Microsoft Is Reportedly Building an AI Language Model to Rival Google and OpenAI

Last week, internal emails sent in 2019 between Microsoft’s Chief Technology Officer Kevin Scott, CEO Satya Nadella and Bill Gates resurfaced as part of the Department of Justice’s antitrust probe into Google. In the emails, Scott writes that he’s “very worried” about the company’s lack of progress in the AI space compared to OpenAI and Google.

That same year, Microsoft invested $1 billion USD to form a partnership with OpenAI, seemingly as a way to remedy its own lack of an AI chatbot. That partnership was reaffirmed in 2023 with another $10 billion USD invested by Microsoft into OpenAI.

Over a year later, Microsoft is finally breaking ground on its own AI language model, according to The Information. The model is apparently being referred to as MAI-1 and is described as “far larger” than any of Microsoft’s previous AI models. People familiar with the project told the publication that its size means it will be more expensive to train and demand more computing power.

The new CEO of Microsoft AI Mustafa Suleyman is reportedly helming the project. Suleyman previously served as the head of applied AI at DeepMind and in a policy role at Google. In 2022, he launched his own AI startup called Inflection through which he released the AI bot “Pi.” Microsoft later bought out Inflection’s intellectual property and hired the majority of its staff.

The bot will reportedly be powerful enough to compete with models from OpenAI, Google and Anthropic, per The Information. It’s also reported to be completely original to Microsoft, not a carryover from Inflection.