Microsoft launches company to help companies adopt AI

Microsoft announced on Thursday (2) that it is creating a new company that will help customers select AI technologies that work for business and generate a return on investment.

Microsoft Frontier Company, as the new operating entity is called, will start with $2.5 billion in funding from the technology giant to work with customers including Unilever and Novo Nordisk.

Large corporations are increasingly less dependent on contracting AI from a single vendor, such as Anthropic or OpenAI, and are instead opting to utilize a combination of technologies, including open source models, adapting them to their needs.

This strategy is expensive and prolongs the time needed to generate a return on investment. Microsoft Frontier Company will offer customers assistance in selecting and integrating AI tools - from Microsoft and third parties - with those customers' unique internal data. Crucially, customers will be able to keep the results of this work, rather than sending them back to Microsoft. The Windows operating system maker joins companies such as Palantir Technologies, which already uses Nvidia's open source models for this type of work with large clients, and cloud competitor Amazon Web Services, which launched its own embedded engineering unit with a $1 billion investment.

Patrick Moorhead, chief executive of analytics firm Moor Insights & Strategy, said large companies suspect that using models from Anthropic and OpenAI will ultimately give these cutting-edge labs the expertise to compete with them, especially in areas like programming and law. Microsoft owns a stake in OpenAI, the creator of ChatGPT, and added Anthropic models to the Copilot AI assistant earlier this year, in part in response to growing corporate demand for the AI lab's offerings.

Judson Althoff, chief executive of Microsoft Commercial Business, said the new company was born in part from Microsoft's own experience when models like China's DeepSeek and Google's Gemini began catching up with OpenAI.

"Three years ago, when we created Copilot, we made a mistake by only tying it to OpenAI models," Althoff told Reuters. "It would be desirable for the models to expand their intelligence and for there to be this ability to switch to cutting-edge models and fine-tuning."

The combination of data and models mattered more to the customer than any specific model, and they needed the flexibility to quickly switch between AI models, he pointed out.

Source: CNN