Microsoft Project Solara

Learning Objectives

• Understand what Project Solara is and what "agent-first" computing means
• Explain how Solara differs from the app-centric model that defines today's phones and PCs
• Evaluate where agent-first devices might fit in enterprises — and what is still unproven

What Is Project Solara?

Project Solara is Microsoft's platform for building agent-first devices — hardware designed so that an AI agent, rather than a grid of apps, is the main way people interact with the device. It was unveiled at Build 2026 by Microsoft technical fellow Steven Bathiche, and Microsoft describes it as a chip-to-cloud platform: it spans the silicon in the device, the operating system on top of it, and the cloud services the agent reaches out to.

The core thesis is a shift in how computing is organized. For decades the unit of software has been the app — you open one program to write, another to message, another to look something up. Solara's bet is that the unit becomes the agent: you state what you want, and the agent coordinates across services and tools to do it, choosing the right capabilities behind the scenes. Devices then become, in Microsoft's framing, the best physical surface for a given context rather than a screen full of icons.

How It Works

Solara is built on the Android Open Source Project (AOSP) and uses the Microsoft Device Ecosystem Platform (MDEP) — an enterprise-grade version of Android that Microsoft already developed for hardware such as Teams meeting-room devices. On top of that base, Microsoft provides reference hardware designs for partners using Arm-based chipsets from MediaTek and Qualcomm, so device makers can build agent-first hardware without starting from scratch.

A defining feature is what Microsoft calls a just-in-time interface: the agent experience adapts to whatever device it is running on — a small wearable, a desk device, or a larger display — without developers having to redesign the experience for each form factor. Much of the intelligence is "liminal," moving fluidly between the device and the cloud depending on what a task needs.

Because the platform targets enterprises, manageability and security are built in. Solara devices are designed to work with Microsoft Intune for device management, Microsoft Entra ID for identity, and Windows Hello for Business for sign-in, with privacy controls included.

Reference Devices and Pilots

Microsoft previewed two prototype form factors to make the idea concrete:

• A stationary desk device with a touchscreen, microphones, cameras, a presence sensor, and expansion ports — an always-available agent surface at your workspace
• A wearable smart badge with voice and vision input, a camera, a small touchscreen, cellular connectivity, and fingerprint sign-in — a hands-free agent you carry

Microsoft said it is piloting Solara-based concepts internally across hundreds of employees, with external pilots beginning in the months after Build. Named early partners span retail, healthcare, and other sectors — including CVS Health, Target, Best Buy, Levi's, and AccuWeather — exploring purpose-built agent devices for their frontline and customer-facing work.

Strengths

• A coherent agent-first vision: Rather than bolting an agent onto an app phone, Solara designs the device, OS, and cloud around agents from the start
• Lower barrier for device makers: Reference designs plus an enterprise Android base let partners build specialized agent hardware faster and cheaper
• Adaptive interface: The just-in-time UI lets one agent experience span many form factors without per-device redesign
• Enterprise-ready by design: Built-in Intune, Entra ID, and Windows Hello support address the manageability, identity, and security that businesses require
• Real pilot partners: Named companies across retail and healthcare give the vision concrete early testbeds

Limitations & Considerations

• Early and unproven: Solara is a platform and a set of pilots, not a product with a track record; whether agent-first devices find real demand is unknown
• Depends on agent quality: The model only works if agents are reliable enough to trust with real tasks — a bar today's agents do not always clear
• Enterprise-first: The initial focus is purpose-built business and frontline devices, not a mainstream consumer device
• Ecosystem risk: Success requires device makers, developers, and customers to adopt a new computing model at once — a hard coordination problem
• No general availability: Timelines beyond the initial pilots, pricing, and the breadth of shipping hardware are not yet established

Best Use Cases

When the app model still wins:

• General-purpose personal computing → today's phones and PCs remain the right tool for broad, self-directed work
• Tasks needing precise manual control → app-first interfaces still beat delegating to an agent when you want to drive every step

Getting Started

1. Review Microsoft's Build 2026 Project Solara materials and reference-design documentation to understand the platform's pieces
2. If you build hardware, evaluate the Microsoft Device Ecosystem Platform base and the MediaTek and Qualcomm reference designs for agent-first devices
3. If you run an enterprise, consider where an agent-first device — a desk surface or a wearable badge — could fit a frontline or customer-facing workflow
4. Watch how the early pilots with partners such as CVS Health and Target play out before planning any deployment of your own

Key Takeaways

• Project Solara is Microsoft's chip-to-cloud platform for agent-first devices, unveiled at Build 2026 by Steven Bathiche
• It reimagines computing around agents instead of apps, providing reference hardware designs, an enterprise Android operating system (MDEP), and a just-in-time interface that adapts across form factors
• Two prototypes — a desk device and a wearable smart badge — make the idea concrete, with Intune, Entra ID, and Windows Hello management built in
• It is in internal use and early external pilots with partners including CVS Health, Target, Best Buy, Levi's, and AccuWeather — not a generally available product
• The bet depends on agents becoming reliable enough to trust with real work; whether agent-first devices find demand is the open question