This week the price of artificial intelligence fell in a way it never had before.
On June 1, a lab called MiniMax released M3, an open-weight model that matches the most expensive frontier models on a professional coding test (SWE-bench Pro) and charges about 12 times less to use. It wasn't an isolated case: in a single week, more than 25 new open-weight models appeared. DeepSeek made a 75% price cut permanent. The big providers' prices fell between 40% and 70% so far this year.
The concrete result: today an open model running on a 20-to-50-dollar-a-month server can handle customers, write and follow up on sales, without paying a fee for each person who uses it. The capability that cost a fortune a year ago now fits inside the budget of a streaming subscription.
Most people will read this as a price war between labs. They'll think:
"Great. Artificial intelligence is cheaper. I'll get to use more of it."
And go back to their day without changing anything.
This isn't about who has the cheapest model.
It's about a pattern that repeats every time a valuable capability stops being expensive, and that almost no one names while it's happening:
1. When a resource becomes abundant, it stops being the bottleneck. A year ago the advantage was access. "The company with the best model wins." That sentence means nothing now. Everyone has frontier-level capability for the price of a coffee. The resource everyone competed for just turned into air: it's everywhere and costs nothing.
2. The bottleneck never disappears. It moves to the neighboring resource that's still scarce. When bandwidth became free, what got scarce was attention. When information became free, what got scarce was the judgment to filter it. Now that thinking became almost free, what's scarce is direction: knowing which of the thousand things that cheap intelligence could do is the one that actually matters this week.
3. This flips who wins. Before, the advantage belonged to whoever could pay for the best model. Now anyone can. The advantage belongs to whoever has the clearest answer to a question money doesn't solve: "where do I point this?" One person with sharp judgment now executes better than a funded team with a fuzzy plan and exactly the same models.
Imagine that tomorrow you're handed ten brilliant analysts who work for free, without sleeping, around the clock.
Would you know exactly what to ask them to do on Monday morning?
Because that's what you just received. The question is no longer whether you have the intelligence at your side. You do. The question is whether you know where to point it, and that question has been hard long before artificial intelligence existed.
When a resource gets democratized, a single opportunity doesn't appear. A whole chain does. And the best ones are almost never where everyone is looking.
Millions of people now have access to an intelligence they don't know how to use. They're not missing the model. They're missing knowing what to ask it first. Helping someone go from "I have the tool" to "I know my next step" is a whole business.
The market will fill with things that generate more text, more images, more of everything. What's scarce will be the opposite: something that helps a founder choose the one thing that moves their business this week and ignore the rest.
Plenty of people with clear judgment don't want to touch the infrastructure. Building the service that executes on cheap intelligence, for those who already know what they want, is a layer with immediate demand.
The question that really matters isn't what to do with cheap intelligence. It's: what resource just became abundant in the terrain you know best, and what gets scarce right next to it?
For years, the question was how to get more intelligence. That question just died. Now everyone has an infinite brain at their side, free, waiting for orders. The winner is whoever knows what question to ask it. The bottleneck is no longer thinking. It's knowing about what.