Satya Nadella Just Declared War on OpenAI: Microsoft's MAI Models Signal a Stunning Breakup

For years, Microsoft held the AI world's most enviable position: it owned the cloud (Azure), the developer platform (GitHub), and the enterprise desktop (Windows), while letting OpenAI build the brains. That era ended on Wednesday.

At its annual Build developer conference in San Francisco, CEO Satya Nadella unveiled Microsoft's first family of proprietary reasoning models — the MAI series. The debut model, MAI-Code-1-Flash, is a 35-billion-parameter reasoning engine built for high-efficiency code generation at low token cost. It is available now in private preview on Microsoft Foundry.

The subtext was impossible to miss: Microsoft no longer wants to be OpenAI's landlord. It wants to be its rival.

Project Soltera: The Android Microsoft Never Built

The most surprising announcement came from Steven Bathiche of Microsoft's Applied Sciences Group. Project Soltera is not Windows. It is not a cloud service. It is an Android-based operating system designed from the ground up for AI agents.

Soltera is described as "a chip-to-cloud platform for an open, multi-agent world." To prove it's real, Microsoft showed two devices:

• A employee badge with Wi-Fi, Bluetooth, a touchscreen, and a fingerprint reader

• A compact desktop unit resembling an Amazon Echo Show

Both are designed for one purpose: making AI agents the primary interface for workplace computing. Think of Soltera as Microsoft's answer to what happens when no one wants to open an app anymore — they just talk to an agent.

Microsoft Scout: Your Always-On Assistant

Nadella also unveiled Scout, an "always-on" assistant based on OpenClaw — the open-source software that exploded in popularity late last year. Scout prepares for meetings, manages schedules, and drafts emails. It is essentially Microsoft's version of Google's Gemini Spark, and for now, it is available only to a limited group of customers.

During a breakout session, OpenClaw creator Peter Steinberger made a striking prediction: "It's only been a year since we've had coding agents, half a year since they got good, and now there's no stopping them."

Microsoft also announced Execution Containers for OpenClaw — secure environments where companies can run autonomous agents without data leakage risks.

The Quantum Clock is Ticking

While the AI news dominated headlines, Microsoft quietly dropped a bombshell in quantum computing. The company's Majorana 2 topological quantum chip is 1,000 times more reliable than its predecessor. Qubits on Majorana 2 survive for an average of 20 seconds — compared to mere milliseconds on Majorana 1.

Microsoft claims this paves the way for a quantum computer capable of solving commercially useful problems within three years. The global race to build useful quantum machines just got a new leader.

Microsoft Discovery: AI for Scientists

Finally, Microsoft unveiled Microsoft Discovery, an agentic AI platform built for researchers. Mining giant BHP is already using it to find copper-leaching solutions "in months instead of years." GSK is using it for drug discovery. The platform combines public scientific literature with internal knowledge, running simulations for hours or days autonomously.

What It All Means

Microsoft is no longer playing the AI game as a partner. It is building its own models (MAI), its own operating system for agents (Soltera), its own assistant (Scout), and its own quantum future (Majorana 2). The message to developers is clear: build on Azure, build with MAI, and build for a world where AI agents live everywhere — on your desktop, on your badge, and in your lab.

OpenAI, Google, and Anthropic have been warned. The $100 billion question is whether Microsoft can execute.