Microsoft has made $30B from OpenAI

THE AI BRIEF

Today's signal: The Microsoft-OpenAI revenue split just shifted in OpenAI's favor by $97 billion. Plus Google's first AI-generated zero-day, Mira Murati's lab finally shipped, and four other things to track this morning.

In today’s issue:

• Main story: Microsoft has made $30 billion from OpenAI

• Also worth knowing: OpenAI/Anthropic acquisition talks, Gemma 4 speed gains, Chrome's silent AI install, and browser agent pricing

• Free webinar: Build Notion Agents to Automate Complex Tasks, a 30-minute Lightning Lesson, tomorrow. Save your seat →

THE READ

Two numbers from the same set of filings rewire how every Copilot conversation will go for the next four years.

The Information reported Monday that Microsoft has booked more than $30 billion in revenue from OpenAI-tied products. About $23 billion of that is Azure rentals OpenAI itself paid to train and serve its models. The rest came from reselling OpenAI's models to enterprises through Azure OpenAI Service. Microsoft put $13 billion into OpenAI starting in 2023, and has now earned back more than twice that figure.

A separate Information story put numbers on the other side of the trade. Under the original 2023 agreement, OpenAI owed Microsoft 20% of its revenue, which could have hit $135 billion by 2030. The restructured deal from last fall caps that payment at $38 billion. OpenAI keeps $97 billion it had previously owed. In exchange, Microsoft gave up its exclusive right to resell OpenAI models, and accepted a shorter partnership window.

AI READY PRO · FREE UNTIL FRIDAY

I will get you AI trained in 30 days for free

After several months of running it quietly with top AI operators and teams, I’m excited to launch AI Ready Pro today to our newsletter subscribers like you.

It’s a 30-day personalized AI training program that is personalized to you and where you are in your AI journey. It’s the culmination of 100’s of hours spent teaching AI to top Fortune 500 AI teams and operators (and working with Mark Cuban).

Each day, you receive a 10-15 minute exercise to complete to improve your AI skills.

This isn’t vague AI theory or a way to pitch you a tool. It’s 30 days of learning to use AI in your real work. So you come out the other side with the skills to use AI to actually improve you and your teams output.

Until this Friday (May 15), it’s free for newsletter readers who complete the extended assessment. This assessment will help us personalize the learning experience to you.

If you lead a team, AI Ready Team is open today too, also for readers first. It’s the same engine, but to train your whole organization. You’ll get a detailed view of:

• Your team’s AI readiness level + detailed strengths & weaknesses

• Ranked list of AI automation opportunities

• Get tactical advice on how to advance AI efforts in your org

Leading a team? Take the Team assessment instead.

The simple read: Microsoft already won on the deal it cut three years ago, and OpenAI just bought back its future. Both can be true. Microsoft's $30 billion came mostly from computing that any frontier lab would have rented from somebody. The $97 billion OpenAI clawed back is what funds its new Broadcom, Cerebras, and Oracle commitments without immediately running out of cash.

For an operator buying enterprise AI, the more interesting effect sits one layer down. Microsoft no longer has exclusive distribution of OpenAI's models. Anthropic is now available inside Microsoft 365, Google is selling Gemini Enterprise into the same accounts, and OpenAI just launched its own services arm that sells directly into large enterprises. Two years ago, Microsoft was effectively the only door to GPT-grade capability inside an enterprise stack. That is no longer the case. If your renewal cycle is in the next six months, it is a different negotiation than the one your procurement team modeled a year ago.

The bear case is worth holding. Most companies will not switch off Microsoft over a contract change. The friction is in identity, data, tenant settings, and the workflows that already run on Azure. The deal does not make Microsoft easy to leave, but it makes Microsoft easier to negotiate with. I want to be honest about that distinction, because what I keep hearing from operators is "we got a better price this cycle," not "we moved."

SPONSORED BY CLUTCH

Hire secure AI teammates that work 24/7.

Hire pre-built AI teammates. Give your engineers and operators a platform to ship their own AI apps. Stop losing sleep about what is running where.

Clutch is the platform behind both: pre-built agents for the workflows your ops team should automate first, plus the integration plane your team's vibe-coded apps and Claude Code projects plug into. One platform. Real production. Visible and safe by default.

Built for ops, engineering, and security teams that are tired of the shadow-AI surface area inside their own company.

ALSO WORTH KNOWING

Google's threat group reported the first AI-generated zero-day used in a real attack. 
GTIG's 2026 AI threat report says adversaries are now using LLMs to discover and weaponize zero-day vulnerabilities, not just write phishing emails. The chief analyst called it "the tip of the iceberg," and the report frames LLMs as active participants in offensive chains, not passive advisors.

Thinking Machines published its first product post. 
Mira Murati's lab detailed "interaction models," AI systems built for continuous real-time collaboration across voice, video, and text rather than turn-by-turn chat. It is the lab's first concrete product direction since its $2 billion seed and reads as a direct counter to the agentic-first push from OpenAI, Anthropic, and Google.

Anthropic committed to spend $200 billion on Google Cloud and TPUs. 
The new deal stacks on top of last week's Memphis supercluster takeover, putting Anthropic on three different compute footprints (Amazon Trainium, Google TPU, the former xAI cluster). Multi-cloud is no longer optional at the frontier lab tier.

Palo Alto Networks said three weeks of model-assisted penetration testing matched a full year of human pentester coverage. 
The figure came in the company's 2026 internal AI security report. The defensive side of the same trend that Google's GTIG report describes from the offensive side.

WATCHING TOMORROW

OpenAI's Parameter Golf research drop today is short on operator implications, but will set up Thursday's read on how research efficiency claims map to actual inference costs. Microsoft's Build keynote is also on deck for late this week, where Satya Nadella is expected to address the post-exclusivity Copilot roadmap directly.

Back tomorrow,
Haroon