Reports suggest that Microsoft is building AI models to rival its longtime partner OpenAI. Under AI Chief Mustafa Suleyman, the company has developed the Phi-4 multimodal and Phi-4-mini models, aiming to reduce reliance on OpenAI.
Satya Nadella, Microsoft’s CEO said in a podcast: “We’re a full-stack systems company, and we want to have full-stack systems capability,” highlighting the company’s focus on integrating models into comprehensive systems and products.
READ: In shareholder letter, Satya Nadella touts Microsoft’s record AI growth, expands future vision (October 24, 2024)
This comes amid growing tensions between the two companies. Microsoft has been pushing for access to OpenAI’s o1 model details, a request that has been denied. Microsoft’s in-house MAI-1 model, with 500 billion parameters, is said to rival OpenAI’s offerings, marking a potential shift in Microsoft’s AI strategy. “Sam Altman and Suleyman are not exactly best friends,” Salesforce CEO Marc Benioff said.
Microsoft seems to be moving away from complete reliance on OpenAI, by testing AI models from competitors like xAI, Anthropic, and Meta. “I do believe the models are becoming commoditized, and in fact, OpenAI is not a model company, it is a product company,” Nadella pointed out.
Meanwhile, OpenAI’s new cloud strategy, Stargate Project, hints at reducing dependence on Microsoft Azure.
OpenAI also signed a $12 billion deal with the GPU-heavy cloud service provider CoreWeave on Monday. The agreement aims for CoreWeave to deliver AI infrastructure to OpenAI, expanding the AI research giant’s compute capacity for training and delivering its latest models at scale for its hundreds of millions of users around the world.
The deal also factors into the rift between OpenAI and Microsoft, since Microsoft was CoreWeave’s biggest customer prior to this deal. In fact, in 2024, Microsoft accounted for 62% of CoreWeave’s revenue, which grew to a stunning $1.9 billion — nearly an eightfold increase from just $228.9 million in 2023.
Big tech companies are increasingly dependent on one another to develop advanced AI models, as many rely on shared resources, partnerships, and technologies to build their systems. With dependencies on cloud infrastructure, data, and computing power, creates an interconnected web where each player relies on the strengths of others.
READ: Satya Nadella, CEO of Microsoft Corporation (October 11, 2023)
To reduce this dependence, companies could focus on developing their own proprietary technologies, invest in in-house research for foundational models, and build independent infrastructures.
As AI becomes more commoditized, some may also turn to open-source alternatives or form new alliances to reduce reliance on any single partner, fostering a more competitive and diversified AI ecosystem.