**Summary**
At Build 2026 Microsoft unveiled a comprehensive push to become self‑sufficient in AI, reducing reliance on external partners like OpenAI and Anthropic. Key announcements include:
- **In‑house model family** – seven new models (MAI Thinking One, MAI Code 1 Flash, MAI Image 2.5, MAI Transcribe 1.5, MAI Voice 2, etc.) trained from scratch on commercially licensed data, with MAI Thinking One claimed to outperform GPT‑5.5 on quality while being ~10× cheaper to run.
- **Agent stack** – a new “always‑on” autopilot agent called **Scout**, integrated across Microsoft 365 (Teams, Outlook, OneDrive, SharePoint) that autonomously schedules meetings, prepares materials, flags risks, and learns from user context via **Microsoft IQ**.
- **Microsoft IQ** – a unified intelligence layer (Work IQ, Fabric IQ, Foundry IQ, Web IQ) that grounds agents in organizational data, business semantics, documents, and web search, reducing hallucination and improving usefulness.
- **AI security for developers** – **Mdash**, a multimodel agentic security system deploying >100 agents to hunt exploitable bugs in code.
- **Quantum hardware** – **Majorana 2**, a topological quantum chip with dramatically improved qubit coherence (≈20 s average lifetime, 1,000× more reliable than its predecessor), positioning Microsoft toward a commercially viable quantum computer by 2029.
- **AI‑driven R&D platform** – **Microsoft Discovery**, now generally available, lets companies run teams of AI agents to accelerate frontier research (including Microsoft’s own quantum work).
Overall, Microsoft is building a full‑stack AI ecosystem—models, agents, intelligence layers, security tools, developer integrations, and quantum hardware—hosted on Azure to control costs, margins, and the end‑to‑end AI value chain, directly challenging its former partners and competitors.
1. Microsoft revealed seven in‑house AI models at Build 2026.
2. Microsoft introduced a new agent stack at Build 2026.
3. Microsoft launched an always‑on personal agent named Scout.
4. Microsoft announced the Microsoft IQ intelligence layer.
5. Microsoft introduced an AI security system for developers.
6. Microsoft unveiled a major quantum chip upgrade called Majorana 2.
7. The headline model is MAI Thinking One, Microsoft’s first reasoning model trained from scratch.
8. MAI Thinking One was trained on clean commercially licensed data without distillation from third‑party systems.
9. MAI Thinking One has 35 billion active parameters.
10. Some sources list MAI Thinking One’s context window as 256,000 tokens; Microsoft’s developer coverage also cites 128,000 tokens.
11. After tuning for consulting firm McKenzie, MAI Thinking One outperformed OpenAI’s GPT‑5.5 on quality while projecting ~10× better cost efficiency based on public pricing data.
12. Microsoft launched MAI Code 1 Flash, a coding model that converts text descriptions into source code for apps and websites.
13. MAI Code 1 Flash is rolling out across GitHub Copilot and Visual Studio Code.
14. In blind evaluations run by independent human rating partner Serge, MAI Thinking One was preferred over Anthropic’s Claude Sonnet 4.6 for coding tasks.
15. MAI Thinking One matches Claude Opus 4.6 on coding benchmarks including SWEBench Pro.
16. MAI Code 1 Flash is available in private preview on Microsoft AI Foundry.
17. Microsoft announced MAI Image 2.5 and a flash variant supporting text‑to‑image and image‑to‑image generation.
18. MAI Image 2.5 is already live in PowerPoint and rolling out on OneDrive.
19. MAI Transcribe 1.5 supports high‑accuracy transcription across 43 languages with streaming coming soon.
20. MAI Voice 2 and its flash variant are available in more than 15 additional languages and can produce new voice options.
21. Microsoft IQ is now generally available as a unified intelligence layer for Copilot and AI agents.
22. Work IQ captures how users work inside Microsoft 365, understanding people, emails, documents, meetings, organizational systems, external sources, and their relationships.
23. Work IQ APIs are set to become available on June 16th.
24. Fabric IQ runs on Microsoft Fabric and works as a semantic foundation for structured business data, described like an ontology.
25. Foundry IQ handles unstructured information, pulling from documents such as wikis, policies, contracts, and the live web.
26. Web IQ (WebQ) gives agents real‑world grounding through web search, is model‑agnostic, and native to the Model Context Protocol (MCP).
27. Microsoft says Web IQ returns relevant information blocks nearly 2.5× faster than the next best alternative.
28. Scout operates under its own governed identity, making its actions traceable to a known actor in the company directory.
29. Scout’s credentials are scoped to the task, protected end‑to‑end, redacted from logs, and managed like a first‑party Microsoft service.
30. Scout is integrated across Microsoft 365 apps: Teams, Outlook, OneDrive, and SharePoint.
31. Scout connects to chats, email, calendar, contacts, documents, browser resources, local desktop resources, and MCP servers.
32. Users interact with Scout via Teams; the desktop app extends its reach into the browser and local system.
33. Scout can proactively schedule and coordinate meetings across time zones, flag important meetings, generate prep materials, identify upcoming deliverables, automatically block calendar time, and spot risks such as stalled decisions.
34. Scout learns over time through Work IQ, building context around how the user works, what they care about, and what needs to happen next.
35. Early desktop Scout was used by Microsoft employees; it is now expanding to a select group of customers in private preview and Frontier organizations.
36. Access to Scout requires Frontier enrollment, Intune policy configuration, and an opt‑in at a station.
37. Users with a GitHub Copilot license can download and install the Scout experience.
38. Scout is built on OpenClaw, an open‑source technology; Microsoft is contributing policy conformance upstream.
39. Scout can only access approved resources; sensitive actions may require human signoff, and Microsoft Purview policies are enforced before any data is sent or written.
40. Microsoft introduced code name Mdash, a multimodel agentic security system that deploys >100 agents to search for exploitable bugs in code.
41. Mdash agents reason about data flows, business logic, exploit chains, and context‑aware fixes inside the developer portal.
42. Majorana 2 is Microsoft’s next‑generation topological quantum chip.
43. Majorana 2 was developed with help from Microsoft Discoveries Agentic AI.
44. Microsoft claims Majorana 2 is 1,000× more reliable than its previous qubit generation.
45. The mean qubit lifetime of Majorana 2 is now 20 seconds, with some instances lasting up to 1 minute.
46. Majorana 2 currently contains 12 qubits.
47. Microsoft says the combination of reliability, 1‑microsecond operations, and very small qubit size (~1/100 mm) puts it on a path toward a commercially valuable scalable quantum computer by 2029.
48. Zulfi Alam, corporate VP of Microsoft Quantum, expects a quantum machine in 2029 that can solve commercially viable reasonable problems.
49. Microsoft’s approach is based on topological qubits and Majorana‑linked physics tied to a quasi‑particle first theorized in the 1930s by Italian physicist Ettore Majorana.
50. Microsoft previously retracted a 2018 Nature paper claiming evidence for the Majorana particle; the new chip and supporting research have not yet been peer‑reviewed.
51. Majorana 2 uses a lead‑based superconductor stack, whereas Majorana 1 used aluminum, to help shield fragile qubits from cosmic disturbances.
52. Ceton Nyak, a Microsoft technical fellow, said the team must improve each year to reach a computer with major commercial and societal value, comparing progress to being 1,000× better than last year.
53. Microsoft Discovery is now generally available and lets companies use teams of AI agents for frontier R&D.
54. Discovery agents can search, research, reason through problems, generate hypotheses, optimize experiments, and help validate ideas while human experts stay in control.
55. Microsoft’s quantum team uses Discovery to manage workflows, automate measurements, improve fabrication, find hidden flaws, and propose better solutions.
56. Discovery helped build an AI agent that cut the cycle of adjusting hundreds of parameters for a topological quantum state by orders of magnitude, adjusting voltages in parallel and mapping conditions continuously.
57. One Discovery agent found an uncalibrated temperature sensor hidden in fabrication data.
58. Microsoft has committed $13 billion to OpenAI, invested up to $5 billion in Anthropic, and sells both companies’ models through Azure.
59. Microsoft is simultaneously an investor, partner, distributor, infrastructure provider, and direct competitor to OpenAI and Anthropic.