In a surprising development that reshapes competitive dynamics in the AI sector, OpenAI has signed an agreement to use Google Cloud's computing power for training and running its AI models, despite Google being one of its fiercest rivals in the artificial intelligence space.
The deal, finalized in May after months of negotiations, represents OpenAI's latest effort to diversify its compute sources beyond Microsoft Azure. Until January 2025, Microsoft had been OpenAI's exclusive cloud provider since their partnership began in 2019. However, that exclusivity ended when Microsoft's role shifted to having a "right of first refusal" on new capacity, allowing OpenAI to seek additional providers when necessary.
This partnership comes at a critical time for OpenAI, which recently announced its annualized revenue has reached $10 billion as of June 2025, nearly doubling from $5.5 billion in December 2024. The company is targeting $12.7 billion in revenue for 2025 and an ambitious $125 billion by 2029. However, these growth projections require massive computing infrastructure - OpenAI reportedly lost about $5 billion last year primarily due to infrastructure costs.
For Google Cloud, securing OpenAI as a customer represents a significant win. Google will supply additional computing capacity to OpenAI's existing infrastructure, likely leveraging its advanced Tensor Processing Units (TPUs) that were once reserved for internal use. Google Cloud generated $43 billion in sales in 2024, comprising 12% of Alphabet's revenue, and has positioned itself as a neutral provider of computing resources for AI startups.
The collaboration underscores how the enormous computing demands of modern AI are reshaping competitive relationships in the tech industry. Despite ChatGPT posing a significant threat to Google's search business, and Google's DeepMind competing directly with OpenAI in model development, both companies recognize the practical benefits of this arrangement.
OpenAI has been aggressively expanding its computing capacity through multiple channels. Earlier this year, it partnered with SoftBank and Oracle on the $500 billion Stargate infrastructure program and signed deals worth billions with CoreWeave. The company is also developing its first in-house chip to reduce dependency on external hardware providers. With 500 million weekly active users and 3 million paying business customers, OpenAI's infrastructure needs will only continue to grow as it pursues its ambitious expansion plans.