In a significant strategic pivot, Nvidia has reorganized its cloud computing unit, effectively scaling back its ambitions to operate a public cloud service that would have competed directly with giants like Amazon Web Services (AWS) and Microsoft Azure. This move, reported by The Information and confirmed by internal restructuring, marks a decisive shift for the DGX Cloud service less than three years after its high-profile launch. The decision underscores the complex dynamics between a dominant hardware supplier and its largest customers, who are also its potential rivals in the cloud services arena.
Nvidia Reorganizes Cloud Team, Folds DGX Cloud into Engineering
Nvidia has officially restructured its cloud computing group, folding the DGX Cloud business into its core engineering organization. The unit is now under the leadership of Senior Vice President Dwight Diercks, who oversees software engineering. This reorganization, which occurred in the week leading up to the Christmas holiday in the United States, represents a formal scaling back of Nvidia's initial vision for DGX Cloud as a branded, public-facing cloud platform. The company had already signaled a change in direction in September 2025, announcing it would stop trying to compete directly with AWS and Azure. The latest move solidifies this new focus, transitioning DGX Cloud from a customer-facing product to an internal tool for Nvidia's own research and development.
The Original Vision and Practical Challenges of DGX Cloud
Launched in early 2023, DGX Cloud was conceived as Nvidia's ambitious foray into the cloud services market. The service aimed to abstract the company's powerful DGX AI supercomputers into a managed service, hosted on infrastructure from partners including AWS, Google Cloud, Oracle Cloud, and Microsoft Azure. It offered enterprises dedicated clusters powered by Nvidia's H100 GPUs with the full Nvidia AI software stack pre-installed, promising a streamlined path to cutting-edge AI compute. However, the model faced significant headwinds in practice. Pricing was often higher than comparable GPU instances from hyperscalers, integration with existing cloud toolchains was inconsistent, and the split support model between Nvidia and its hosting partners created operational complexity for customers.
DGX Cloud Hosting Partners (Original Model): AWS, Google Cloud, Oracle Cloud, Microsoft Azure.
Avoiding Channel Conflict with Key Hyperscaler Customers
A primary driver behind the strategic retreat is the need to avoid channel conflict with Nvidia's most important customers. Cloud providers like AWS, Microsoft, Google, and Oracle represent a massive portion of Nvidia's record data center revenue. By operating a first-party cloud service, Nvidia risked becoming a direct competitor to these very companies, who are committing billions of dollars to purchase its GPUs. This friction was deemed counterproductive. The reorganization aligns Nvidia's interests more clearly with its partners, positioning DGX Cloud as a development and validation platform that ultimately helps hyperscalers deploy Nvidia hardware more effectively, rather than as a rival service.
The New Role: An Internal Platform for AI and Silicon Development
With its new mandate, DGX Cloud will serve as a critical internal platform for Nvidia's engineers. Its primary functions will now include AI model development, software validation, and pre-silicon and post-silicon testing for upcoming GPU platforms like the anticipated Blackwell series. This "internal testbed" strategy allows Nvidia to simulate demanding AI workloads at scale to identify hardware and software bottlenecks, accelerating its own product iteration cycles. This shift leverages the platform's capabilities to directly strengthen Nvidia's core competency in chip and system design, turning what was an external commercial effort into a potent internal R&D tool.
Economic Realities and the Hyperscaler Silicon Landscape
Operating a competitive global cloud platform is a capital-intensive endeavor requiring massive ongoing investment in data centers, networking, and operations—areas outside Nvidia's core expertise in silicon and software. Furthermore, the hyperscaler partners that hosted DGX Cloud have been aggressively developing their own custom AI silicon, such as AWS's Trainium, Google's TPUs, and Microsoft's Maia chips, to manage costs and supply risks. While these companies remain heavily dependent on Nvidia GPUs for leading-edge AI, the competitive landscape made the long-term viability of a first-party Nvidia cloud service uncertain. Stepping back from this space allows Nvidia to concentrate its resources on maintaining its architectural leadership.
Reported GPU Leaseback Deal: Nvidia leased 18,000 GPUs (value: USD 1.5 billion) over four years from a cloud partner, highlighting complex supply dynamics.
Implications for Nvidia's Broader Cloud Strategy
This restructuring does not signify a reduced commitment to cloud technologies overall. Instead, it reflects a refinement of strategy. Nvidia will continue to heavily invest in and expand its essential software ecosystems like CUDA, TensorRT, and its inference frameworks, which are designed to run optimally across all major cloud platforms. DGX Cloud, in its new form, will act as a proving ground for these technologies. The move is widely seen as a pragmatic decision that stabilizes Nvidia's crucial partnerships, reaffirms its role as an enabler of AI infrastructure, and allows it to focus capital on high-margin R&D. It concludes a two-year experiment in moving up the stack, yielding a clearer, more focused path forward as the indispensable "picks and shovels" provider of the AI era.