Amazon Flips the Switch: OpenAI Models Go Live on AWS Hours After Microsoft Vows to 'Exploit' New Deal
Image: Bloomberg AI
Main Takeaway
Amazon launched OpenAI models overnight while Microsoft CEO Satya Nadella publicly declared he'll "exploit" the same deal for Azure customers.
Jump to Key PointsSummary
Amazon didn't wait
Amazon flipped the switch before the ink dried. Less than 24 hours after Microsoft agreed to surrender its exclusive grip on OpenAI, AWS already had GPT-4o, o3, and a brand-new agent framework live in the console. No staged rollout. No beta list. Just a giant "Add to Cart" button for enterprise customers who've been begging for native OpenAI access since 2022.
The speed shocked even AWS veterans. TechCrunch reports that internal teams green-lit the launch at 3 a.m. Seattle time, hours after the amended Microsoft agreement hit the wires. The new agent service—basically an API wrapper that chains together multiple GPT calls—wasn't even in AWS's public roadmap last week. Now it's front-page billing material.
Microsoft's public pivot
Microsoft isn't sitting still. CEO Satya Nadella told TechCrunch in a brief hallway interview that he sees the amended deal as pure upside. "We fully plan to exploit it," he said, grinning. The comment marks a sharp turn from Microsoft's previous defensive posture around OpenAI exclusivity.
Sources familiar with Azure's roadmap say Nadella's words weren't empty. Teams inside Microsoft have already started migrating OpenAI endpoints to a new billing structure that lets them offer the same models to customers without paying OpenAI's standard API fees. One engineer described it as "getting the milk without buying the cow."
What landed overnight
AWS customers woke up to three concrete offerings: A managed GPT-4o endpoint priced at $0.002 per 1k tokens, undercutting Azure's rate by 8%. An o3 reasoning model with built-in guardrails that Amazon swears can handle PHI workloads (we'll see). And the sleeper hit: an agent orchestration layer that lets developers stitch together GPT calls, Lambda functions, and SageMaker models without writing glue code.
Amazon's press deck claims 2,300 companies requested access during the closed preview. Those companies are getting emails right now with direct console links. No wait-list. No sales call. Just click and deploy.
The new math
Every AWS customer who spins up OpenAI through Amazon is one less Azure bill. But Microsoft's new arrangement means they keep the 2032 model access guarantee while paying zero marginal cost for OpenAI's tech. Nadella's "exploit" comment suddenly makes perfect sense—they can match or beat AWS pricing while pocketing pure margin.
The race is on. AWS has first-mover advantage and a slick orchestration layer. Microsoft has cost-free access and an existing enterprise base that already runs their entire stack. Both sides just got handed the same weapons.
Key Points
AWS launched OpenAI models (GPT-4o, o3) plus new agent framework within 24 hours of Microsoft dropping exclusivity
Microsoft CEO Nadella publicly stated they'll "exploit" the new deal structure for Azure's benefit
AWS undercut Azure pricing by 8% on GPT-4o tokens at launch
2,300 companies had early access to AWS OpenAI services via closed preview
Microsoft now gets OpenAI tech at zero marginal cost under amended agreement
Questions Answered
April 28, 2026, less than 24 hours after Microsoft agreed to end exclusivity.
GPT-4o endpoints, o3 reasoning models, and a new agent orchestration service.
AWS GPT-4o pricing is $0.002 per 1k tokens, about 8% cheaper than Azure's previous rate.
Yes, Microsoft immediately matched AWS pricing and extended Azure startup credits through 2027.
Source Reliability
35% of sources are highly trusted · Avg reliability: 68
Go deeper with Organic Intel
Simple AI systems for your life, work, and business. Each one includes copyable prompts, guides, and downloadable resources.
Explore Systems
