Microsoft teases new era of AI-driven devices

SAN FRANCISCO — Microsoft (MSFT.O) on Tuesday announced a sweeping slate of AI initiatives, from autonomous workplace assistants and gadgets to Nvidia-powered PCs and a new in-house reasoning model, in a major push to move beyond apps and remake computing around artificial intelligence.

At its annual software developer conference, Microsoft Build in San Francisco, executives showcased a broader shift in company strategy. The tech giant is pushing to replace the traditional model of navigating software with one in which AI agents carry out complex tasks autonomously.

By pairing those agents with new devices, powerful PCs, and its own models, Microsoft is trying to control more of the end-to-end AI system — called a "stack" — and lock in enterprise customers as competition from rivals OpenAI and Anthropic intensifies.

The Nvidia-Powered 'Dream Machine'

Microsoft showed off a new computer called the Surface RTX Spark Dev Box, loaded with an Nvidia (NVDA.O) chip. CEO Satya Nadella called the computer a "dream machine" and noted that he is already on the waiting list to buy it.

Nvidia has stated that computers fitted with its new RTX Spark PC chip will bring AI capabilities directly to personal computers. The Surface RTX Spark Dev Box follows a laptop that Microsoft introduced with Nvidia this week. During the event, executives showed the device running an AI model with 120 billion parameters — a rough measure of a model's complexity — that most standard PCs would be unable to load.

While this new wave of PCs is priced to compete with Apple's (AAPL.O) premium offerings, market analysts note it may take some time for businesses to widely adopt the new hardware.

Project Solara: Rethinking Hardware

Executives also revealed Project Solara, a family of prototypes that includes devices the size of a smart speaker or keycard badge, powered by chips from Qualcomm (QCOM.O) and MediaTek (2454.TW).

These devices feature screens and microphones, but instead of running a traditional operating system and apps like a smartphone, they host AI agents that communicate with cloud-computing systems to carry out specific, specialized tasks — such as documenting a medical visit with a nurse.

"Whenever these new platforms come, you get to rewrite even the rules of how new platforms operate," Nadella said during his keynote address. "That's what we're trying to get done with Project Solara, so that you, as developers and enterprises, have the flexibility to imagine the form factors that you want and have your agents be ubiquitous."

Microsoft is actively competing against rivals to sell cloud-based AI tools for coding and other enterprise tasks. Concurrently, it is trying to nudge customers toward running AI technologies directly on laptops and desktop computers powered by its Windows operating system.

Integrating OpenClaw for Enterprise

Microsoft also announced it is developing tools to help Windows run OpenClaw, an open-source software that can direct groups of AI agents to carry out everyday tasks for users.

The goal is to make OpenClaw — which has gained immense popularity in China and previously helped rival Apple sell Mac computers — safe for businesses to use on computers containing sensitive corporate data. During an onstage demonstration, executives showed how a corporate IT department could use these tools to prevent users from inadvertently deleting all the files on their desktops.

"You can totally run OpenClaw inside your company now," said Peter Steinberger, the software engineer who created OpenClaw, during his onstage appearance.

New Autonomous Agents and In-House AI Models

As part of its software evolution, Microsoft will introduce a new AI agent called Scout within its Copilot ecosystem. Scout can autonomously carry out tasks like gathering emails or messages that require specific decisions by the user to move forward.

Furthermore, Microsoft provided an update from its dedicated AI unit focused on "superintelligence." Aiming to catch up to Anthropic and OpenAI, the unit released what it called the most efficient transcription AI model of any cloud hyperscaler, alongside a new image model designed to vie with Google's offerings.

MAI Thinking-1: Microsoft's first proprietary reasoning model matched the performance of Anthropic's Claude Opus 4.6, launched earlier this year. (Anthropic recently announced Opus 4.8 last week).

Independence: These new models underscore Microsoft's strategic efforts to build frontier AI independent of OpenAI, the startup lab it has long backed and partnered with.

A Frontier Healthcare Partnership

One effort distinguishing Microsoft AI's work is its focus on medical diagnostics, which was initially announced late last year. The company has now reached a landmark deal with the Mayo Clinic to build frontier healthcare AI, drawing on Microsoft's reasoning and compute capabilities alongside Mayo Clinic's clinical expertise and data.

In a joint interview, Microsoft AI CEO Mustafa Suleyman and Mayo Clinic CEO Gianrico Farrugia said the partnership emerged from direct meetings between Farrugia and Nadella.

The ultimate goal is to improve patient outcomes with AI that acts as an active team member to get "to a diagnosis faster and better," Farrugia said. -Reporting by Stephen Nellis and Jeffrey Dastin in San Francisco; Editing by Muralikumar Anantharaman, Rod Nickel, Matthew Lewis and Cynthia Osterman/Reuters