OpenAI Suspends New Subscriptions for $200 Pro Plan
Thibault Sottiaux, Head of Product at OpenAI, announced that to ensure the experience of existing users and maintain access to Astra, the company is suspending new subscriptions and upgrades to the $200 Pro plan. He stated that this plan puts the most pressure on the system, and the decision was made to make minimal adjustments to preserve the widest coverage. Other plans and APIs will continue as usual, and existing accounts will not be affected as the company works to expand capacity as quickly as possible.
The help center has also stated that, effective immediately, new openings for the ChatGPT Pro $200 plan (Pro 20X) are suspended, as well as upgrades from free, Go, Plus, or the $100 Pro plan to the $200 plan. Existing $200 subscribers will continue to renew as usual; new subscriptions and existing accounts for the $100 Pro plan are unaffected. If a cancellation or downgrade is requested, rights will be retained until the end of the current billing period, and can be retracted within that period. The duration of the suspension and the daily new subscription scale have not been disclosed.
On the previous day, Sottiaux had already warned that the demand for Astra is "unprecedented"; even though there has been steep growth, it has not reached a similar scale, prioritizing the quality of existing services. If this continues, new Pro subscriptions will remain suspended. TechCrunch and others attribute this to the strain Astra puts on infrastructure. The Pro 20X is described as approximately 20 times the usage of Plus, and does not have the same 5-hour limit as some other plans, resulting in the highest consumption per user.
The company has not shut down the API or lower-tier plans, effectively reserving scarce computing power for existing $200 heavy users and pay-as-you-go API calls, blocking new subscriptions at the highest tier. Codex has recently been raising limits and resetting quotas for paid users multiple times, indicating that capacity tightness is a recent phenomenon rather than a one-time failure. The help documentation lists that Pro still includes advanced models, Codex, deep research, live images, memory, and upload capabilities; what is suspended is the entry point, not the existing functionality list.
During the same period, OpenAI's developer channel launched the GPT-Live-1 voice interface, but Sottiaux clearly categorized the API as "still available." The $200 plan has been singled out, indicating that the bottleneck lies in high-multiplicity text and proxy usage, rather than a complete halt on sales.
In market mechanics, this is a supply quota, not a price reduction promotion. The buyers are new users wanting to purchase the highest tier Astra quota, who are temporarily locked out; the sellers are OpenAI, prioritizing computing time for existing Pro and API calls. Funds are stalled on new $200 orders, shifting towards the $100 plan, Plus, and token-based APIs. Beneficiaries are the existing $200 users and API customers who can still expand; those under pressure are individuals and small teams planning to upgrade to the 20x quota. The event is driven by peak demand for Astra, and the pace of expansion will determine when the suspension ends.
On a supplementary note, once canceled, users cannot upgrade back to the $200 plan during the suspension period. Service restoration depends on new capacity, and the official date has not been provided.
Source: Public Information
ABAB AI Insight
OpenAI did not first cut free or Plus, but rather suspended new subscriptions at the highest price point. The logic is that the $200 plan has a volume approximately 20 times that of Plus, and a single user occupies far more of the Astra cluster than their customer acquisition value; shutting off the new entry point causes less harm to existing paying users than a complete throttling. Sottiaux's warning the day before and the implementation the next day indicate that the warning window was only one day, and the cluster capacity had already thinned to the point where it could not accommodate the 20x tier. This is similar to how Microsoft and Google manage cloud resources with queuing and regional quotas, just with the focus shifting from cabinets to model tiers.
The capital path is "existing users prioritized over incremental revenue." Losing one $200 subscription is exchanged for preventing existing high-value users from dropping off and preventing the API from being crowded out by web Pro. Expansion is still about buying chips and data centers, but the delivery speed is not keeping up with the slope after Astra's release. Money continues to flow into the API and the $100 plan, while the $200 plan is temporarily taken off the shelf. The motivation is not to abandon high customer value, but to avoid the 20x usage overwhelming the entire network experience, triggering larger-scale cancellations.
This is comparable to the queuing for Plus when ChatGPT-4 first launched, and the limited purchases for AAA game releases: both use the suspension of the highest configuration to protect online quality. The industry phase has shifted from expanding the subscription funnel to controlling concurrency at the most expensive tier. Whoever can first expand Astra's capacity will be able to reopen the $200 faucet; until then, the model release itself is a purchase limit.
The structural change represents a transfer of pricing power. The mechanism is: the price of high-tier subscriptions buys a multiplier, which directly translates into watts; when the watts are insufficient, the highest-priced new subscriptions are the first to be suspended, shifting the market from "who is willing to pay $200" to "who is already online." The API remains open on a pay-as-you-go basis because its consumption can be constrained by billing, while the web 20x tier cannot.
ABAB News · Cognitive Law
The most expensive tier is suspended first because the multiplier consumes watts first.
When the demand curve for new products is steeper than the expansion curve, the shelf moves before the price.
Preserving existing user experience buys time for the next round of expansion.