The GPT Store gives builders reach, not the customer relationship

OpenAI opens a store for 3 million custom GPTs and Microsoft sells Copilot from one seat. Reach gets cheap; owning the customer gets more valuable.
OpenAI now has a shelf for more than 3 million custom GPTs, the versions of ChatGPT that users build without code, all created in the two months since it introduced them. The GPT Store opened on 10 January to paying ChatGPT subscribers only, with a revenue program for builders promised for the first quarter: US first, paid by user engagement, criteria still to come.
The store hands builders reach and keeps everything else. Pricing, ranking and customer data stay with the platform, and a payout by engagement inside someone else's store is closer to rent than to a customer relationship. That widens the gap investors should price between companies that own the customer and companies that rent it. Revenue built on a marketplace listing deserves a discount for platform risk; revenue from direct contracts with businesses does not.
The same week fixed the price of the category. On launch day OpenAI introduced ChatGPT Team for smaller organisations at $25 per user per month billed annually, or $30 monthly, with a commitment not to train on their data. Five days later Microsoft removed the 300-seat minimum for Copilot for Microsoft 365, now sold from a single seat at $30 per user per month, and launched Copilot Pro for individuals at $20 a month. The two largest sellers of AI assistants have set a public anchor of $20 to $30 per seat and opened their own distribution channels. A thin product that adds a prompt and an interface to a general model now competes with a free listing inside the assistant the customer already pays for.
Used well, the store is a cheap demand test rather than a business. It shows which workflows people come back to. The company worth building is what the general assistant cannot do: a product wired into the customer's systems, carrying domain data and priced against the value of the outcome rather than against a $20 seat. The end of the seat minimum helps here as well. Small and mid-sized companies now buy AI from a budget line, and vertical products that fit that line with clear security terms, admin controls and measurable time saved will meet buyers who have already approved the category.
The case for building inside the store rests on terms OpenAI has not yet published. If the builder program lets a GPT earn enough to sustain a team, and if builders win the right to charge users directly instead of seeing all payment stay inside OpenAI's own plans, the listing becomes a channel worth a company. Copilot seat adoption among small businesses is the other open question, the first real test of whether a $30 add-on pays back for companies without enterprise budgets.
Until those answers arrive, platform stores make reach cheap and margins thin for everyone who stays inside them. What Fanaura looks for is an AI-native company that uses such channels for discovery and keeps the contract, the data and the price for itself, a core test in our thesis. The listing can be rented. The customer has to be owned.
Building an AI-native company? Apply in five minutes. →
More from Fanaura
Building a vertical AI-native product that owns its customers and prices on outcomes? Send us the project.
Fanaura invests its own capital and the capital of its limited partners (LPs), usually through a separate SPV for each deal. If we see a fit, we get in touch.
Get the news by email
One short note each weekday. No marketing, and you can unsubscribe at any time.