From ChatGPT to agentic AI: AWS to host OpenAI’s expanding workloads in $38bn deal

Amazon Web Services (AWS) has entered into a landmark agreement with OpenAI, securing a multi-year, $38 billion infrastructure partnership that positions Amazon Web Services as a core provider for OpenAI’s generative and agentic artificial intelligence workloads. The strategic move will allow OpenAI to scale its compute needs using Amazon Web Services’ global infrastructure, including NVIDIA GPU clusters and Amazon EC2 UltraServers, starting immediately and expanding over the next several years.

The agreement is among the largest infrastructure commitments ever disclosed in the artificial intelligence ecosystem and signals a clear shift in how model developers are sourcing compute resources. OpenAI, known for powering ChatGPT and building frontier large language models, will now use Amazon Web Services to host a wide range of tasks, from inferencing to training next-generation models. The infrastructure, designed to scale to tens of millions of CPUs and hundreds of thousands of GPUs, will support OpenAI’s growing need for reliable, low-latency, and highly secure environments to execute compute-heavy operations.

The transition to Amazon Web Services is not a wholesale abandonment of Microsoft Azure, which continues to host portions of OpenAI’s infrastructure under its longstanding strategic investment. Instead, this partnership marks an expansion and diversification of OpenAI’s infrastructure backbone as it accelerates delivery of real-time and long-context AI systems.

What are the technical specifics of the AI infrastructure OpenAI will use on Amazon Web Services?

The foundation of this partnership rests on Amazon Web Services’ Amazon EC2 UltraServers, which integrate high-density GPU configurations using NVIDIA’s GB200 and GB300 hardware. These are arranged into tightly coupled clusters capable of high-throughput model training and ultra-low-latency inference tasks. The system architecture leverages specialized interconnects and optimized networking stacks to reduce performance drag between parallelized tasks, which is especially critical for the continuous operation of agentic AI systems.

Amazon Web Services stated that it already runs AI clusters at a scale exceeding 500,000 chips globally, giving OpenAI access to a proven infrastructure that combines reliability, redundancy, and power-efficient scaling. This includes the ability to adapt workloads dynamically depending on training cycles, inference volume, and region-specific requirements.

Sam Altman, chief executive officer of OpenAI, indicated that scaling frontier models requires not just chips, but a reliable ecosystem of compute and operational control. He said the partnership with Amazon Web Services reinforces the multi-cloud ecosystem necessary to make advanced artificial intelligence universally available. Matt Garman, chief executive officer of Amazon Web Services, stated that OpenAI’s infrastructure requirements represent the highest level of artificial intelligence scaling seen to date and that Amazon Web Services was uniquely positioned to meet those needs with immediate availability and flexible deployment architecture.

How does this partnership align with OpenAI’s broader ecosystem strategy beyond Microsoft Azure?

The agreement underscores OpenAI’s intent to diversify its infrastructure strategy. While OpenAI remains tightly integrated with Microsoft Corporation and its Azure cloud services, especially for enterprise integrations and developer tooling, this new deal demonstrates a calculated move to avoid over-concentration in any single hyperscale provider.

OpenAI’s growth trajectory has accelerated sharply since the public launch of ChatGPT and the release of GPT-4, which created demand not just for consumer-facing applications but also for enterprise use cases that require stable, mission-critical compute. By turning to Amazon Web Services, OpenAI adds a second operational pillar that can absorb workload surges and serve as a geographic or architectural redundancy partner.

This approach mirrors what other artificial intelligence companies such as Anthropic, Mistral, and Cohere are beginning to pursue. With supply chain constraints on advanced GPUs, access to available compute infrastructure has emerged as a strategic bottleneck, pushing model developers to strike long-term cloud commitments in order to de-risk capacity.

For Amazon Web Services, the partnership provides a high-visibility validation of its leadership in infrastructure performance and readiness. By securing one of the most closely watched artificial intelligence companies in the world, Amazon Web Services is also sending a strong signal to enterprises looking to adopt large-scale models or develop proprietary workflows.

What role does Amazon Bedrock play in this strategic partnership?

Beyond core infrastructure hosting, the relationship between OpenAI and Amazon Web Services is also evolving at the platform level. OpenAI’s open-weight foundation models have been made available on Amazon Bedrock, a fully managed service that provides access to foundation models via an API without managing underlying infrastructure.

Amazon Bedrock allows enterprises to integrate models such as those built by OpenAI into their applications with minimal operational overhead. Since being introduced on the platform earlier in 2025, OpenAI models have become one of the most widely adopted on Amazon Bedrock. Companies such as Bystreet, Peloton, Thomson Reuters, Verana Health, and Triomics are using OpenAI models through Amazon Web Services for a range of agentic and analytical applications including document analysis, scientific reasoning, automated research synthesis, and mathematical operations.

The Bedrock integration offers another path for OpenAI to monetize its models and allows Amazon Web Services to grow its generative artificial intelligence market share without building every model in-house. This also expands the reach of OpenAI models into enterprise workflows hosted directly within Amazon Web Services infrastructure, where companies already store their data and run critical workloads.

How are investors and institutional analysts interpreting the Amazon–OpenAI deal?

Although OpenAI remains privately held, the implications of the deal for Amazon.com Inc. (NASDAQ: AMZN), which owns Amazon Web Services, are drawing significant attention from institutional investors. Market reaction following the announcement was positive, with analysts highlighting the scale and duration of the contract as an important signal of Amazon Web Services’ role in the next phase of artificial intelligence deployment.

The $38 billion figure was interpreted as an upfront and forward-looking commitment, likely to be realized over a seven-year period, with phased rollouts extending into 2026 and beyond. This long-term revenue visibility aligns with Amazon Web Services’ high-margin business model and provides confidence in its ability to win hyperscale artificial intelligence workloads beyond internal Amazon use cases.

Institutional sentiment around Amazon stock has grown more bullish as artificial intelligence becomes a core driver of cloud computing growth. While Microsoft Corporation has captured significant mindshare due to its early OpenAI integration into Office 365 and Azure Copilot, Amazon Web Services is increasingly seen as the infrastructure engine behind many of the largest deployments in the artificial intelligence sector. Investors tracking hyperscaler capital expenditures, particularly in GPU sourcing, networking upgrades, and power provisioning, are viewing this deal as a validation of Amazon Web Services’ investment strategy.

What does this deal indicate about the future of artificial intelligence infrastructure scaling?

The agreement between OpenAI and Amazon Web Services reflects a broader trend: artificial intelligence companies are increasingly realizing that access to compute not just innovation in model architecture is becoming the primary constraint to advancement. The emergence of agentic artificial intelligence systems, which can reason, plan, and execute complex tasks across multiple domains, has significantly raised the bar for infrastructure requirements.

Traditional scaling models that relied on a single provider or region are giving way to hybrid and multi-cloud approaches optimized for task-specific latency, regional compliance, and redundancy. OpenAI’s use of Amazon Web Services infrastructure to support both training and inference suggests that the company is preparing for a more complex, globalized deployment model where resilience, performance, and capacity can no longer be centralized.

Analysts expect other artificial intelligence firms to follow suit by securing long-term compute arrangements to lock in GPU availability and manage costs amid rising demand and volatile hardware supply chains. This also has implications for sovereign infrastructure initiatives, with governments and large enterprises seeking to control or co-locate compute for regulatory and competitive reasons.

As Amazon Web Services scales OpenAI’s workloads, the future of artificial intelligence may increasingly be shaped by those who can deliver not just the smartest models, but also the most dependable and expansive compute foundation.

Key takeaways: What the AWS–OpenAI $38 billion deal reveals about the future of AI infrastructure

• Amazon Web Services has entered a multi-year, $38 billion agreement with OpenAI to host and scale advanced artificial intelligence workloads globally.
• The deal enables OpenAI to immediately begin using Amazon EC2 UltraServers with hundreds of thousands of NVIDIA GB200 and GB300 GPUs, and the ability to scale up to tens of millions of CPUs.
• OpenAI will use the Amazon Web Services infrastructure for multiple applications, including ChatGPT inference, next-generation model training, and agentic AI task execution.
• This marks a significant expansion beyond OpenAI’s existing Microsoft Azure relationship, signaling a more diversified multi-cloud infrastructure strategy.
• The infrastructure includes low-latency GPU clusters, advanced networking, and high-throughput AI performance optimization tailored to OpenAI’s evolving needs.
• The partnership deepens existing collaboration through Amazon Bedrock, where OpenAI’s models are already widely used by customers like Thomson Reuters, Peloton, and Verana Health.
• Investors are viewing the deal as a major validation of Amazon Web Services’ position in hyperscale cloud computing, with positive implications for Amazon.com Inc. stock performance.
• Institutional sentiment suggests this type of long-term compute contracting will become increasingly common as frontier artificial intelligence workloads outpace traditional infrastructure capacity.
• Analysts believe the deal underscores how compute access, not just algorithmic innovation, is becoming the defining battleground in the race to scale artificial intelligence.