AWS Signs Record $38bn Deal With OpenAI to Power Next‑Gen AI
The deal marks a major shift in cloud‑AI alliances and accelerates compute scale for both firms.
The power couple nobody saw coming
OpenAI just signed a $38 billion deal with Amazon Web Services, and it’s not just about servers. It’s about power (literal, computational, and strategic).
The agreement, announced Monday, hands OpenAI access to a vast arsenal of AWS infrastructure: hundreds of thousands of NVIDIA GPUs (think GB200s and GB300s) and the ability to scale to tens of millions of CPUs via EC2 UltraServers. OpenAI gets to start using that hardware now. The full setup will be in place by the end of 2026, and the expansion window runs until at least 2032.
It’s one of the largest cloud compute deals ever signed. But the real story isn’t just the size. It’s the fact that OpenAI, Microsoft’s golden child and exclusive partner—until recently—just slipped out the side door and shook hands with Microsoft’s fiercest cloud rival.
ChatGPT, now hosted on Amazon
What’s OpenAI using all that silicon for? Training and running its most advanced models, including the ones behind ChatGPT and its forthcoming successors. That means everything from consumer-facing tools to agentic systems capable of autonomous reasoning, code generation, research assistance, and likely a dozen other things still under NDA.
“Scaling frontier AI requires massive, reliable compute,” said OpenAI CEO Sam Altman, in a statement that could be politely interpreted as “we needed more options.” AWS, of course, was happy to oblige. Matt Garman, AWS’s newly minted CEO, called the deal proof of why Amazon is “uniquely positioned” to handle the scale of OpenAI’s ambitions.
It’s a big win for Garman, who took over AWS in a year that’s seen the cloud giant scrambling to catch up to Microsoft and Google in the generative AI space. Signing OpenAI,arguably the most influential AI company on the planet—is about as loud a comeback signal as you can send.
Microsoft: Still invited to the wedding?
This doesn’t mean OpenAI is ghosting Microsoft entirely. Redmond still holds a multi-billion-dollar stake in OpenAI, and Azure will remain a key platform. But exclusivity? That’s dead. And the timing is interesting. OpenAI only regained the freedom to negotiate its own compute deals after a governance shakeup earlier this year that restructured its board and clarified its independence.
If you’re Microsoft, the message is brutal: even your closest ally wants a backup plan.
Compute is currency
The deal also reflects a broader truth in AI development: it’s no longer just about clever algorithms—it’s about having access to enough chips to actually run them. As models balloon in size and complexity, infrastructure has become the real bottleneck. That’s why AWS is clustering GPUs like rare minerals, wiring them together in low-latency fabrics, and bragging about topping 500,000 chips in a single system.
It’s not just about raw power either. OpenAI’s workloads require precision-tuned environments; data centres built for training runs that can take weeks, inference at massive scale, and architectures that can bend to fit the next idea Altman’s team dreams up. Amazon is betting that its cloud is flexible enough to be that sandbox.
Bedrock, now with more OpenAI
The partnership extends beyond backend compute. OpenAI’s models will also become more deeply integrated into Amazon Bedrock, AWS’s foundation model platform for enterprises. The goal: help customers drop ChatGPT-style smarts into their products without needing a PhD—or a GPU cluster the size of Denmark.
It’s a sales pitch aimed at exactly the kind of companies AWS already courts: logistics giants, media firms, pharma labs, retailers. All looking for ways to stay relevant in a world being reshaped by machine intelligence.
The cloud wars just got personal
This is more than a tech deal. It’s a tectonic shift. For years, the cloud giants fought with speeds and feeds. Now they’re battling for control of the AI layer that will sit on top of it all.
With this move, OpenAI gets redundancy, leverage, and a lot of bargaining power. Amazon gets credibility, relevance, and a marquee tenant whose name alone will move contracts.
And the rest of the industry? Buckle up. The AI gold rush is still in its early days. But the biggest claim stakers are already choosing sides—and buying a lot of compute to defend their turf.