Microsoft Charts New AI Strategy Following Reworked OpenAI Agreement

After signing a revised agreement with OpenAI, the parent company of ChatGPT, in 2025, the tech giant is venturing into new research areas in the hopes of bringing its AI models up to speed with competitors. In 2019, Microsoft made a foresightful investment in OpenAI, which was accompanied by an agreement that cemented the two companies' relationship in more than just financial terms.

OpenAI would have no choice but to use Microsoft's fledgling Azure cloud unit going forward. In contrast, Microsoft would stay out of OpenAI's way if it meant building artificial intelligence (AI) systems that could outperform humans in knowledge-based tasks—a phenomenon known as superintelligence.

Microsoft and OpenAI Re-assessing their Deal

Due to the rapid growth of ChatGPT in 2023 and 2024, OpenAI and Microsoft started to re-evaluate their agreement. Neither company wanted to be dependent on a single source for such a crucial component. Last autumn, as part of their reorganisation, OpenAI decided to use other cloud providers; thus, they relaxed the agreement. Microsoft, meanwhile, might try to develop artificial intelligence in its own unique way. Microsoft retained approximately 27% ownership in OpenAI during that period. This week, OpenAI's $122 billion fundraising round concluded, which may have reduced that share. An initiative to create "frontier" models has been launched by Microsoft in an effort to compete with models from ChatGPT, Anthropic, Google, and China.

Mustafa Suleyman, CEO of Microsoft AI, announced the creation of a new superintelligence team in November, shortly after OpenAI's semi-divorce. This marked the start of Microsoft's drive to gain greater control over its artificial intelligence technology stack.

"If we want to be on the cutting edge, we need to build our own foundation models," Suleyman stated in an interview in February. There is no universally accepted definition of "superintelligence", yet Microsoft's is spot on: Superintelligence is an enterprise product from Microsoft that operates strictly according to the specifications set by its clients. The initiative was restructured in March by CEO Satya Nadella. To concentrate entirely on frontier models, Suleyman delegated responsibility for artificial intelligence services, including the pricey Microsoft 365 Copilot AI assistant.

Investors Not Happy With Azure Sales

A key internal AI resource dispute surfaced during Microsoft's second-quarter results call in January. Azure sales may have exceeded expectations, according to CFO Amy Hood, who explained the failure to investors by saying that Microsoft chose to keep data centre capacity for internal services and research, which is why Azure sales fell short.

Microsoft was unable to meet all of its internal and external computing demands with the $160 billion it spent on capital expenditures over the previous six quarters. However, Suleyman's initiative has been impacted by the conflict between renting out AI servers and using them internally. He characterised the new artificial intelligence models he released this week for internal purposes—like transcriptions—as "mid-class" and said they wouldn't compete with the top frontier models. The internal compute shortage, which he said would be resolved this year, was the limiting factor, he said.