In a help article updated in late October 2026, OpenAI explained that it may restrict new ChatGPT subscriptions when too many paid subscriptions are linked to the same payment card or device. There is also a cap on the number of paid accounts that can be used on a single device, and the company describes these measures as protecting "privacy and security." The limits apply to the Go, Plus and Pro plans; Free and business workspaces do not count toward the subscription limits. For users who run short of usage, OpenAI advises buying credits on the account they are already using or checking when their usage limits reset. It has not disclosed the specific caps or an effective date. Comparing the related official documents shows how OpenAI is building out its consumer billing structure and how it intends to absorb demand for additional usage.
No numbers or effective date: caps tied to payment cards and devices
According to OpenAI's help article, the limits apply in two situations: when signing up and after subscribing. At sign-up, if too many paid subscriptions are linked to the same payment card or device, the checkout screen shows an error to the effect that the purchase cannot be completed because there are too many subscriptions associated with the card or device (original: "Too many subscriptions associated with this card or device. We can't complete this purchase"). OpenAI states explicitly that this differs from a payment decline by a bank or card issuer. After subscribing, the number of paid accounts that can be used on one device is limited, and once the cap is reached, the affected accounts can no longer be used on that device. Even an account that worked normally at the time of purchase may be signed out following a later check.
In other words, a new subscription rejected at checkout points to a limit on subscriptions tied to the card or device, while a warning at sign-in or being signed out during use points to the per-device limit. These are separate from the message and token limits that apply to everyday ChatGPT use. OpenAI explains the difference in the article's FAQ: ordinary usage limits define how much can be used under a subscribed plan, whereas the card and device limits can result in a new subscription being refused outright.
The five plans covered are ChatGPT Go, Plus, Pro 100, Pro 200 and Pro 500. Free and the Business, Enterprise and Edu workspaces are not counted. ChatGPT also lets users sign in to up to two accounts at once, but this is independent of the per-device subscription limit. Signing out of one account therefore does not necessarily make another paid account usable. A second account can itself be created for free, starting on the Free plan.
Families who share a payment card may also be affected. The help article says that when several people subscribe to paid plans with the same card, the limit can be reached even if each person uses a separate device. Likewise, where devices are shared at home or in public facilities, the limit could be reached simply by each user signing in to their own account. Subscriptions purchased before the limits were introduced may continue as existing subscriptions, but there is no guarantee that a user who cancels can switch to another subscription.
OpenAI has not disclosed concrete figures, such as how many subscriptions one card can support or how many paid accounts one device can use. It also gives no effective date for the limits and no explanation of how devices are identified. The covered plans and the existence of the limits are stated clearly, but the conditions under which a cap is actually reached remain opaque. More detail may appear later in the help article or in the app, but the current explanation does not let users check how many more subscriptions their card or device can support before they sign up.
From Go to Pro 500: different ways to add usage depending on the plan
Go, at $8 a month in the US, and Plus, at $20, offer different options for adding usage once the allowance runs out. Organizing the descriptions in OpenAI's pricing page and five related help articles by plan gives the following picture.
| Plan | Monthly price | Card/device limit | Ads | Usage credits | Paid instant reset | Using plan allowance in partner apps |
|---|---|---|---|---|---|---|
| Free | $0 | Not covered | May be shown | Available to some users only | Not available | Not available |
| Go | $8 (US) | Covered | May be shown | Available to some users only | Not available | Available |
| Plus | $20 | Covered | None | Available | Available | Available |
| Pro 100/200/500 | $100/$200/$500 | Covered | None | Available | Available | Available |
| Business, Enterprise, Edu | Separate business pricing | Not covered | None | Separate system for workspaces | Not available | Not available |
Availability of each feature may vary by account and region. Go, the cheapest paid consumer plan, is also subject to the card and device limits; ads may be shown, and it does not support paid instant resets. Because credit purchases are limited to some users, it offers fewer choices than Plus and above when the allowance runs out.
Pro has additional ways to secure more usage. The top-tier fast mode, "Ultrafast," is exclusive to Pro 500, and Pro 100, Pro 200 and Pro 500 come with monthly API credits worth $100, $200 and $500 respectively. However, according to the pricing page, using the API credits requires activation, unused credits expire at the end of the billing period, and the terms may change or end. Pro is available only as a monthly subscription. At DevDay 2026, OpenAI also announced that Plus and Pro allowances can be used across 16 partner tools.
This pricing structure shows that Plus and above, while subject to the card and device subscription limits, can buy additional credits and instant usage resets within the same account. Go, by contrast, is subject to the subscription limits yet has limited ways to add usage, and ads may be shown. Taking out several cheap paid accounts to expand usage is restricted; users who need more must buy on their current account or move to a higher plan. OpenAI has not explained the specific aims behind this pricing structure or the subscription limits.
Paid resets pull the weekly allowance forward; fast modes burn usage up to 8x faster
A paid instant reset is available on individual Plus and Pro accounts, and buying one restores the ChatGPT Work and Codex allowance at that moment. The help article says both the 5-hour and weekly windows are reset, though OpenAI's pricing document says Pro has no 5-hour limit in the first place. The key point is that an instant reset does not add a new entitlement separate from the normal allowance. OpenAI says it "pulls your normal weekly allowance forward rather than adding a separate usage entitlement."
The reset takes effect at the moment of purchase, and even if weekly allowance remains, it is not carried over; the account switches to a fresh allowance. The new weekly cycle begins when the first request is sent to Work or Codex after the reset, and the next automatic reset comes seven days after that. The next reset therefore does not necessarily occur exactly seven days after purchase. Resets cannot be saved or scheduled for a specific time, and as a rule they are not refundable. The purchase price appears at checkout, but the help article gives no specific amount.
Usage credits, by contrast, are a pay-as-you-go balance consumed after the allowance included in the plan has been used up. They can be used in Codex, ChatGPT Work, Word, Excel and PowerPoint, and are valid for 12 months from the purchase date. As a rule they are neither refundable nor transferable to other users, but they support "auto top-up," which automatically buys more when the balance falls below a set amount, and a monthly purchase cap can also be set. However, when multiple tasks run at the same time, the credit balance may temporarily go negative.
The pace at which allowances and credits are consumed also depends on the speed mode selected. In the fast modes for ChatGPT Work and Codex, with Standard as the baseline, the plan's included allowance is consumed 2.5 times as fast in Fast and 8 times as fast in Ultrafast. For purchased credits, the multipliers are set at 2x for Fast and 6x for Ultrafast. Compared with purchased credits, then, the plan's included allowance is consumed 1.25 times as fast in Fast and about 1.33 times as fast in Ultrafast.
These ratios come from dividing 2.5 by 2 for Fast and 8 by 6 for Ultrafast, the latter rounded to the second decimal place. Ultrafast is limited to GPT-6 Astra and GPT-6.1 Sol, and OpenAI cautions that these figures indicate the pace of allowance consumption and do not guarantee faster processing. Because the conversion ratio between one unit of allowance and one unit of credit has not been published, these multipliers alone cannot show which billing method is more economical.
These mechanisms show that OpenAI has built fine-grained options for adjusting usage within the same account. An instant reset pulls the normal weekly allowance forward and auto top-up replenishes the credit balance; both are ways to keep working without taking out a new paid account. Changing the speed mode differs from an additional purchase in that it adjusts the pace at which allowance or credits are consumed. The help article on card and device limits also states that credits are not meant for combining multiple subscriptions or moving usage between accounts.
900 million weekly users, 50 million paying subscribers, and ad revenue
When it announced $110 billion in new funding on February 27, 2026, OpenAI said ChatGPT had more than 900 million weekly active users, more than 50 million paying individual subscribers and more than 9 million paying business users. Simply dividing 50 million individual subscribers by 900 million weekly active users gives roughly 5.6%. However, both figures are published lower bounds, and not every paying subscriber is necessarily among the weekly active users, so this is not an actual paid conversion rate. The calculation also excludes the more than 9 million paying business users. The DevDay 2026 summary on September 29 said weekly users had reached 1.2 billion, but no subscriber count for that point has been published.
As a new way to turn this large user base into revenue, OpenAI has introduced ads on Free and Go. Ads may appear on both plans but not on Plus, Pro, Business, Enterprise or Edu. According to the help article, the US ad test began on February 9, and ads are not shown to accounts identified as belonging to users under 18. Availability is expanding region by region.
The relationship between ads and usage allowance also shows up in the ad-free setting available on the Free plan. Switching to ad-free removes ads but lowers the message limit and disables some features, such as image generation and deep research. The settings screen also explains that users can do more when ads are shown, while ad-free use limits the number of messages. In other words, on Free, ad revenue helps support part of the features and usage offered at no charge.
In its January ads announcement, OpenAI said it would expand Go to all regions where ChatGPT is available, including the US, offer it at $8 a month in the US, and test ads on Free and Go. It stated as basic principles that ads would not influence answers and that conversation privacy would be protected, and said it intends to diversify its revenue sources. According to the company, the ads business reached $1 billion in annualized revenue in under 200 days from launch and has expanded to more than 40 countries. Annualized revenue, however, converts recent revenue into a one-year figure and differs from confirmed annual revenue. OpenAI positions ads as a revenue source alongside consumer subscriptions, enterprise products and the usage-based API, and says the ad-supported free tier lets more than 1 billion weekly active users use ChatGPT.
Figures for OpenAI's overall revenue vary by news outlet. Reuters reported on September 29 that annualized revenue was approaching $70 billion, while Bloomberg reported on October 8 a figure of about $50 billion. According to Bloomberg, the Financial Times (FT) reported that the roughly $20 billion gap stems from OpenAI's investors adjusting the way revenue through cloud providers is recorded, among other things, to make comparison with Anthropic easier. Neither figure was officially announced by OpenAI; both are based on accounts from people familiar with the matter. Reuters also reported in the same article that enterprise revenue has more than doubled since July and that consumer revenue in the third quarter of 2026 exceeded the full-year 2025 total, but the share of revenue from ads and additional credits has not been disclosed.
From ads to December's shared token allowance: how billing changed in 2026
Organizing OpenAI's 2026 changes to consumer billing in chronological order places the card and device subscription limits in the latter part of a series of changes.
| Date | Main events |
|---|---|
| January | Announced ad policy. Expanded Go to all regions and offered it at $8 a month in the US. Said it would test ads on Free and Go |
| February 9 | Began US ad test |
| February 27 | Announced $110 billion in new funding. Disclosed more than 900 million weekly active users and more than 50 million paying individual subscribers |
| May | Introduced Ads Manager, letting advertisers place ads themselves |
| September 29 | Announced Pro 500 at DevDay 2026. Announced a reduction in Pro 200 usage allowance by email the same day. Reuters reported annualized revenue approaching $70 billion (per sources) |
| October 8 | Bloomberg reported annualized revenue of about $50 billion (per sources) |
| Late October | Reversed the Pro 200 allowance reduction. Explained per-card and per-device subscription limits in help article |
| December (planned) | Plans to introduce a method that combines ChatGPT and Codex allowances and measures them in tokens for Plus, Pro and Business |
OpenAI began testing ads in the US in February and, on September 29, announced a cut to the Pro 200 allowance alongside the Pro 500 announcement. The Pro 200 reduction was reversed in about a month, however, and subscribers who had been subject to the lower allowance saw it doubled. The Work and Codex allowances for Pro 100 and Pro 500 were also doubled, while Plus's allowance was left unchanged. After these changes, the help article explained the per-card and per-device subscription limits in late October, and a new method combining ChatGPT and Codex allowances is planned for December.
Under the shared allowance starting in December, ChatGPT and Codex usage will be measured in tokens, and consumption will vary with the model chosen and the work performed. More advanced models and more complex processing consume more tokens, while ordinary chat with the Instant model does not consume allowance. When the allowance limit is reached, OpenAI points to three options: wait for the next reset, change to a higher plan, or buy additional credits. If ChatGPT and Codex usage is unified under a common token measure, usage purchased within the same account should become easier to track on a common basis.
OpenAI describes the card and device subscription limits as privacy and security protections, and when usage runs short it points to buying credits or instant resets, moving to a higher plan, or waiting for the next automatic reset. Taken together, these mechanisms suggest that taking out several paid accounts to expand usage is restricted, and that additional usage is centered on purchases and plan changes within a single account. Free offers a way to get more usage in exchange for seeing ads, but for Go, where ads may be shown, no explanation clearly states an exchange between ads and usage.
This view could change, however, if OpenAI later formally shows a way to use multiple paid subscriptions together to increase usage. For now, two points deserve attention: whether the conversion ratio between the shared token allowance starting in December and purchased credits will be published, and whether the per-card and per-device caps and their effective date will be disclosed. The former would make it easier to compare the usage included in monthly plans with pay-as-you-go costs on the same basis. The latter would let users judge before subscribing how many additional subscriptions their card or device can support.