OpenAI has made a significant move to diversify its cloud computing presence. The company announced a seven-year agreement with Amazon Web Services (AWS) worth $38 billion. This deal greatly expands OpenAI’s dependence on external cloud infrastructure and reflects the changing alliances in the global AI landscape.
This agreement positions Amazon as a key technology partner for OpenAI, the creator of ChatGPT, which is usually seen as Microsoft’s main AI collaborator. It comes at a time when demand for high-performance computing is skyrocketing.
A Shift Away from Microsoft
For years, OpenAI’s most demanding training workloads took place almost entirely on Microsoft’s Azure cloud. That exclusivity was relaxed recently after a renegotiated partnership that allowed OpenAI more freedom to get computing capacity from other providers.
The new AWS agreement is the first major action under this more flexible deal. It is also one of the largest cloud agreements in the AI industry so far.
AWS will provide the resources needed to train OpenAI’s next generation of advanced models and support the scale-up of ChatGPT, which is rapidly growing in the consumer and enterprise markets. According to Amazon, OpenAI will use the newly launched EC2 UltraServers, machines built with hundreds of thousands of Nvidia GPUs aimed specifically at AI training tasks.
After the announcement, Amazon’s shares jumped nearly 5% in early trading, highlighting investor excitement about cloud providers winning long-term AI contracts that secure steady revenue for years.
“Scaling frontier AI requires massive, reliable compute”
OpenAI’s CEO Sam Altman highlighted the scale of the infrastructure necessary to keep the company on its current path. He stated, “Scaling frontier AI requires massive, reliable compute.” Altman added that the partnership with AWS “strengthens the broad compute ecosystem” needed to advance AI into its next phase while making new capabilities accessible to more users.
Matt Garman, AWS’s CEO, described the deal as a foundation for OpenAI’s future goals. He mentioned, “We expect AWS to be a backbone for OpenAI’s most demanding workloads,” noting the company’s commitment to next-generation GPU and AI-optimized technology.
AWS Builds Capacity for OpenAI
In an interview with CNBC, Dave Brown, Amazon’s vice president for compute and machine-learning services, said AWS is developing “completely separate capacity” for OpenAI. This dedicated infrastructure is intended to meet the performance and availability needed for training increasingly complex models.
“Some of that capacity is already live, and OpenAI is using it today,” Brown explained, but he did not specify which models or training sessions have already started using AWS hardware.
Amazon’s Concurrent Support for Anthropic Adds Complexity
OpenAI isn’t the only AI company Amazon is supporting. The tech giant has invested over $8 billion in Anthropic, the creator of the Claude AI models, and has promised significant cloud resources for the startup’s plans.
Earlier this year, Amazon opened Project Rainier, an $11 billion data center campus primarily focused on Anthropic’s training activities. This facility uses hundreds of thousands of Amazon-designed Trainium 2 chips, which aims to reduce reliance on Nvidia hardware while creating a more integrated AI infrastructure.
Amazon’s simultaneous partnerships with OpenAI and Anthropic show its ambition to lead the infrastructure aspect of the AI boom.
Microsoft Balances OpenAI and Anthropic
The new OpenAI–AWS agreement comes at a critical time for Microsoft. While Microsoft remains OpenAI’s largest investor and main cloud partner, it is also integrating Anthropic’s Claude models into various platforms, including:
- Microsoft 365 Copilot
- GitHub Copilot
- Azure AI Foundry
These actions suggest a broader strategy to secure access to leading AI systems even if OpenAI expands its cloud options or makes its models available across the tech industry.
Despite the changes, Microsoft has retained significant leverage. The updated contract keeps Microsoft’s exclusive rights among major cloud providers to distribute OpenAI’s proprietary models through the Azure OpenAI Service until at least 2032.
Additionally, OpenAI has agreed to spend an extra $250 billion on Microsoft services over the coming years, ensuring that Azure continues to be a vital part of its operations.
Why OpenAI Is Using Multiple Clouds
The recent deal with Amazon fits into a larger trend. In the past few months, OpenAI has also made major agreements with Oracle and Google, collectively committing hundreds of billions of dollars to secure future AI computing resources.
Several factors fuel this multi-cloud approach:
1. Unprecedented Compute Demand
Training advanced models requires huge clusters of GPUs or other AI-optimized hardware, available nonstop for months. No single cloud provider has enough capacity to meet OpenAI’s peak needs alone.
2. Supply Chain Challenges
The current global shortage of Nvidia chips has forced AI companies to source hardware from multiple partners to avoid training delays.
3. Strategic Independence
Relying on just one cloud provider makes AI labs vulnerable to outages, geopolitical supply issues, and changing partner incentives. Using multiple clouds serves as a kind of operational safety net.
4. Competitive Pressure
OpenAI’s main competitors, including Anthropic, Google DeepMind, and Meta, are rapidly increasing their training scale. Securing long-term computing contracts helps ensure OpenAI keeps up.
Amazon and OpenAI Strengthen Their Relationship After Open-Weight Model Release
Before this agreement, OpenAI and Amazon had already been working more closely together. Earlier this year, Amazon included OpenAI’s first open-weight models in over five years in AWS Bedrock and SageMaker.
Since these models were released under an open-source license and were not limited by the exclusive API agreement OpenAI has with Microsoft, Amazon could distribute them without any legal issues.
This cooperation laid the groundwork for Amazon’s offer to serve OpenAI in two ways: by providing open models for developers and large clusters for cutting-edge research.
Concerns About the Sustainability of AI Infrastructure Spending
While Big Tech races to secure AI infrastructure capacity, analysts worry about the long-term economics of these investments.
Corporate clients and investors are asking:
- Can generative AI revenue justify the hundreds of billions spent on GPUs, data centers, and cooling systems?
- Will oversupply become a risk if AI adoption slows?
- Are AI providers at risk of becoming overly dependent on a few cloud giants?
For now, companies like Amazon, Microsoft, and Google see AI infrastructure as a key investment, similar to the early days of smartphones or the move to cloud computing. However, the scale of spending is much larger than past technology cycles, and the timeline for seeing financial returns remains unclear.
A Defining Moment in the AI Compute Landscape
OpenAI’s choice to collaborate with Amazon on a $38 billion cloud contract shows a growing awareness that the future of AI will depend as much on access to computing power as on model quality or innovations in algorithms.
With this deal, AWS becomes a vital part of OpenAI’s infrastructure strategy, and the overall cloud market becomes even more competitive. The agreement also highlights the increasingly intricate relationships among major tech companies, all of which are working together, competing, and investing in various aspects of the AI ecosystem.
As the industry pushes toward larger and more capable models, these partnerships will determine not only who leads the AI frontier but also how sustainable the sector will be in the long run.
