The Vendor Map Just Flipped: OpenAI on AWS, Microsoft Off OpenAI, Anthropic Going Public
June 4, 2026
Executive Summary
Microsoft used Build 2026 to declare independence from OpenAI. The company shipped MAI-Thinking-1, its first in-house frontier reasoning model, the Scout autopilot agent, and a repositioning of Windows as the OS for agents. Anthropic confidentially filed for IPO at a $965B valuation. OpenAI frontier models and Codex hit GA on AWS Bedrock.
Top Stories
1. Microsoft Ships MAI-Thinking-1, Its First In-House Frontier Reasoning Model

Microsoft launched MAI-Thinking-1 at Build 2026, a sparse Mixture-of-Experts model with 35B active parameters and a 256K token context window. Trained end-to-end on commercially licensed data, with no distillation from OpenAI or any third party. Benchmarks: 97.0% on AIME 2025, 94.3% on GPQA-Diamond ahead of Gemini 2.5 Pro and o3, parity with Claude Opus 4.6 on SWE-Bench Pro, and a win over Claude Sonnet 4.6 in blind side-by-side evaluations. Now in private preview via Microsoft Foundry.
Business Impact: Microsoft no longer depends on OpenAI for frontier reasoning. Expect MAI models to surface harder in your next Azure renewal. Run a parallel eval against your current frontier model. Licensing terms will matter as much as benchmarks.
2. Microsoft Repositions Windows as the OS for AI Agents, Launches Scout

Microsoft introduced Scout, the first in a new agent category called Autopilots, always-on agents with their own identity that act on the user's behalf. Scout runs across Outlook, Teams, OneDrive, and the broader Microsoft 365 surface, built on the OpenClaw open-source framework with Entra identity and Purview compliance. Microsoft also announced Windows Agent Framework APIs, an Agent Store, and a native GitHub Copilot desktop app with parallel sessions via git worktrees. Scout is experimental through the Frontier program.
Business Impact: If you run a Microsoft 365 estate, your AI strategy needs to account for agents operating persistently inside your tenant, with their own identity. Build identity, audit, and access policies for agent-class principals now. Compliance overhead will exceed licensing cost.
3. Anthropic Files Confidential S-1 After $65B Raise at $965B Valuation

Anthropic announced a $65 billion Series H on May 28 at a $965 billion valuation, then confidentially filed its draft S-1 with the SEC on June 1. Annualized revenue is $47 billion. Reports place the target listing in October at $1.75 to $1.8 trillion with a raise up to $75 billion, the largest IPO in history if it lands.
Business Impact: Anthropic now runs on a public-market clock alongside OpenAI's confidential filing two weeks ago. Quarterly disclosures from both will reset the market's view of AI unit economics. If your vendor scorecard prices Anthropic as a private-company risk, recalibrate.
4. OpenAI Frontier Models and Codex Hit GA on AWS Bedrock

GPT-5.5, GPT-5.4, and Codex are now generally available on Amazon Bedrock. Pricing matches OpenAI first-party rates and usage counts toward AWS commitments. Codex runs through its App, CLI, and IDE integrations, with inference routed through Bedrock and AWS-native IAM, VPC isolation, and encryption. OpenAI also rolled out role-specific Codex plugins and a Sites preview. The company reports 4 million+ weekly Codex users.
Business Impact: OpenAI is multi-cloud at production scale. If your AWS commitments limited you to Anthropic on Bedrock, that constraint is gone. Re-evaluate vendor routing for coding agents.
Quick Bytes
- Microsoft MAI suite: Microsoft also released MAI-Code-1-Flash, MAI-Image-1, MAI-Voice-1, and three other MAI models alongside Thinking-1. (Microsoft AI, June 2)
- Claude outage: Claude suffered a partial outage on June 2 starting around 2:10am ET, resolved within hours. (TechRadar, June 2)
- OpenAI x Rosalind: OpenAI announced a partnership with Rosalind Biodefense for biothreat resilience research. (OpenAI, June 1)
Industry Impact
The week's pattern is platform consolidation and vendor independence. Microsoft moved off OpenAI dependency with MAI-Thinking-1. OpenAI moved off Microsoft cloud exclusivity with AWS Bedrock GA. Anthropic moved toward public-market accountability. The three frontier vendors now each control their own model, their own runtime, and their own distribution. For enterprise teams, the routing logic that worked twelve months ago no longer applies. The next 90 days are a re-architecture window.
How Azumo Helps
Switching frontier vendors or operating multi-model is engineering work, not a sales decision. Azumo brings senior AI engineers, nearshore from LATAM, with experience in model evaluation, RAG architectures, agent orchestration, and multi-cloud deployment. We integrate with your team, in your time zone and your sprints. 300+ AI and software projects delivered, SOC 2 compliant, 95% NPS.
Sources
- Introducing MAI-Thinking-1 — Microsoft AI
- Microsoft Build 2026: MAI Keynote Transcript — Microsoft AI
- Microsoft Build 2026: MAI-Thinking-1 Is First In-House Reasoning Model, Trained Without OpenAI Data — TechTimes
- Introducing Microsoft Scout: Your always-on personal agent — Microsoft 365 Blog
- A new category of agents: Microsoft reveals Scout, its first Autopilot — TechRadar
- Anthropic confidentially submits draft S-1 to the SEC — Anthropic
- Claude maker Anthropic files for IPO, setting up public-market test of AI boom — CBS News
- Anthropic, maker of Claude, files with the SEC to go public in an IPO — Washington Post
- Anthropic confidentially files its S-1 first, but the IPO race with OpenAI is just beginning — Fortune
- OpenAI models, Codex, and Managed Agents come to AWS — OpenAI
- OpenAI models GPT-5.5 and GPT-5.4 and Codex now available on Bedrock — Amazon
- OpenAI models and Codex on Amazon Bedrock are now generally available — AWS Machine Learning Blog
- Claude is down for many: outage tracking — TechRadar
