Microsoft Plans to Build Advanced AI Models by 2027

Microsoft wants to be a key player in the development and production of leading-edge artificial intelligence models. And the tech giant has a 12- to 18-month plan in place to reach that level.

As reported by Bloomberg, the news is a confirmation of sorts that its models are not yet up to the standard of OpenAI, Anthropic, and Google, seen as the three leaders in AI model research and sophistication. Microsoft has leaned on its exclusive licensing arrangement with OpenAI since 2020, integrating GPT into most of its Office suite, but now sees in-house AI models as its future.

“We must deliver the absolute frontier,” said Mustafa Suleyman, chief executive of Microsoft AI, to Bloomberg. “Certainly by 2027, the objective is to really get to state-of-the-art.”

Suleyman’s role at Microsoft was recently narrowed to focus on model development, while ex-Snapchat executive Jacob Andreou was put in charge of Copilot. The next 12 months will see a ramp-up in computing power to build leading-edge foundational AI models, similar to GPT or Gemini.

Suleyman sees the future of AI as one of heavy automation and agent usage, recently saying that most office work will be automated in the next 12 to 18 months. To that end, the company is investing more heavily in AI agents and automated systems for office work.

Microsoft’s place in the AI market

It is unclear how many people use Microsoft Copilot, but it is nowhere near ChatGPT on the consumer side and is not attracting as many enterprise customers as Anthropic’s Claude. While some big tech companies, such as Apple and Amazon, appear to have stepped back from foundational model development, Microsoft still clearly has ambitions to be a major player in this market.

The software giant has continued to launch new Copilot products and appears more focused on capturing enterprise customers than consumers. It launched Copilot Cowork last month, but it uses Claude to automate workplace tasks, which does little to support its push towards in-house models. Its recent AI image model is a better sign of things to come, representing a major step up from MAI-Image-1 and built for creatives.

The relationship with OpenAI has cooled somewhat since its initial $1 billion investment and follow-up $10 billion deal in January 2023. Microsoft secured an exclusive commercial license for GPT technology in 2020, along with a revenue-sharing agreement in 2023.

As OpenAI has shifted further towards a for-profit structure, it has loosened some of these constraints, prompting Microsoft to explore alternatives such as Anthropic and accelerating its own model development.

For more on Microsoft’s AI ambitions, check out its latest MAI-Image-2 model launch and what it signals for the company’s creative AI push.

The post Microsoft Plans to Build Advanced AI Models by 2027 appeared first on eWEEK.