OpenAI paused $200 Pro signups — your AI tier can sell out

On September 10, OpenAI stopped selling its $200/month ChatGPT Pro plan. Not deprecated, not repriced — you simply cannot buy one right now. Demand for GPT-6 Astra outran capacity, and Pro seats are the heaviest load per user, so Pro is the lever they pulled. Existing subscribers keep working. Everyone else gets a closed door and no reopening date. If you were planning to put three more people on Pro this month, that plan is now a vendor capacity problem, not a budget line.

What actually happened

Per TechCrunch, OpenAI paused new signups and upgrades to the Pro tier while leaving the API, Go, Plus, Business and Enterprise plans available. OpenAI's Thibault Sottiaux described the reasoning directly: Pro seats "put the most strain on our systems and we wanted to take the smallest step that allows us to continue giving the broadest access possible." He had earlier called demand for Astra "really unprecedented."

No restart date was given. The company says it is scaling capacity.

Note what stayed open: the API. Consumer seats got rationed; programmatic access did not. That is not an accident, and it is the most actionable sentence in this story.

Why vendor capacity matters for your business

This is the second time in about eight weeks that a frontier vendor has rationed access by closing the front door rather than raising the price — Moonshot paused Kimi K3 subscriptions roughly 48 hours after launch in July when it ran out of GPUs. Treat it as a pattern, not a fluke. The newest, highest-tier, most compute-hungry product is the first thing to become unbuyable.

The practical split: seat-based dependencies and API dependencies fail differently.

A seat-based dependency fails at hiring. Your process works fine for the six people who already have logins, and then you onboard a seventh and discover the process does not exist for them. That failure is invisible until the exact moment it is expensive.

An API dependency fails at the request, which is loud, instrumented, and something you can write a fallback for. Rate limits and 503s are engineering problems. "You cannot purchase this product" is not.

So take twenty minutes and write down every AI tool your operation actually depends on, and mark each one S or A. For anything marked S that sits in a revenue path — quoting, support triage, content production — either move it behind an API you control or accept in writing that you cannot scale that process on demand. Both are fine answers. Not knowing which one you picked is not.

And the general shape holds regardless of vendor: your AI supplier is capacity-constrained, is allocating that capacity by tier, and is not going to consult you about it. Same conclusion we reached during the ChatGPT outage that took Codex with it — build the second route while the first one is working.

Sources: TechCrunch: OpenAI puts Pro subscriptions on hold due to Astra demand, Computerworld: OpenAI pauses $200 Pro tier as Astra demand strains capacity.