Nvidia GPU Provider Market Expands to 323 as Access Routes Multiply

Companies now have more ways to obtain Nvidia GPUs as the provider market expands beyond traditional cloud platforms and specialized AI infrastructure firms increase available capacity.

The market reached 323 GPU providers as of Sept. 23, up from 209 in the previous review. That is an increase of about 55%. The review examined 77 providers in depth and included interviews with more than 200 neocloud users.

Large cloud platforms such as Amazon Web Services, Google Cloud and Microsoft Azure offer Nvidia GPUs alongside storage, networking, security and other services. Specialized neoclouds focus on accelerated computing and may provide regional capacity, newer systems or bare-metal access.

Companies with enough capital can also purchase GPU servers and run them in on-premises data centers. That approach offers more control but requires electricity, space, networking and technical staff. Some customers instead place purchased hardware in cloud facilities.

CoreWeave said Sept. 17 that it had signed approximately $25 billion in additional customer commitments early in the third quarter. Its contracted power reached approximately 4.2 gigawatts as of Aug. 11, up from approximately 3.7 gigawatts on June 30.

Nvidia's filings classify cloud providers, AI model developers, equipment makers, distributors and system integrators as direct customers; AI clouds, enterprises and public-sector organizations are indirect customers. The distinction means the ultimate GPU user may not be Nvidia's direct customer.

Nvidia reported $96.221 billion in revenue for the fiscal second quarter ended July 26, 2026, including $89.023 billion from its Data Center business. It forecast fiscal third-quarter revenue of $108.0 billion, plus or minus 2%.

One direct customer accounted for 16% of fiscal second-quarter revenue. Three direct customers represented 16%, 15% and 13% of Nvidia's revenue during the first half of fiscal 2027.

Nvidia expects regional neoclouds, sovereign AI providers, enterprises, edge systems and air-gapped data centers to contribute about half of its Data Center business.

The company also described arrangements in which it commits to purchase part of a neocloud facility's capacity and shares in revenue above a minimum floor. Nvidia said those arrangements are supported by independent capital and are not loans.

Nvidia CEO Jensen Huang said, “We don't own a cloud ourselves. We are a neutral partner to every sovereign and neocloud.”

Nvidia listed CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius among the companies operating Vera Rubin systems. For customers, the expanding market creates more options based on ownership, location, hardware generation, delivery timing and operating responsibility.