AI Frontiers: OpenAI Just Moved Into Google’s House

There’s a sentence that would have sounded absurd two years ago: OpenAI’s models are now available inside Google Cloud.

Not through some unofficial workaround. Not as a side experiment. Through Google Cloud’s Model Garden, on Google’s Marketplace, with Google Cloud spend commitments counting toward the purchase. And OpenAI’s Codex agentic workflows are expected inside the Gemini Agent itself — Google’s own AI product.

The two biggest names in AI just became distribution partners.

What the Deal Actually Is

OpenAI’s frontier models are joining Google Cloud’s Model Garden — the same catalog where Google offers its own Gemini models, alongside third-party options. Enterprise customers can access and deploy OpenAI models without leaving the Google Cloud environment they’re already operating in.

The commercial angle matters: customers with existing Google Cloud spend commitments can apply those toward OpenAI model purchases. For large enterprises already locked into multi-year Google Cloud contracts, this removes the friction of spinning up a separate OpenAI billing relationship. It’s a meaningful unlock for procurement teams in regulated industries who need everything to flow through approved vendors and centralized governance.

General availability is coming in the next few weeks. Access is currently request-based.

Why This Is Stranger Than It Looks

Google and OpenAI are direct competitors. Gemini competes with GPT-4o. Google DeepMind competes with OpenAI Research. Vertex AI competes with OpenAI’s API. These aren’t adjacent companies — they’re fighting for the same enterprise contracts, the same developer mindshare, the same foundation model dominance.

And yet here is Google Cloud’s CEO framing the goal as giving customers “a broad selection to improve quality and lower costs” — which is true, and also happens to mean selling a competitor’s product through your own platform.

Microsoft remains OpenAI’s primary cloud partner. This doesn’t change that. But it does mean OpenAI is now distributed through all three major clouds — Azure natively, AWS via Bedrock, and now Google Cloud via Model Garden. The model itself has become infrastructure, and the clouds have become shelves.

What This Signals About Where the Market Is Going

A few years ago, the assumption was that each major cloud would build or exclusively back one foundation model — Microsoft with OpenAI, Google with Gemini, Amazon with Anthropic and its own models. Vertical integration. Walled gardens.

That assumption is quietly collapsing.

Google Cloud now sells Claude (via the Anthropic partnership), Llama (via Meta), and now GPT-4o and OpenAI’s agent stack. The model layer is becoming a commodity shelf, and the value is shifting to infrastructure, governance, compute, and integration — things the clouds actually own.

For enterprises, this is genuinely good news. Running GPT-4o for one task and Gemini for another, billed through the same cloud account, with the same access controls and audit logs, is a real operational improvement. Best model for the job, without the integration overhead.

For the model providers, it’s more complicated. Broader distribution means more revenue, but it also means the cloud platforms — not the model makers — become the primary customer relationship. The cloud controls the billing, the governance, the data residency. The model is just a thing you call.

Final Thought

When your fiercest competitor becomes your biggest distribution channel, the market has shifted under your feet.

OpenAI didn’t move into Google’s house because they’re friends. They moved in because the model wars are giving way to something less dramatic and more durable: an industry where AI models are utilities, clouds are the grid, and the real competition is in what you build on top.

The shovel sellers won the gold rush. The clouds might be about to do the same thing to AI.