Amazon jumps on OpenAI’s new releases — and it’s not even a surprise anymore

A day after Microsoft gave up its exclusive grip on OpenAI’s models, AWS was already out the gate with a fresh batch of offerings. That includes OpenAI’s newest models and something they’re calling an “agent service.” If you blinked, you might have missed the pivot.

Let’s be real: this isn’t shocking. The writing has been on the wall for months. Microsoft’s exclusive deal with OpenAI was always going to be a temporary arrangement — the kind of thing that looks good on a press release but makes less sense as the market matures. Once Microsoft agreed to let go, the floodgates opened. AWS, being AWS, didn’t waste any time.

The agent service part is what caught my attention. Amazon is positioning this as a way to build autonomous AI agents that can perform tasks — think booking flights, managing schedules, or automating customer support workflows. It’s not a new concept; we’ve seen similar moves from Google Cloud and Microsoft themselves. But having it on AWS infrastructure, with OpenAI’s models under the hood, adds a different flavor. The question is whether developers will trust a third-party agent layer when they could just build their own with the raw API.

I’ve been playing with the preview. The setup is smoother than I expected — you can define tasks in natural language, and the agent figures out the steps. It’s not perfect. There’s still a tendency to hallucinate when the task gets too complex, and the pricing feels like it’s designed for enterprise budgets, not indie hackers. But for a first cut, it’s usable.

What’s more interesting is what this means for the cloud wars. AWS now has direct access to OpenAI’s latest models, which means they can compete head-to-head with Microsoft’s Azure OpenAI Service. Amazon’s advantage is its existing enterprise relationships and the sheer breadth of its ecosystem. If you’re already running your infrastructure on AWS, adding OpenAI models becomes a no-brainer. No need to manage a separate Azure account or deal with cross-cloud latency.

Of course, there’s a catch. OpenAI’s models are still best-in-class for many tasks, but the gap is narrowing. Google’s Gemini, Meta’s Llama, and even Anthropic’s Claude are catching up fast. Amazon knows this. That’s why they’re hedging their bets by also offering models from other providers through Bedrock. The OpenAI offering is just one piece of a larger strategy.

I’m not entirely sold on the agent service yet. It feels like a solution in search of a problem for most users. But for enterprises that need to automate complex workflows without hiring a team of engineers, it could be a game-changer. The real test will be in the next few months, when we see how well these agents handle real-world edge cases.

One thing’s for sure: the era of exclusive AI model deals is over. Microsoft had its run, but the market is moving toward openness — or at least, multi-cloud pragmatism. Amazon’s move is a clear signal that they’re not going to sit on the sidelines while others shape the AI infrastructure landscape. They’re playing to win, and they’re playing with OpenAI’s best tools.

Whether that’s enough to shift the balance of power remains to be seen. But for now, if you’re an AWS customer, you’ve got a shiny new set of options. Use them wisely.