastructure booms wind up creating new industries.

Where is the best place to invest to capitalize on AI infrastructure?

Don't get me wrong: this analysis is not to say that hyperscalers have become unimportant. AWS, Azure, and Google Cloud still remain the clearinghouses for much of the world's rented compute. Enterprise software names that are able to attach new AI-driven use cases to customer licenses will benefit if the ACIE demand is in fact real.

At the end of the day, hyperscalers still need Nvidia's systems to keep their AI products competitive. Neoscalers need GPUs and public clouds to exist to even have a viable business. Enterprises and sovereign governments rely on Nvidia's ecosystem spanning software, networking, and rack-scale design. This is all to say that accelerating ACIE growth shows how Nvidia's customer base is widening while the pick-and-shovel layer of capacity remains fairly concentrated.

Investors that fixate on the three cloud giants are stuck watching the first chapter of the AI narrative. Meanwhile, those watching enterprise software without asking who supplies the underlying compute are distracted by a third chapter that is still being written. The second chapter -- the one Nvidia's quarter actually highlights -- shines a light on the vendor that sells the AI factory to both. If AI adoption is truly broadening rather than looping, then Nvidia's scarce, high-margin systems will continue to be where evidence shows up first. In my eyes, that's what makes Nvidia such a no-brainer stock to buy and hold in the AI infrastructure era.

Adam Spatacco has positions in Alphabet, Amazon, Microsoft, and Nvidia. The Motley Fool has positions in and recommends Alphabet, Amazon, Microsoft, and Nvidia. The Motley Fool has a disclosure policy.