OpenAI just ended its Azure exclusivity. GPT-5.5—their most capable agentic model—is now available on AWS Bedrock, giving enterprises a second production path for deploying frontier AI. For CTOs managing multi-cloud strategies and CFOs evaluating AI spend, this changes the procurement conversation.

The news: OpenAI launched GPT-5.5 this week alongside a partnership expansion with AWS. The model is available through Amazon Bedrock at $5 per million input tokens and $30 per million output tokens—double GPT-5.4's pricing, but delivering what OpenAI calls "state-of-the-art intelligence at half the cost of competitive frontier coding models" when measured by output quality per dollar.

Why CTOs Care: Multi-Cloud AI Deployment

For the past two years, production OpenAI deployments meant Azure. That created vendor lock-in challenges for organizations already standardized on AWS. Bedrock changes the equation.

What's now available on AWS:

• GPT-5.5 and other OpenAI models through Bedrock APIs
• Codex coding agent integrated into AWS development workflows (CLI, desktop, VS Code)
• Bedrock Managed Agents powered by OpenAI's agent framework
• AgentCore runtime for enterprise-scale agent deployment

The technical advantage: these run within your existing AWS environment. No separate infrastructure to configure. Model inference stays on Bedrock, execution follows OpenAI's harness, and everything inherits AWS security controls, encryption, and audit logging you've already implemented.

Ed Anderson, Distinguished VP Analyst at Gartner covering cloud and AI infrastructure, described the move as "significant for AWS customers, who now have access to the latest OpenAI models through Amazon Bedrock. Customers can expand their AI strategies to leverage the strengths of multi-cloud environments and select the best models for each use case."

The deployment decision: If you're already running production workloads on AWS with established governance frameworks, Bedrock lets you evaluate OpenAI models alongside Anthropic, Meta, Mistral, Cohere, and Amazon without fragmenting your infrastructure across cloud providers.

Why CFOs Care: Cost Structure and Procurement

GPT-5.5 lists at $5/$30 per million input/output tokens—exactly 2x GPT-5.4's $2.50/$15 pricing. That looks expensive until you dig into usage efficiency.

The cost analysis:

OpenAI claims GPT-5.5 "uses significantly fewer tokens to complete the same Codex tasks" compared to GPT-5.4. On Artificial Analysis's Coding Index, GPT-5.5 delivers "state-of-the-art intelligence at half the cost of competitive frontier coding models."

The math: if GPT-5.5 completes tasks in 40% fewer tokens (conservative estimate based on benchmark efficiency gains), effective cost per task approaches GPT-5.4 despite 2x per-token pricing. And if it replaces a competitive model at 2x the cost, you're saving 50% while upgrading capability.

Procurement advantage: Bedrock usage can be applied toward existing AWS commitments. If you've already committed $10M+ annually to AWS infrastructure, GPT-5.5 consumption draws from that pool—no separate OpenAI contract negotiation, no new vendor onboarding, simplified cost management.

For finance teams evaluating AI spend across departments, consolidating model access through Bedrock means one procurement relationship, one billing structure, one compliance review instead of managing contracts with OpenAI, Anthropic, and others separately.

The Benchmarks: What GPT-5.5 Actually Does

OpenAI positions GPT-5.5 as their strongest agentic coding model. Here's what that means in production terms:

Terminal-Bench 2.0: 82.7% accuracy (state-of-the-art)

• Tests complex command-line workflows requiring planning, iteration, and tool coordination
• Measures ability to complete multi-step tasks without human intervention

SWE-Bench Pro: 58.6% success rate

• Evaluates real-world GitHub issue resolution
• Solves more tasks end-to-end in a single pass than previous models

Expert-SWE: Outperforms GPT-5.4

• OpenAI's internal frontier eval for long-horizon coding tasks
• Median estimated human completion time: 20 hours per task

Real-world context: OpenAI reports that 85% of the company uses Codex every week across software engineering, finance, communications, marketing, data science, and product management. Senior engineers testing the model said GPT-5.5 was "noticeably stronger than GPT-5.4 and Claude Opus 4.7 at reasoning and autonomy."

One engineer at NVIDIA who had early access went as far as to say: "Losing access to GPT-5.5 feels like I've had a limb amputated."

Michael Truell, Co-founder & CEO at Cursor, noted: "GPT-5.5 is noticeably smarter and more persistent than GPT-5.4, with stronger coding performance and more reliable tool use. It stays on task for significantly longer without stopping early, which matters most for the complex, long-running work our users delegate to Cursor."

What This Means for Enterprise AI Strategy

The OpenAI-AWS partnership reshapes how enterprises can approach AI deployment:

For multi-cloud organizations:

• Evaluate OpenAI models alongside alternatives without fragmenting infrastructure
• Use existing AWS security/governance frameworks instead of building separate controls
• Consolidate AI spend under AWS commitments instead of managing multiple vendor contracts

For Azure-standardized organizations:

• Still no change—Azure remains a production path
• But AWS customers are no longer forced into Azure for OpenAI access

For CFOs managing AI budgets:

• 2x per-token pricing vs. GPT-5.4, but potentially lower cost-per-task due to efficiency
• Half the cost of competitive frontier coding models (per Artificial Analysis)
• Simplified procurement through AWS commitments

For CTOs evaluating agentic AI:

• Strongest publicly available benchmarks for long-horizon coding tasks
• Production-ready deployment with Bedrock Managed Agents
• Real internal adoption signal (85% of OpenAI employees use Codex weekly)

The Competitive Landscape

OpenAI's move to AWS Bedrock follows a reset of its Microsoft partnership that ended Azure exclusivity. This creates a three-way race in enterprise AI infrastructure:

Azure OpenAI Service: Established, production-proven, deep Microsoft integration AWS Bedrock: Multi-model platform, existing AWS governance, consolidated billing Google Vertex AI: Gemini models, GCP integration, enterprise security

For enterprise leaders, the strategic question shifts from "How do we get OpenAI access?" to "Which cloud platform aligns with our broader AI strategy?"

What to Watch

Pricing evolution: GPT-5.5 at $5/$30 is the launch price. Watch for Batch API pricing (typically 50% discount), volume discounts for enterprise contracts, and competitive pressure as Anthropic, Google, and others respond.

Model performance: OpenAI claims GPT-5.5 uses "significantly fewer tokens" to complete tasks. Track your own usage data—if you're not seeing 30-40% token reduction vs. GPT-5.4, effective cost-per-task may not justify the 2x pricing.

AWS commitment utilization: If you're under-utilizing AWS commitments, Bedrock consumption helps avoid waste. If you're already at capacity, GPT-5.5 usage may trigger overages.

Agent deployment complexity: Bedrock Managed Agents promise production-ready deployment, but early adopters should expect integration work connecting agents to enterprise data sources, tools, and workflows.

Multi-cloud governance: Running OpenAI models on both Azure and AWS creates governance complexity—which teams use which platform? How do you maintain consistent security controls across environments?

The Bottom Line

OpenAI's AWS Bedrock launch gives enterprises a second production path for deploying frontier AI. For organizations already standardized on AWS, it eliminates the Azure-or-nothing choice. For multi-cloud shops, it enables model evaluation within existing infrastructure.

The cost equation is nuanced: 2x per-token pricing vs. GPT-5.4, but potentially lower cost-per-task if efficiency gains materialize. CFOs should track usage data, not just list prices.

The strategic opportunity: consolidate AI model access through a single cloud platform, simplify procurement, and evaluate OpenAI alongside alternatives without fragmenting governance.

Decision framework:

• If AWS-standardized: Bedrock is the obvious path for OpenAI access
• If Azure-standardized: No change unless multi-cloud strategy drives exploration
• If multi-cloud: Evaluate cost, governance complexity, and commitment utilization before choosing platform
• If evaluating agentic coding: GPT-5.5 benchmarks (82.7% Terminal-Bench, 58.6% SWE-Bench Pro) are strongest publicly available

OpenAI ending Azure exclusivity doesn't change the core question: does your organization need frontier AI capability, and can you deploy it responsibly? But it does change the procurement answer—you now have options.

Want to calculate your own AI ROI? Try our AI ROI Calculator — takes 60 seconds and shows projected savings, payback period, and 3-year ROI.

Continue Reading

If you're evaluating enterprise AI strategy, you might find these articles helpful:

• Enterprise AI Governance: Building Controls That Scale
• Multi-Cloud AI Strategy: When Does It Make Sense?
• AI Model TCO: Hidden Costs Beyond Per-Token Pricing

Sources

• OpenAI Release Notes - May 2026
• Introducing GPT-5.5 | OpenAI
• OpenAI models, Codex, and Managed Agents come to AWS | OpenAI
• Amazon Bedrock Managed Agents | AWS
• AWS and OpenAI Partnership Brings New AI Capabilities to the Amazon Bedrock Platform | National CIO Review
• GPT-5.5 Pricing & Benchmarks | Artificial Analysis
• GPT-5.5 vs GPT-5.4: Pricing, Speed, Context, Benchmarks | LLM Stats