Это русская версия материала. Для полноты языковых маршрутов текст основан на существующей основной версии.
On April 29, 2026, OpenAI officially opened three services on Amazon Bedrock: GPT-5.5 model, Codex Agent, and the new Bedrock Managed Agents. This is OpenAI's first major multi-cloud deployment after ending its exclusivity agreement with Microsoft — a watershed moment for enterprise AI infrastructure.
What Happened
Exclusive Partnership Relaxed
OpenAI and Microsoft reset their partnership: exclusivity terms relaxed, OpenAI can now distribute models to multiple cloud providers while core collaboration remains intact.
AWS Bedrock Launch Details
| Service | Content |
|---|---|
| GPT-5.5 | OpenAI's latest flagship, multiple reasoning strength configs |
| Codex Agent | Coding Agent with autonomous code generation, review, iteration |
| Bedrock Managed Agents | New managed Agent service, enterprises deploy and orchestrate AI Agents directly on AWS |
Why This Is Structural Change
From Azure Exclusivity to Multi-Cloud Distribution
Previously, Azure was the only cloud hosting OpenAI models for enterprise use. Now AWS becomes the second enterprise-grade path:
- AWS-invested enterprises no longer need Azure accounts and compliance processes for GPT
- Multi-cloud strategy shifts from "optional" to "default"
- Vendor lock-in risk significantly reduced
Bedrock Managed Agents: AWS's Counter
Notably, Bedrock didn't just "add a GPT model" — it launched Managed Agents the same day:
| Platform | Agent Service | Launch |
|---|---|---|
| AWS Bedrock | Managed Agents | 2026.4.29 |
| Microsoft Agent 365 | Enterprise Agent governance | 2026.5.1 GA |
| Google Vertex AI | Agent Builder | Existing |
Impact on Industry
1. Azure AI's Moat Gets Shallower
Azure OpenAI Service's core value was "exclusive access to the most advanced models." When GPT-5.5 is also available on AWS, that proposition weakens significantly.
2. Enterprise Procurement Logic Changes
Past: Choose Azure → Get GPT + Azure ecosystem
Now: Choose AWS/Azure/GCP → All can get GPT → Decision returns to infrastructure itself
AI models become "commoditized" infrastructure layer. Competition returns to:
- Inference latency and throughput
- Compliance and data sovereignty
- Integration depth with existing tools
- Pricing
3. OpenAI's Independence Increases
Through DeployCo ($500M initial commitment) and multi-cloud distribution, OpenAI is building commercial infrastructure independent of any single cloud vendor.
Pricing Comparison
| Provider | GPT-5.5 Input | GPT-5.5 Output | Added Value |
|---|---|---|---|
| OpenAI Direct | $2.50/M | $15.00/M | Latest features |
| Azure OpenAI | $2.50/M | $15.00/M | Azure integration, compliance |
| AWS Bedrock | $2.50/M | $15.00/M | Bedrock Agents, AWS integration |
Model pricing remains consistent — differentiation shifts to added services and integration.
Action Advice
| Your Situation | Recommendation |
|---|---|
| Already using OpenAI on Azure | No need to migrate short-term, but evaluate AWS Bedrock as backup |
| In AWS ecosystem wanting GPT | Best time now — Bedrock native integration + Managed Agents |
| Multi-cloud planners | Leverage GPT availability across clouds, build model routing layer |
| AI vendor evaluation | End of exclusivity means reduced lock-in risk, increased negotiation leverage |