Google Cloud Unifies GPT-6, Claude, and Gemini on Single Invoice for Enterprise
For the first time in the business AI era, companies can use one cloud bill to pay for all three major families of closed-source AI models: Google's Gemini, Anthropic's Claude, and OpenAI's GPT-6. Google Cloud made this possible by adding OpenAI's latest models to its Model Garden platform. This makes billing easier, but it also comes with a tradeoff. Putting all three AI model families on one cloud provider goes against the advice of most enterprise architecture guidelines, which recommend using multiple cloud providers to reduce risk.
Google Cloud Vice President Michael Gerstenhaber announced the news in an official blog post. He confirmed that three OpenAI models are entering private preview for selected business customers. GPT-Astra is designed for advanced software development, using computers, and scientific research. GPT-6.1 Sol is built for AI agents that write code and handle professional tasks at a lower cost. GPT-6 Luna is a fast, affordable option for everyday tasks that happen in large numbers. General availability is expected within weeks, and customers can already sign up. OpenAI's coding agent, Codex, is also expected to come to Google Cloud as part of the wider launch.
The way Google handles billing could be just as important as the models themselves. All GPT-6 use through Model Garden, along with existing Gemini and Claude use, appears on one bill. The charges count toward Committed Use Credits and use up customers' Flexible Savings Plans (FSP). Business purchasing teams that now manage separate API agreements with OpenAI or Microsoft Azure, Anthropic, and Google can reduce paperwork and make billing simpler by bringing these services together.
One Bill, Three AI Giants: How the Commercial Mechanics Work
Google has confirmed that customer data will not be used to train any of the models it hosts. Business customers now consider this a basic requirement. Google's cloud privacy promise says that customer prompts, background information, and model responses stay within each customer's designated data boundary. With GPT-6 added to its catalog, Google Cloud hosts more than 200 models from Google, other companies, and open-model projects. According to Google's official announcement, it is the only large-scale cloud provider to offer all three leading families of closed-source AI models in one place.
GPT-6 Model Lineup and Pricing: What Google Cloud Has and Hasn't Said
The three GPT-6 models offer different balances of price and performance. This lets companies choose a model that fits each task instead of using one model for everything. OpenAI's direct API prices provide a guide because Google Cloud has not yet published its own Model Garden prices for these models. According to the OpenAI model documentation page, GPT-Astra costs $10 per million input tokens and $50 per million output tokens. GPT-6.1 Sol costs $2 per million input tokens and $10 per million output tokens. GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens.
The 100-fold difference between Astra's and Luna's output prices matters when designing AI systems. It suggests that Astra is intended for tasks that need deep reasoning, such as writing complex code, carrying out multi-step scientific analysis, and running AI agents that work on their own. Luna is designed for tasks that happen often but do not need the highest level of accuracy, such as sorting content into categories or directing customer questions to the right place. Sol sits between the two. It is aimed at professional tasks that need reasoning close to Astra's level but cost less.
Business purchasing teams should pay attention to the pricing gap. Google Cloud has not announced its own Marketplace prices for GPT-6 as of this writing. Whether Google charges more than, less than, or the same as OpenAI's direct API prices will affect whether combining billing also saves money. Companies planning their AI budgets should treat the prices above as reference points, not as confirmed Google Cloud rates.
Inside Model Garden's Infrastructure: How Google Hosts a Competitor's Models
Companies thinking about using GPT-6 through Google Cloud instead of OpenAI's direct API should understand how the systems differ. Google Cloud's Model Garden, now part of the Gemini Enterprise Agent Platform, hosts third-party models such as GPT-6 through a service called Managed Model-as-a-Service (MaaS). The model runs on computing equipment managed by Google. Customers access it through an API endpoint, a connection point for software, that follows the same basic structure as other Model Garden models, according to the Model Garden architecture documentation.
Two features designed for business customers make Google Cloud's service different from using OpenAI's API directly:
Dedicated Serving Instances (DSI): Businesses that use AI heavily can request computing resources reserved only for their own requests. Their work does not have to compete for space in a shared queue with other customers, and nearby customers' activity should not cause the same changes in processing speed. DSI offers promised processing capacity and lower delays for AI tasks that must work reliably in production.
VPC Service Controls: Business security teams can set network boundaries so that all requests to GPT-6 stay within their chosen Google Cloud project. The requests do not need to travel over the public internet, according to the VPC Service Controls documentation. This is an important security difference from using OpenAI's API directly, where requests travel over public internet infrastructure.
The business reason Google is willing to host a competitor's models is simple: computing power can support many different AI products. Google earns money from the computing resources used to process GPT-6 API requests, no matter whose model runs on its equipment. Every token processed through Google Cloud uses computing resources that Google bills for. When a company sends its OpenAI API requests through Google Cloud, it also increases its spending with Google. This can make the company more dependent on Google, even if Google's own AI models offer no special advantage for that company's tasks.
Consolidation vs. Multi-Cloud: What Enterprise Architects Need to Know
The claim that Google Cloud now offers businesses more choice among AI models is true in one sense but can be misleading in another. OpenAI's GPT-6 models are now available through AWS, Azure, and Google Cloud, so companies can choose which cloud provider runs their OpenAI workloads. But Google's announcement also encourages businesses to move more of their total AI spending to one cloud provider.
A standard multi-cloud setup uses services from several cloud providers to avoid relying too heavily on any one company. It spreads work across providers rather than bringing everything together. A business that previously used OpenAI directly or through Azure, Anthropic, and Google had three separate ways to access AI models. Moving all three to Google Cloud's single bill reduces paperwork, but it also concentrates infrastructure risk in Google Cloud. If Google's systems slow down, fail, or change their contract terms, the company's access to all three AI model families could be affected at the same time.
This does not mean companies should reject the combined billing option. It means they should consider the tradeoff that the announcement does not discuss. Business technology leaders should compare the offer with their existing Google Cloud spending commitments. If a company already has large Google Cloud contracts, combining billing may provide real value without greatly increasing its dependence on Google. But if the company has deliberately spread its systems across AWS and Azure, the change may work against its goal of using multiple cloud providers.
How OpenAI Lost Its Azure Exclusivity — and Why That Made This Possible
OpenAI's arrival in Google Cloud's model catalog required years of contract changes before it became legally possible. For much of OpenAI's history, Microsoft Azure was its exclusive cloud partner. This relationship was supported by Microsoft's reported $13 billion investment in OpenAI.
That exclusive arrangement ended in stages. In February 2026, AWS and OpenAI announced a deal that made AWS the exclusive third-party cloud distribution provider for OpenAI Frontier, OpenAI's platform for AI agents. The deal was reportedly worth up to $50 billion. However, the exclusive rights applied to the Frontier platform, not to the underlying model API. This left room for other distribution agreements.
After Microsoft publicly challenged parts of the AWS deal, Microsoft and OpenAI renegotiated their agreement. Under revised terms announced in April 2026, the Microsoft-OpenAI renegotiation terms gave Microsoft a non-exclusive license to OpenAI's intellectual property through 2032. Microsoft also stopped making revenue-sharing payments on OpenAI products resold through Azure. OpenAI gained the freedom to sell its models through any cloud provider, according to reporting by Directions on Microsoft and a detailed analysis of the revised Microsoft-OpenAI agreement. Microsoft remains OpenAI's primary cloud partner, and OpenAI products are expected to launch on Azure first when possible. But today's announcement makes clear that first no longer means only.
OpenAI President and co-founder Greg Brockman described the Google Cloud partnership as "an important step in making our most capable models and Codex available where customers are already building," according to the Google Cloud announcement. Brockman co-founded OpenAI and, in 2025 and 2026, also co-founded "Leading the Future," a political action committee that supports reducing AI regulation. The group raised more than $100 million with backing from Andreessen Horowitz, according to TechCrunch's coverage of the PAC's launch. Brockman donated $25 million to MAGA Inc. in 2025 and $25 million to the Leading the Future PAC in 2026. These political activities offer a signal of how OpenAI's leadership expects rules for advanced AI to develop. TechTimes readers assessing OpenAI's approach to business governance should consider them as part of that assessment.
Internal Tension at Google: Serving Rivals While Building Gemini
Google Cloud's decision to host OpenAI's most capable models shows that the company believes its cloud market share and developer ecosystem are worth more than the competitive cost of helping a rival AI company reach more customers. The business logic is straightforward: a cloud provider benefits whenever an AI model runs on its systems, even if that model competes with the provider's own AI products.
However, there is a real conflict over resources. The Decoder noted in its report on the announcement that computing power used to run GPT-6 for business customers is not available for Google's own Gemini development and model processing. This competition for computing resources reportedly helped lead to the departure of Noam Shazeer. He co-wrote the 2017 paper "Attention Is All You Need," which introduced the transformer architecture that supports modern large language models. Shazeer returned to Google in 2024 after the company licensed his Character.AI technology for $2.7 billion. He left Google in June 2026 to join OpenAI.
For now, Google Cloud has decided that the benefits of bringing business purchases together outweigh the internal conflict. It wants to be the only large-scale cloud platform where a purchasing team can sign one contract and access GPT, Claude, and Gemini. Whether this approach will remain attractive as Google's Gemini models compete more directly with GPT-6 Astra-level capabilities will become clearer over the next few product cycles.
What This Means for Enterprise AI Buyers
The main point for business decision-makers is simple: if your company already relies on Google Cloud, combining billing may be a strong option. You can get three AI model families through one bill, one set of security and compliance controls, and one vendor relationship for model access. This can replace the need to manage separate API keys, usage dashboards, and contracts with OpenAI, Anthropic, and Google.
The benefits are smaller for companies that do not mainly use Google Cloud or have deliberately spread their AI systems across several providers. For these companies, starting a new Google Cloud relationship just to access GPT-6 through its billing system is unlikely to save money overall. Simpler billing may not be worth the cost of making a new commitment to a cloud provider.
GPT-6 models are now available in private preview through Google Cloud's Model Garden. Companies can sign up for early access through the Model Garden sign-up page. Wider availability is expected within weeks.
Frequently Asked Questions
What is the difference between GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna on Google Cloud?
Google Cloud's Model Garden offers three levels of OpenAI's GPT-6 models. GPT-Astra is the most capable and expensive option. It is designed for advanced software development, computer use, and scientific research. Based on OpenAI's direct API prices, it costs $10 per million input tokens and $50 per million output tokens. Google Cloud has not yet published its own prices. GPT-6.1 Sol aims to provide reasoning close to Astra's level for AI agents that write code and other professional tasks, at a lower price of $2 per million input tokens and $10 per million output tokens. GPT-6 Luna is designed for everyday tasks that happen in large numbers and need to be cost-effective. It costs $0.10 per million input tokens and $0.50 per million output tokens, according to the Google Cloud blog post describing the models. Astra's output tokens cost 100 times as much as Luna's. Companies should therefore check each type of task separately so they do not pay Astra prices for work that Luna can handle.
Does using OpenAI models through Google Cloud mean my data stays on Google's infrastructure?
Under Managed Model-as-a-Service, the system Google uses to host third-party proprietary models in Model Garden, requests to run the models are processed on computing equipment managed by Google. Google's privacy terms say customer data will not be used to train the hosted models. Business customers can also use VPC Service Controls to keep requests within the boundaries of their Google Cloud project, preventing them from traveling over the public internet. This security option is not available when using OpenAI's API directly. Companies with strict rules about where their data must be stored or processed should confirm which Google Cloud regions will support GPT-6 when it becomes generally available. They should also note that Zero Data Retention options and Confidential Compute are listed as coming soon, so they will not be available at launch.
If OpenAI models are now on Google Cloud, AWS, and Azure, is that a multi-cloud AI strategy for my organization?
Not necessarily. This difference matters to business technology leaders. OpenAI offering its models through several cloud providers means that its models can be accessed through multiple clouds. Whether your company has a multi-cloud strategy depends on where its systems actually run. If your team moves its OpenAI, Anthropic Claude, and Gemini requests to Google Cloud and uses its combined billing, your infrastructure is becoming more dependent on one cloud provider, not more spread out across several. The simpler purchasing process is a real benefit, but so is the risk of concentrating more work in one place. Companies that already use multiple clouds should decide whether easier billing is worth becoming more dependent on Google Cloud.
When will GPT-6 be generally available on Google Cloud, and how do I sign up?
General availability is expected within weeks of the private preview launch announced today. Companies can already sign up for early access through the Google Cloud Model Garden documentation and the Gemini Enterprise Agent Platform interface. Businesses with existing Committed Use Contracts should check with their Google Cloud account team to confirm that GPT-6 spending will count toward their existing commitment credits.