Amazon Web Services said on June 1 that OpenAI’s frontier models and its Codex coding agent are now generally available on Bedrock, the cloud’s managed platform for foundation models. The move, one month after the two companies restructured their relationship in a limited preview, means any AWS customer can call GPT-5.5, GPT-5.4 and Codex through the same APIs and governance controls they already use across the cloud. Pricing matches OpenAI’s direct rates, and usage counts toward existing AWS spending commitments.
The availability is the concrete result of an agreement signed in April. OpenAI and AWS announced an expanded partnership on April 28 that ended the exclusive cloud arrangement OpenAI had maintained with Microsoft, and the launch arrived in the middle of a week in which Microsoft’s own Build conference was taking place. AWS vice president Swami Sivasubramanian confirmed that GPT-5.5 is “available natively on Amazon Bedrock,” alongside the full lineup of OpenAI models.
The geography of the rollout shows the caution involved. GPT-5.5, OpenAI’s most capable frontier model, is available in the US East (Ohio) region only. GPT-5.4, positioned as the price-performance option for most enterprise workloads, runs in Ohio and US West (Oregon), and will reach AWS GovCloud on June 3, making it the first OpenAI model available in a government cloud environment. Codex, the coding agent used by more than five million developers weekly, runs through its CLI, desktop app and integrations with Visual Studio Code, JetBrains and Xcode, with all inference routed through Bedrock.
For enterprise customers, the significance is operational rather than technical. Until now, an organization that wanted OpenAI models and used AWS as its primary cloud faced a choice: manage a split-cloud architecture to reach Azure, or do without. Bedrock removes that friction. Every API call inherits the security posture the customer already uses on AWS, including IAM roles, virtual private cloud isolation, encryption and CloudTrail logging, and usage flows through existing billing and commitment agreements. The governance integration, more than the model quality, is what teams evaluating the launch have cited as the reason to adopt it.
The competitive logic on both sides is explicit. For AWS, adding OpenAI completes a catalog that already includes Anthropic, Meta, Mistral and Amazon’s own models, and it answers a question enterprise buyers have asked for two years: why can’t we get the models we want on the cloud we already run? For OpenAI, Bedrock distribution expands reach into the largest enterprise computing base on earth, through a channel that does not require customers to build a new relationship with the company. OpenAI described the launch as the start of a broader path for customers to bring frontier AI into the environments where they already build, govern and ship.
The arrangement reshapes the economics of model distribution. Under the old regime, Microsoft’s exclusive licensing prevented OpenAI from offering API-accessible products through competing clouds, and AWS customers who wanted GPT had to route through Azure. The April restructuring carved out a path for multi-cloud availability, and the general availability launch confirms that the carve-out is real. Microsoft retains its investment in OpenAI and its own deep integration with the company, but it no longer holds the exclusive key to OpenAI’s enterprise distribution.
The launch also intensifies competition among the model providers themselves. Bedrock now hosts Anthropic’s Claude, backed by a multibillion-dollar AWS investment, and OpenAI’s GPT models on the same managed infrastructure. Customers can compare them in the same console, with the same governance, and switch based on workload fit rather than cloud allegiance. That is a structural change: the era of exclusive cloud partnerships for frontier models is ending, replaced by a market where the models are commodities and the cloud is the channel.
The reaction on Hacker News, where the announcement drew a thread with more than 300 points, focused less on the news itself than on what it means for the companies involved. Some commenters noted that Anthropic’s revenue surge owed something to the absence of OpenAI on AWS, and that the balance would shift as GPT becomes a native option on the platform. Whether that reading is fair, the underlying point is sound: distribution determines revenue in enterprise AI, and OpenAI has just doubled its distribution.
For AWS, the launch is also a defense of Bedrock’s position. The platform has competed with Azure AI and Google’s Vertex by offering the widest catalog of models under one governance umbrella, and losing OpenAI would have been a permanent gap. Now the catalog is complete, and the argument for Bedrock is simple: every major model family, one platform, the controls you already trust.
The remaining question is price competition. OpenAI’s models on Bedrock cost the same as they do directly, which preserves margins but invites comparison with cheaper rivals hosted on the same platform. AWS is not in the business of subsidizing model prices; it profits from usage. The bet is that customers choose Bedrock for governance and stick with it for scale, whatever model wins the benchmark race. That is a bet on the cloud itself, and it just got bigger.
