What is OpenAI on AWS? OpenAI on AWS is the general availability of OpenAI's frontier models, GPT-5.5 and GPT-5.4, together with the Codex coding agent and Amazon Bedrock Managed Agents, inside Amazon Web Services. As of June 1, 2026, commercial real estate firms that already run their data and applications on AWS can call OpenAI models on AWS through Amazon Bedrock without leaving their existing cloud, governance, and billing perimeter. That matters because OpenAI's strongest models were previously tied to Microsoft Azure for most enterprises. For a wider view of the platforms reshaping the industry, see our guide to the best AI tools for commercial real estate.
Key Takeaways
- GPT-5.5, GPT-5.4, and Codex reached general availability on Amazon Bedrock on June 1, 2026, ending the period when OpenAI's top models were effectively Azure-only for most enterprises.
- CRE firms standardized on AWS can now run OpenAI models alongside Anthropic Claude and Meta Llama, choosing the best model per task rather than per cloud.
- Bedrock inherits AWS controls such as IAM, VPC, KMS, and CloudTrail, creating audit trails that matter for high-risk uses like tenant screening, valuation, and lending.
- Pricing matches OpenAI first-party rates, is billed per token with no seat licenses, and counts toward a firm's existing AWS spend commitments.
- OpenAI prompts and responses on Bedrock are not used to train models, addressing a top data-governance concern for firms handling rent rolls and investor data.
OpenAI Models on AWS Explained
The release bundles three offerings on Amazon Bedrock. First, the frontier models: GPT-5.5, positioned as OpenAI's most capable model for agentic coding, data analysis, and multi-step tasks, plus the faster and cheaper GPT-5.4. Second, Codex, OpenAI's coding agent, which holds context across an entire repository and runs through the Codex CLI and IDE integrations such as Visual Studio Code and JetBrains. Third, Bedrock Managed Agents powered by OpenAI, which pair OpenAI's agentic capabilities with AWS infrastructure to build and run long-running agents.
General availability arrived on June 1, 2026, about a month after AWS and OpenAI announced an expanded partnership. GPT-5.5 is initially available in the US East (Ohio) region, while GPT-5.4 runs in US East (Ohio) and US West (Oregon) to support data residency. Bedrock already hosts Anthropic Claude, Meta Llama, and Cohere, so adding OpenAI turns it into a true multi-model hub, a 2026 trend we covered when AWS added Grok to Bedrock. OpenAI says Daybreak, its cyber-defense models and Codex Security, will follow.
Why OpenAI Models on AWS Matter for CRE Investors
For years, the practical question for a CRE technology team was not just which AI model is best, but which model your cloud will let you run. A firm with its data lake and applications on AWS faced friction using OpenAI's strongest models, which lived on Azure. OpenAI models on AWS remove that friction. The need is real: roughly 92% of corporate occupiers have started AI programs, yet only about 5% report achieving most of their goals. The gap is rarely the model itself; it is integration and governance. With the AI in real estate market projected to reach $1.3 trillion by 2030 at a 33.9% compound annual growth rate, the firms that solve integration now will compound the advantage. For deal work, pairing these tools with a strong reasoning model like Claude Opus 4.8 for CRE underwriting gives teams genuine model choice.
Key Benefits of OpenAI Models on AWS for CRE
- Model choice without cloud migration: Use GPT-5.5 and GPT-5.4 in the AWS account you already run, with no second cloud and no moving of sensitive property data.
- Unified governance: Every call inherits IAM permissions, VPC and PrivateLink isolation, KMS encryption, and CloudTrail logging, so AI activity is tracked like any other AWS service.
- Predictable cost: Token-based pricing matches OpenAI first-party rates, needs no per-seat licenses, and draws down a firm's existing AWS commitments.
- Agentic workflows: Codex and Bedrock Managed Agents run multi-step tasks, from an internal underwriting tool to a due diligence pass across hundreds of documents.
- Data protection: Prompts and responses are not used to train models or shared with the provider, a baseline when inputs include rent rolls and investor data.
Real-World CRE Applications
Consider a multifamily team evaluating a 200-unit property priced at $40 million. The team can build a Bedrock agent that reads the trailing twelve months (T12) statement and rent roll, then computes the figures that drive the deal. Net operating income (NOI) is gross revenue minus operating expenses, and it excludes debt service, capital expenditures, and depreciation. If the agent confirms NOI of $2.4 million, the cap rate is NOI divided by purchase price, or 6.0%. A 50 basis point compression to a 5.5% cap rate, holding NOI constant, implies a value near $43.6 million, a difference the agent can flag in seconds.
The same workflow can stress test financing. Debt service coverage ratio (DSCR) is NOI divided by annual debt service, so $2.4 million against $1.8 million of debt service yields about 1.33x, above a typical 1.25x lender threshold. The agent can then model internal rate of return (IRR), the discount rate that sets the net present value of all projected cash flows to zero across the hold period. This does not replace an underwriter; it turns hours of spreadsheet work into a reviewable draft. The pattern extends to document review, which we cover in our guide to AI in commercial real estate due diligence. CRE investors wanting hands-on AI implementation support can reach out to Avi Hacker, J.D. at The AI Consulting Network.
Governance, Compliance, and Vendor Risk
Running AI on regulated decisions raises the stakes. The EU AI Act's general purpose AI obligations take effect August 2, 2026, and the law treats housing decisions, including tenant screening, as high-risk. In the United States, HUD Fair Housing Act guidance already addresses AI in tenant screening and advertising, and state rules keep shifting, with Colorado's AI Act framework scaled back and delayed to January 1, 2027. The common thread is documentation: regulators expect firms to show how an automated decision was reached and to offer notice and human review.
This is where running OpenAI models on AWS inside Bedrock helps. Because every call generates IAM and CloudTrail records, a firm can show who used which model, with what data, and when, supporting the impact assessments these rules require. For how leading labs formalize this, see our analysis of OpenAI's Frontier Governance Framework. Multi-model access also cuts vendor concentration risk: if one provider changes pricing or availability, a firm can shift workloads without rebuilding its stack. If you are ready to build governed AI workflows on your existing cloud, The AI Consulting Network specializes in exactly this.
How CRE Firms Should Approach OpenAI Models on AWS
Start narrow. Pick one high-volume, low-risk workflow such as lease abstraction or rent roll normalization, and prototype it with GPT-5.4 to control cost. Set guardrails first: scope IAM roles, enable CloudTrail logging, and keep production tenant data out of prompts until your data-handling policy is documented. Measure against a human baseline, then compare OpenAI, Claude, and Llama on the same Bedrock account so model selection is driven by evidence, not by your cloud choice. Our pillar on AI tools for real estate investors maps where these models fit. CBRE and JLL have built dedicated technology and data center practices because the plumbing now decides who captures AI's value. For guidance on choosing and governing AI models, connect with The AI Consulting Network.
Frequently Asked Questions
Q: What does OpenAI on AWS actually include?
A: It includes OpenAI's frontier models GPT-5.5 and GPT-5.4, the Codex coding agent, and Amazon Bedrock Managed Agents powered by OpenAI. All reached general availability on Amazon Bedrock on June 1, 2026, with native AWS governance and billing.
Q: How is pricing structured for OpenAI models on AWS?
A: Pricing matches OpenAI first-party rates and is billed per token, with no seat licenses or per-developer commitments. Usage counts toward your existing AWS spend commitments, which makes costs easier to forecast.
Q: Is my CRE data safe when using these models through Bedrock?
A: AWS states that prompts and responses are not used to train models and are not shared with the model provider. Calls also inherit IAM, VPC, KMS, and CloudTrail controls, so access is logged and encrypted.
Q: Should a CRE firm pick OpenAI, Claude, or another model?
A: There is no single right answer. Because Bedrock now hosts OpenAI, Anthropic Claude, Meta Llama, and others in one place, test each model on your real tasks, such as underwriting or lease abstraction, and route work to whichever performs best.
Q: Does this change anything for AI compliance in real estate?
A: It can help. The audit logging and access controls in Bedrock support the documentation that rules like the EU AI Act and HUD Fair Housing guidance expect for high-risk decisions such as tenant screening and valuation. The model location does not remove your compliance obligations, but it makes them easier to evidence.