Nvidia may have just admitted that it cannot take Amazon, Google and Microsoft in Cloud

Nvidia has abandoned its plan to compete directly with Amazon Web Services—the cloud market leader commanding roughly 30% share—along with Google Cloud and Microsoft Azure, the big three dominating the industry.

The chip giant is restructuring its DGX Cloud division to focus on internal development rather than selling services to external customers, with the cloud team now serving Nvidia's own engineers building AI models, The Information reported.The company reassigned cloud business unit head Alexis Black Bjorlin, who joined from Meta in 2023, while integrating the DGX Cloud team into its engineering organization led by Vice President Dwight Diercks, according to internal memos reviewed by The Information.

The restructuring reverses CEO Jensen Huang's ambitious March 2023 vision of rivaling industry leaders.Several former employees told The Information that the initiative faced insurmountable challenges from the start. The company had initially promoted early customers including ServiceNow, SAP, and Amdocs, positioning DGX Cloud as delivering superior performance compared to standard cloud configurations.

However, Nvidia struggled to build momentum in a market already dominated by its largest chip customers.

Technical hurdles and fear of alienating major chip buyers doom cloud venture

DGX Cloud consistently failed to attract sufficient customers despite its promise of superior chip performance compared to traditional cloud providers, former employees told The Information. The service's fundamental problem was technical: operating across different cloud providers' data centers made troubleshooting nearly impossible, as fixes implemented in Amazon's AWS facilities wouldn't work in Google or Microsoft environments.The Information reported that Nvidia had leased servers from major cloud providers, customized them to exacting standards, then attempted to rent them to AI developers at premium rates. DGX Cloud charged $36,999 per month for H100 instances when it launched in 2023, but that pricing became unsustainable as AWS slashed its own H100 and A100 prices by up to 45%.More critically, Huang grew reluctant to expand the business, fearing it would alienate AWS, Google, and Microsoft—companies generating roughly half of Nvidia's revenue through GPU purchases, sources told The Information.

The cloud giants viewed DGX Cloud as a competitive threat, potentially accelerating their development of proprietary AI chips. Amazon notably declined to participate in the DGX Cloud program when approached, while Microsoft, Google, and Oracle agreed to host the service.

Nvidia pivots to protect 80% AI chip monopoly over cloud ambitions

The strategic withdrawal reflects Nvidia's calculation that protecting its AI chip monopoly matters more than capturing cloud revenue. While the company previously told investors DGX Cloud could generate $150 billion in future revenue—exceeding AWS's current annual sales—that ambitious projection now appears abandoned as the company refocuses on its core business.Nvidia's newer marketplace service, DGX Cloud Lepton, also struggled to gain traction and will be absorbed into the engineering team, The Information reported. The platform, which allowed cloud providers to list unused Nvidia server capacity, saw limited participation, with some providers joining only because Nvidia required it for joint marketing campaigns.Despite retreating from cloud competition, Nvidia maintains dominance in AI chip sales, controlling over 80% of the market. The company plans to invest $26 billion leasing servers from cloud providers—becoming one of the largest customers of the companies it briefly tried to challenge. Nvidia will "continue to invest in DGX Cloud" to support internal R&D and provide software capabilities to cloud partners, a spokesperson said.