Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Microsoft is making Elon Musk’s start-up xAI’s artificial intelligence models available to its cloud computing customers, in the latest signal of the technology giant’s cooling relations with ChatGPT maker OpenAI.
The Seattle-based group said on Monday that developers using its Azure AI Foundry platform would be able to purchase xAI’s latest Grok models under the same terms as if they were purchasing licenses for OpenAI’s equivalent products.
The move means that users will receive “service parity”, such as preferential access to cloud computing power, regardless of whether they choose to use OpenAI or xAI’s models.
Microsoft, which is OpenAI’s biggest backer, has increasingly sought to disentangle itself from the San Francisco-based start up led by Sam Altman. The move also comes as Musk is in the midst of a bitter legal dispute with Altman over OpenAI’s plans to convert into a for-profit enterprise.
“What we’re trying to do is simplify the purchasing experience for customers and the user experience to make it look more like [what] we do for the OpenAI models,” said Eric Boyd, corporate vice-president of Microsoft’s Azure AI Platform.
Microsoft has pumped more than $13bn in investment into OpenAI since 2019. But tensions have grown between the groups because of the ChatGPT maker’s demands for more computing power from Microsoft, while also competing with its main benefactor by selling AI products designed for enterprise customers.
Offering xAI’s models on the same commercial terms as OpenAI will generate interest across the tech industry, given the animus between the chief executives of the rival groups.
“We don’t have a strong opinion about which model customers use. We want them to use Azure,” Boyd said. “If [customers are] on Azure and they’re finding the thing that they need, we’re going to be quite happy with that outcome.”
He added: “We have great partnership with OpenAI . . . contrary to whatever you may read.”
The software giant also said on Monday that it would start to rank AI models to enable customers to choose the “top performing” option for particular tasks.
Microsoft already offers developers more than 1,900 models from a host of companies including freely available “open” models developed by the likes of China’s DeepSeek, Meta and French start-up Mistral.
Separately, Microsoft will tell developers at its Build conference in Seattle this week that it is adopting Amazon-backed start-up Anthropic’s Model Context Protocol. MPC, which is also being used by Google and OpenAI, standardises the way so-called AI “agents” communicate and has fast become the industry norm.
The move shows a willingness for rival tech groups to co-operate on a key area of AI development, creating a path for more powerful digital assistants to interact with one another to complete tasks.
The Build conference will show how Microsoft is increasingly forging its own path in the race to commercialise generative AI.
Satya Nadella, Microsoft’s chief executive, has distanced the company he leads from OpenAI’s ambitious vision for general artificial intelligence — where systems match or surpass the abilities of humans.
Instead, Nadella has argued that leading models will be “commoditised”, or have less value than being able to sell companies AI-enabled applications and digital assistants built on top of these programmes.