Microsoft launches firm to help companies adopt AI

SAN FRANCISCO — Microsoft said on Thursday it is creating a new company that will help ​customers select AI technologies that work for their businesses and generate returns on their investment.

Microsoft Frontier Company, as the new operating entity is called, will kick off ‌with $2.5 billion in funding from the tech giant to work with clients such as Unilever and Novo Nordisk.

Large corporations are ⁠relying less on renting out AI from ​a single provider, such as Anthropic or ⁠OpenAI, and are instead using a mix of technologies, including open-source models, tailoring them to ‌their needs. This is a ‌costly affair and stretches the time it takes to generate a return on ⁠their investment.

Microsoft Frontier Company will offer customers help ⁠to select and integrate AI tools, from Microsoft and outside, with that customer's unique internal data. Critically, the customers will get to keep the results of that work rather than send it back to Microsoft.

The Windows operating system maker joins the likes of Palantir Technologies, which is already using Nvidia's open-source models for such work with large ‌customers, and cloud rival Amazon Web Services, which kicked off ​a $1 billion embedded-engineer unit of its own.

Patrick Moorhead, CEO of analyst firm Moor Insights & Strategy, said large businesses suspect that using models from Anthropic and OpenAI will eventually grant these frontier labs expertise to compete with them, especially in fields such as coding and law.

Microsoft partly owns ChatGPT-maker OpenAI and had added Anthropic's models to its Copilot AI assistant earlier this year, partly in response to booming enterprise demand for the AI lab's ​offerings.

Judson Althoff, CEO of Microsoft Commercial Business, said the new firm was born partly out of Microsoft's own ‌experience when models ‌such as ⁠China's DeepSeek and Google's Gemini began to catch up to OpenAI.

"Three years ago, when we built Copilot, we made a mistake by binding it to OpenAI models only," Althoff told Reuters. "You wanted models to amplify your intelligence and be able to have that sort of swappability for ‌state-of-the-art and fine-tuning."

The combination of ​data and the models mattered more to the customer ‌than any particular model, and ⁠they needed the ​flexibility to switch among AI models quickly, he said.

Related topics