Why OpenAI, Microsoft, and Anthropic Fund One Union
Three fierce AI rivals — OpenAI, Microsoft, and Anthropic — co-fund the same teachers' union training program. Here's the coopetition logic behind the move.
OpenAI and Anthropic are rivals in the most literal sense. They chase the same researchers, the same enterprise contracts, and the same benchmark records. Microsoft is a major backer of OpenAI and also ships its own competing classroom product. And yet, in July 2025, all three put money into the exact same thing: the National Academy for AI Instruction, a teacher-training initiative run with the American Federation of Teachers.
If competition were the only force in play, this would make no sense. Companies do not usually hand their rivals a share of the same goodwill. But there is a well-worn name for what is happening here — coopetition, cooperating on the layer where everyone benefits while competing hard on the layer where money changes hands. Understanding that split explains not just this $23 million partnership, but where the AI-in-education market is actually heading.
Key Takeaways
- They compete on products, cooperate on trust. OpenAI, Microsoft, and Anthropic each sell a rival classroom tool, yet all fund the same neutral training body. The training layer is shared infrastructure; the product layer is the battlefield.
- A fragmented training market would hurt all of them. If every vendor ran its own incompatible teacher training, schools would stall in confusion. One credible, union-led program lowers that friction for the whole category.
- Teacher trust is the real bottleneck. Adoption dies when educators distrust AI output. Funding a teacher-owned academy buys credibility no vendor could manufacture alone.
- The union is the neutral ground. The AFT can say "we vetted this" in a way OpenAI's marketing never could. That neutrality is precisely what makes the money worth spending.
- The same names recur across offers. Anthropic funds this union academy and runs Claude for Education; OpenAI funds it and runs ChatGPT for Teachers. Watching who cooperates and who competes tells you where the market is consolidating.
What Is Coopetition, and Why Does It Fit Here?
Coopetition is what happens when competitors collaborate on a shared foundation because none of them can win without it. Airlines share booking standards and still fight over fares. Chipmakers fund the same manufacturing research and still undercut each other on price. The pattern shows up wherever the size of the whole market matters more, for now, than any one company's slice of it.
AI in K-12 is exactly that kind of market. It is early, it is nervous, and it is bottlenecked not by technology but by human confidence. A teacher who does not trust a model's output stops using every tool, not just one vendor's. So the thing all three companies most need — a workforce that is comfortable directing AI — is a resource none of them can build alone without looking self-serving. Pooling money into a neutral academy solves that. The AFT's own announcement frames the effort as teacher-led infrastructure, which is the whole value: it is credible because it is not any single company's sales channel.
Where They Cooperate vs. Where They Compete
The cleanest way to see the logic is to separate the two layers explicitly.
| Layer | What happens | Who "wins" |
|---|---|---|
| Training / fluency | All three fund the AFT academy; teachers learn to evaluate and direct AI in general | The whole category — bigger, more confident market |
| Product / model | Each ships a distinct offer (ChatGPT for Teachers, Copilot for Education, Claude for Education) | One vendor per school, per contract |
| Trust / credibility | A neutral union owns the training narrative | Every funder borrows the union's credibility |
| Revenue | Districts eventually pay for seats, storage, admin tooling | Whoever the trained teachers choose |
Read top to bottom and the strategy is obvious. Cooperation happens where a rising tide lifts all boats. Competition happens the moment a district reaches for a purchase order. As TIME and EdWeek both noted when the partnership was announced, the striking part is not that these companies compete — it is that they agreed on who should sit in the neutral seat.
Why not just run your own training?
Each company could train teachers directly, and to some extent each does. But vendor-run training carries an unavoidable tax: skepticism. When OpenAI teaches you to use AI, part of your brain files it under "sales." When a union you already trust teaches you — using a curriculum funded by several companies, not one — that tax drops. Credibility is the product being purchased here, and credibility cannot be self-issued. This is the deeper reason a union, not a vendor, leads the effort, which gets a full treatment in Why a Union, Not a Vendor, Leads Teacher AI Training.
Does Cooperation Now Mean They Stop Competing?
Not for a second. The moment a teacher finishes training, three companies want to be the tool they open on Monday morning. OpenAI's move — making ChatGPT for Teachers free through June 2027 — is aggressive product competition aimed squarely at that decision. Anthropic runs its own K-12 push, tracked on this site in the Claude for Teachers series. Microsoft leans on its Copilot and Office footprint. The academy makes the pie bigger; the products decide who eats.
That is why the honest framing is not "rivals made peace." It is "rivals agreed the market is worth growing before they carve it up." For a teacher, the practical implication is freedom: the fluency you build is portable across all three tools. Skills like a Socratic tutor or a broader education agent skills set work the same way regardless of which model wins the contract, because the underlying competency — knowing what to ask and how to check the answer — is vendor-agnostic. That scale ambition, 400,000 teachers by 2030, is worth examining on its own, which is where this series turns next.
Frequently Asked Questions
Why would competing AI companies fund the same program?
Because teacher trust and AI fluency are a shared bottleneck. A fragmented, vendor-by-vendor training market would slow adoption for everyone. Pooling money into a neutral, union-led academy grows the whole category before the companies compete for individual contracts.
Does OpenAI still compete with Anthropic and Microsoft after this?
Yes, fiercely. Each runs a separate classroom product and wants to be the tool teachers actually use. The cooperation is limited to the training and trust layer; product and revenue remain a direct fight.
Is the union just a marketing front for the companies?
The value of the AFT's involvement is that it is not a vendor. A union can vet and endorse training in a way a company's marketing cannot. That independence is the whole reason the arrangement works — and why skeptics still watch it closely.
How much did each company contribute?
Reporting attributes $12.5 million to Microsoft and $10 million to OpenAI, with Anthropic also contributing an amount not disclosed in the coverage reviewed here. The total is described as roughly $23 million.
Does this affect which AI tool a teacher should pick?
Not directly. The training is designed to be tool-agnostic, so the skills transfer across ChatGPT, Claude, and Copilot. The competition among products is a separate decision, covered later in this series.
Part 16 of 100 in the ChatGPT for Teachers series. Previously: The $23M Bet: Inside the National AI Academy. Next: 400,000 Teachers by 2030: Can Training Scale That Fast?. Browse more builder insights or explore AI skills for education at aiskill.market.