The OpenAI–Google Cloud partnership marks a surprising shift in how AI labs choose their allies
Sundar Pichai’s enthusiasm for supporting OpenAI signals how far Google will go to win cloud deals
Even with Microsoft’s billions behind it, OpenAI is hedging its bets by embracing Google’s infrastructure
It’s a partnership that would’ve seemed unthinkable a year ago. OpenAI, the Microsoft-backed research lab behind ChatGPT, has quietly started using Google Cloud to train and serve its AI models. In doing so, it has joined the ranks of Anthropic and Mistral as major AI players relying on Google’s infrastructure, even as they compete with Google in the very products that are defining the AI era.
SurveyGoogle CEO Sundar Pichai confirmed the deal during Alphabet’s Q2 earnings call this week, sounding not just satisfied but energized. “We’re very excited to be supporting OpenAI’s training and inference,” he said, emphasizing Google Cloud’s growing role as a home for next-generation AI workloads.
The takeaway is clear: in the world of AI, yesterday’s rivals might just be tomorrow’s best customers.
Also read: Google’s AI strategy is working, says Sundar Pichai as Alphabet reports strong Q2 results
From exclusive to expansive
For years, OpenAI’s relationship with Microsoft Azure looked exclusive. Azure powered GPT-3, GPT-4, and the foundation of ChatGPT’s API and enterprise integrations. Microsoft invested over $13 billion and became OpenAI’s deployment partner across everything from GitHub Copilot to Bing Chat.
But training frontier AI models isn’t just expensive, it’s logistically brutal. As demand for high-performance compute outpaces supply, even OpenAI needs a backup plan. Google Cloud offers precisely that: global-scale infrastructure, cutting-edge TPUs, and an AI-optimised stack that rivals anything Azure can provide.
Also read: OpenAI partners with Oracle to scale up Stargate AI data centers: All you need to know
This isn’t betrayal, it’s diversification. And it’s becoming the norm in a world where the bottleneck isn’t talent or ambition, it’s access to chipsets and data centers.
The “cloud wars”
What makes this partnership so fascinating is the tension it creates. OpenAI’s ChatGPT is one of the biggest threats to Google Search. Meanwhile, Google’s Gemini models are marketed directly against ChatGPT in nearly every segment, from consumer Q&A to enterprise copilots.
And yet, in the trenches of compute infrastructure, they are now allies, however cautiously. OpenAI gets redundancy and scale. Google gets prestige, revenue, and a powerful signal that even its most formidable rivals trust its cloud.
This is the strange logic of modern AI: you can be enemies in product and partners in hardware. You can undercut someone’s business on the front end, while renting their servers on the back end.
The backdrop to this deal is Google’s surging cloud momentum. In Q2 2025, Google Cloud revenue rose 32% year-over-year to $13.6 billion. That alone would be impressive. But even more striking is Google’s $10 billion increase in capital expenditures, pushing this year’s total to $85 billion, most of it aimed at building and expanding AI infrastructure.
That level of investment isn’t just for Google’s own Gemini ambitions. It’s for becoming the default compute layer for the broader AI ecosystem, even if that includes companies building products designed to dethrone Google itself.
In that sense, Google Cloud isn’t picking sides. It’s betting that no matter who wins the AI arms race, they’ll need somewhere to run their models. And Google intends to be that somewhere.
Frenemies forever?
So, what does this all mean for OpenAI and Google? Are they rivals playing nice? Or pragmatic collaborators doing what the moment demands?
The truth is more complicated and more interesting. These aren’t traditional alliances. They’re tactical arrangements born of necessity and scale. OpenAI doesn’t want to be overly dependent on Azure. Google doesn’t want to miss out on powering the very future it’s trying to build.
In this new landscape, frenemy status isn’t awkward, it’s efficient.
AI’s future will be shaped not just by model performance, but by infrastructure capacity. And in that fight, there’s little room for purity. There’s only performance, cost, and availability.
So yes, OpenAI is now running on Google Cloud. And yes, Google is still trying to outbuild OpenAI. But in the cloud wars, every GPU counts, even if it’s helping a rival sharpen their edge.
Also read: AI Factories to Agentic Web: NVIDIA and Microsoft’s vision for Future of AI
Follow Us
Vyom Ramani
A journalist with a soft spot for tech, games, and things that go beep. While waiting for a delayed metro or rebooting his brain, you’ll find him solving Rubik’s Cubes, bingeing F1, or hunting for the next great snack. View Full Profile
