The AI Industry

OpenAI Models Arrive on Google Cloud: What the End of Microsoft Exclusivity Means for Enterprise Buyers

GPT-6 Astra, Sol and Luna are headed to Google Cloud in private preview. Here is what changes for pricing, procurement and the cloud race.

Tobias Reyes

AI Industry & Policy Analyst

Published 8 min read
System with various wires managing access to centralized resource of server in data center
In this story 10 sections

OpenAI will sell its frontier models, including GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna, through Google Cloud's Model Garden. Access starts as a private preview for select enterprise customers. The move follows the April end of OpenAI's exclusive distribution deal with Microsoft, and it gives enterprise buyers more leverage on price, terms and vendor choice.

This explainer is for technology leaders, procurement teams and anyone who buys AI through a cloud provider. It covers what was announced, what is still unclear, and how the deal changes bargaining power. Emergent Wire covers the AI industry, and this is one of the clearest signals this year that the labs and the clouds are rearranging who sells what to whom.

What Did OpenAI and Google Announce?

Three mature professionals in a business meeting discussing and signing documents in an office setting.

OpenAI said it will offer its frontier models and its Codex coding agent through Google Cloud in the coming weeks. Reuters reported the deal on October 21, noting that the models include GPT-6 Astra, which will appear in Google Cloud's Model Garden. Customers can buy through Google Cloud Marketplace and apply existing cloud spending commitments to those purchases, according to the Reuters report.

Techzine adds detail on the lineup. Astra targets demanding work such as coding, computer use and scientific research. Sol is close to Astra in capability but costs less, and Luna is a fast, low-cost model for high-volume everyday tasks. All three are in private preview for select enterprise customers only.

Google says it is now the only hyperscaler offering the three most powerful closed-source models, which it names as OpenAI's lineup, Anthropic's Claude Fable 5.1 and its own Gemini 4 Argon, according to Techzine. Gemini remains exclusive to Google Cloud.

Why Does the End of Microsoft Exclusivity Matter?

For years, OpenAI's models were sold on only one major cloud. OpenAI grew up on Microsoft Azure, where its models were exclusive because of Microsoft's earlier investment. That arrangement expired in April, and OpenAI has since expanded to Amazon Web Services and now Google Cloud.

The change matters because cloud platforms are where most large companies buy AI. A buyer that already runs workloads on Google Cloud can now reach OpenAI models without a new vendor contract. That lowers friction, and friction is a real cost in enterprise purchasing.

Hyperscalers are also becoming the main commercial channel for frontier labs, the Reuters report says. It notes that Anthropic drew 47% of its 2025 revenue from Amazon and Google's cloud platforms, according to its IPO filing. We covered the financial side in our look at OpenAI's $30 billion raise and Anthropic's IPO.

How Does Buying Through a Cloud Marketplace Change the Deal?

Close-up of a handshake with financial graphs on laptop screen, symbolizing a successful agreement.

It changes billing first. Techzine reports that usage appears on one invoice and counts toward existing cloud commitments and Flexible Savings Plans. For a company that has promised Google a large annual spend, that can make OpenAI models cheaper to adopt, since the spend draws down a commitment already made.

It also changes who has leverage. When a buyer can compare similar models inside one cloud, a salesperson has less room to hold price. Emergent Wire's analysts expect more buyers to run side-by-side tests of Astra, Sol, Luna, Claude and Gemini on the same platform, and to push for price matches.

The table below compares what changes for three types of buyers.

How Does Buying Through a Cloud Marketplace Change the Deal?
BuyerWhat changesWhat to ask for
Google Cloud customerOne platform, many frontier modelsCommit drawdown, unified billing
Azure customerAlternative to a single supplierPrice matching, revised terms
Regulated enterpriseAccess through existing governance toolsData retention and residency terms
StartupCredits and marketplace spendStartup program eligibility

What Do Buyers Still Not Know?

Quite a lot. The Reuters report gives no financial terms for the Google deal, the Microsoft exclusivity end or the AWS arrangement. It also gives no launch date beyond the coming weeks. Private preview means most companies cannot use the models yet.

Pricing is the biggest open question. Cloud providers typically keep a share of revenue from sales made through their platforms, the Reuters report notes. Whether Google's Marketplace prices match OpenAI's direct rates is not confirmed in the coverage we reviewed.

Data controls are another gap. Google says prompts, context and outputs stay within the customer's data perimeter and are not used for training. Techzine says Zero Data Retention and Confidential Compute are planned to follow but need more work. Regulated buyers should not assume those protections exist today.

How Does This Compare With the Cost Picture?

Business person handling paper receipts, meticulously organized.

Pricing for AI inference has been falling, and a marketplace listing can push it further. Our guide to what a million tokens costs in 2026 shows how wide the price gaps are between model tiers. Sol and Luna are positioned to give buyers cheaper options than Astra.

Google's Michael Gerstenhaber, a vice president of product management, said workloads should be matched to the right model, rather than relying on a single model for every task. That is a sensible rule. A customer service bot does not need a model built for scientific research.

The savings come from routing. A company that sends 80% of requests to a cheaper model and 20% to a premium one can cut its bill sharply. Marketplaces make that kind of routing easier, since all the models sit under one contract.

Is Google Giving Up on Its Own Models?

No. Gemini remains exclusive to Google Cloud, and Google keeps promoting its own models. Reuters says Google Cloud CEO Thomas Kurian described the platform as open and said it welcomes any model popular with developers.

The strategy is to be the neutral shelf. If customers want choice, Google would rather sell them OpenAI models on its cloud than lose the workload to a rival. It also keeps customers inside its security, billing and data tools.

That fits a pattern of rivals cooperating when the economics demand it. Google and OpenAI compete in core AI products and search, but the cost of building and deploying large models pushes them toward shared infrastructure. The compute side is part of that story, which we explored in our piece on why AI labs want to build their own chips.

How Does This Change the Cloud Race?

The cloud race used to hinge partly on who had exclusive access to the best model. That edge has faded. Microsoft's exclusive window ended in April, Amazon followed, and Google has now joined. The contest shifts toward price, tooling, security and how well a cloud runs the model rather than whether it can offer it.

For Microsoft, the loss of exclusivity is a business change more than a technical one. Reuters reports that OpenAI has expanded distribution since the expiry, and an analyst note cited by other outlets says the open questions are how exclusivity terms get revised and how inference revenue is shared. Those details are not public.

For Google, the deal is a bet on neutrality. It sells its own Gemini models, hosts Anthropic's Claude and now offers OpenAI. A customer who wants to compare all three has a reason to stay inside one cloud, and the cloud earns a share of the spending either way.

For smaller clouds and model hosts, the picture is harder. If the biggest providers each carry every leading model, the differences move to price, speed and data handling. Specialist hosts will need to compete on those terms.

Does Model Choice Still Matter for Buyers?

It matters more, because choice is now cheap to exercise. When models sit side by side under one contract, a team can test three of them on the same task in a day. Differences in accuracy, speed and cost then show up in numbers instead of sales slides.

Buyers should resist the urge to pick one winner. A single model for every task usually overpays on simple jobs and underperforms on hard ones. Routing by task, as Google's Gerstenhaber suggested, is the cleaner approach, and a marketplace with several models makes it simple to try.

What Should Enterprise Buyers Do Now?

Close-up of a hand writing in a notebook with a checklist for effective task management.

Start by asking your cloud account team whether you qualify for the preview. Then ask for written answers on pricing, data retention and regional availability. Early access often comes with better terms for design partners.

Work through this short checklist as the preview opens.

  1. List your current AI workloads and the models behind them.
  2. Ask which workloads could move to a cheaper tier like Sol or Luna.
  3. Request marketplace pricing and compare it with direct rates.
  4. Confirm data retention, residency and audit terms in writing.
  5. Run side-by-side quality tests before moving production traffic.

Avoid a rush to switch. A new platform means new integration work, and quality can differ between providers even for the same model name. Test on your own data, not on a vendor demo.

A fair counterpoint is that most companies will not move quickly. Existing integrations, security reviews and procurement cycles slow change. The deal widens options, but a switch still takes months in a large organization.

Conclusion

OpenAI on Google Cloud ends the era of a single-cloud model for frontier AI. Buyers gain choice, simpler billing and negotiating room, while open questions remain on pricing and data controls. Treat the preview as a chance to test, compare and ask for terms in writing. Then decide which workloads belong on which model, and let price and quality set the answer.

Which OpenAI models are coming to Google Cloud?
Techzine reports that GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna will be available through Google Cloud Model Garden. All three start in private preview for select enterprise customers. Astra targets demanding work, Sol offers near-Astra capability at lower cost, and Luna is a fast low-cost model for high-volume tasks.
Is OpenAI still exclusive to Microsoft Azure?
No. Reuters reports that OpenAI exclusive distribution arrangement with Microsoft expired in April. OpenAI has since expanded to Amazon Web Services and now Google Cloud. Microsoft remains a major partner, but enterprise buyers can reach OpenAI models on other clouds, which gives them more leverage on price and terms.
Can I use my Google Cloud commitment to buy OpenAI models?
Yes, according to the announcement coverage. Customers can buy the models through Google Cloud Marketplace and apply existing cloud spending commitments to those purchases. Techzine adds that usage appears on one invoice and counts toward existing commitments and Flexible Savings Plans, which can simplify procurement for current Google Cloud customers.
Are OpenAI models on Google Cloud safe for regulated data?
Google says prompts, context and outputs stay within the customer data perimeter and are not used to train models, with queries passing through isolated gateways. Techzine notes that Zero Data Retention and Confidential Compute are planned but not yet available, so regulated buyers should get retention and residency terms in writing.

Written by

Tobias Reyes

AI Industry & Policy Analyst

12+ years experience

Tobias Reyes spent ten years as a technology strategy consultant, most of that time advising Fortune 500 clients on enterprise software adoption before AI became the whole conversation. He joined Emergent Wire to cover the industry from the incentive layer up: who actually pays for frontier model access, how compute costs shape product decisions, and where policy is likely to reshape the competitive landscape next. He is allergic to roadmap-slide analysis and prefers to start from unit economics and public filings whenever they exist.

Credentials

  • Former technology strategy consultant (10 years, enterprise clients)
  • MBA, focus on technology and operations strategy
  • Regular commentator on AI industry economics and policy