You can now pay for ChatGPT Enterprise out of your AWS commitment, but you negotiate the price blind, your own AWS contract sets how much it retires, and your data stays off AWS. Treat the new listing as a payment route. It is not a hosting option or a residency option, and its main value is the leverage it gives you against both OpenAI and Anthropic at your next renewal.
OpenAI put ChatGPT Enterprise on AWS Marketplace on November 9, 2026, sold by OpenAI and billed by AWS. The AWS listing bundles four products as one offering: Chat, ChatGPT Work, Codex and dots, OpenAI's always-on agents. Eligible consumption counts toward an AWS EDP or PPA commitment "subject to your agreement." There is no price on the page.
What Exactly Did OpenAI List on AWS?
OpenAI listed the full ChatGPT Enterprise seat, including the agent products, as a SaaS subscription you can only buy through a private offer. The listing says pricing is "based on your specific requirements and eligibility" and tells buyers to request a private offer for a quote. It publishes no per-seat price, no seat minimum, no term length and no usage rates.
A private offer is a price and licence that a seller creates for your AWS account alone. AWS's buyer guide defines it as pricing and EULA terms that "aren't publicly available," negotiated with the seller and accepted from the account you designate. The software is the same product you would buy direct.
The division of labour is plain. AWS handles procurement and billing, and OpenAI manages the product experience. OpenAI also offers ChatGPT for Healthcare and ChatGPT for Financial Services through the same marketplace to eligible customers, with eligibility confirmed by an OpenAI or AWS account team under criteria neither company has published.
AWS's pitch, from Matt Yanchyshyn, VP of AWS Marketplace and Partner Services, is that customers "can buy it through their AWS account, govern it alongside the rest of their software spend," per ppc.land's write-up. That benefit is real for procurement and finance. It leaves price, drawdown limits and data location exactly where your contracts already put them.
How Much of Your AWS Commitment Can It Retire?
Your own EDP or PPA amendment decides how much ChatGPT Enterprise spend counts, and AWS does not publish that number. The listing's "subject to your agreement" qualifier means the drawdown share is set contract by contract, as ppc.land notes.
The figure most third-party guides quote is a cap of 25% of the annual commitment. CloudForecast's August 2026 guide says eligible Marketplace purchases are "capped at 25% of the annual commitment," and that they retire commit at full value while you pay the vendor's price, not your EDP-discounted rate. That guide does not cite AWS documentation for the cap, and other consultancies quote different numbers, so read 25% as a common negotiated level and check your own paper.
Two rules from May 2025 matter more than the cap:
- Only "Deployed on AWS" SaaS counts. From May 1, 2025, only SaaS products hosted entirely on AWS and carrying that badge qualify for PPA drawdown. PointFive's write-up of the rule says both the application and control planes must run on AWS. The ChatGPT Enterprise listing shows "Deployed on AWS: Yes."
- Eligibility is judged per invoice. PointFive says eligibility "is determined based on the invoice date, not the contract signing date," and that it "is not currently exposed via APIs or cost reports." A contract signed before the rule changed can still produce invoices after it that retire nothing.
So the practical question for FinOps is narrow. How much Marketplace headroom does your amendment leave after the SaaS you already route through AWS, and will each monthly ChatGPT invoice qualify on the day it lands?
Does Buying Through AWS Keep Your Data on AWS?
No. The listing says so in its own words: buying through AWS Marketplace "does not guarantee" that data stays on AWS infrastructure or in your AWS environment. The listing names OpenAI as the data processor, engaging "subprocessors, including AWS," and the service runs as SaaS managed by OpenAI, outside your account.
OpenAI's data residency and inference residency controls still apply the same way they do for a direct contract. The listing says both work only when enabled, are not available in every region, and do not cover all processing or systems data. Silicon Report's coverage makes the same point: the Marketplace classification as SaaS deployed on AWS does not mean all customer data stays on AWS infrastructure.
That leaves an odd pairing your security team should hear about before procurement signs. The badge that makes the spend eligible is, per the third-party guides above, about where the application and control planes run. The listing's own data language tells you nothing about which regions process your prompts. If your reason for preferring AWS is that your data already lives there, this listing does not give you that. For model access inside your AWS account, the comparison is Amazon Bedrock, which we priced against Azure OpenAI and the OpenAI API in where to buy GPT-6.
The listing also makes no HIPAA statement. ppc.land reports that OpenAI points healthcare buyers to ChatGPT for Healthcare instead.
How Does It Compare With Claude Enterprise on the Same Marketplace?
Claude Enterprise publishes a public price on AWS Marketplace and ChatGPT Enterprise does not, which gives you a reference number to negotiate against. The Claude Enterprise listing, sold by Anthropic and also marked Deployed on AWS, shows a 12-month contract for 20 seats at $4,800, which works out to $20 per seat per month. Additional seats are priced at $0.65753424 per day for the rest of the term, and usage is billed on top at $0.01 per unit "based on Anthropic API rates," a unit the page does not define.
That public number does not match older AWS material. The AWS Marketplace blog post announcing the listing in July 2025 cited $40 per user per month with a 25-seat minimum. Read the current $20 as a base seat plus metered usage, and get a private offer from Anthropic before you treat it as your price.
| ChatGPT Enterprise (AWS listing) | Claude Enterprise (AWS listing) | |
|---|---|---|
| Public price | None; private offer only | $4,800 for 20 seats, 12 months, plus usage |
| Commit drawdown wording | Counts "subject to your agreement" | "Spend draws down directly from your EDP or PPA commitment" |
| Bundle | Chat, ChatGPT Work, Codex, dots | Claude Chat, Claude Code, Cowork, plus Claude Security and Claude Design |
| HIPAA on listing | No statement | "HIPAA-eligible, with a BAA available" |
| Data location on listing | Not guaranteed to stay on AWS | No statement |
The Anthropic listing's drawdown sentence is more confident than OpenAI's, but neither listing can override your EDP amendment. The cap and badge rules above apply to both.
Marketplaces already carry a lot of this category's revenue. Anthropic's amended IPO filing put marketplace sales at 44% of revenue for the nine months to September 30, 2026; we covered the filing's buyer risks in Anthropic's S-1. OpenAI has not disclosed its marketplace share, and neither company disclosed AWS's cut of each sale. Whatever AWS keeps from each sale is built into the private offer you are quoted, disclosed or not.
What Happens at Renewal and Expiry?
Private offers expire, and what happens next is written into AWS's rules, not OpenAI's. AWS's buyer guide says that when an agreement expires you either move to the product's public pricing or lose the subscription. ChatGPT Enterprise has no public pricing on the listing, so plan for the second case and start the renewal offer early.
The same guide says a renewal or upgrade private offer takes effect immediately, and "any previous terms or remaining scheduled payments are cancelled and replaced." Read a mid-term upsell offer (more seats, more dots capacity) as a full contract replacement before you accept it.
Billing control is the other open item. OpenAI added usage analytics and spend controls for ChatGPT and Codex credits in June 2026, but ppc.land notes the announcement does not explain how credit use maps to AWS invoices. That matters more since OpenAI starts billing for dots on November 18. Metered agent spend landing on a central AWS bill is easy to lose sight of.
The strongest case for the listing is real. For a company that is short of its commit and already buying ChatGPT seats, moving that spend onto AWS can turn a shortfall payment into software you were going to buy anyway, with one vendor-onboarding process instead of two. The limits are the cap, the per-invoice badge check and a price you cannot see until you ask.
What to Do Before Your Next AI Seat Renewal
This Week:
- Ask your AWS account team, in writing, what share of your current EDP or PPA can be retired by Marketplace SaaS, how much of that headroom is already used, and whether the ChatGPT Enterprise listing qualifies on each invoice date.
- Pull your current ChatGPT Enterprise or Claude Enterprise contract and note the renewal date, seat count and any committed credits, so the private-offer request has numbers in it.
This Month:
- Request private offers from both OpenAI and Anthropic through AWS Marketplace for the same seat count and term. Use the Claude listing's public $4,800 for 20 seats as the opening reference, and ask OpenAI to quote against it.
- Have your security lead confirm which ChatGPT data residency and inference residency regions you have enabled, and record in the DPIA that the AWS purchase route does not change them.
- Ask OpenAI how credit-based usage for dots and Codex will appear on the AWS invoice, and set a budget alert on that line item before anyone switches dots on for a pilot group.
Before Renewal:
- Compare the AWS private offer with the direct quote at equal terms. If the AWS route is more expensive, the price of drawdown is the difference; decide whether retiring that much commit is worth it.
- Put the expiry behaviour in the contract: a named renewal notice period and pricing protection, because the listing has no public price to fall back to.
The Bottom Line
With both OpenAI and Anthropic now selling their workplace seats on AWS, the AI seat becomes another line on the hyperscaler bill, competing for the same commit as everything else you buy there. We saw the model-API version of this when OpenAI models arrived in Google Cloud's Model Garden and in our Bedrock vs Vertex AI vs Foundry comparison, where the cloud commit you already hold pushed the platform choice. Your Marketplace headroom is finite and shared with every other SaaS tool you route through AWS, so decide which vendor gets it before a sales team decides for you.
Get the drawdown cap in writing before you ask OpenAI for a quote.
Continue Reading
- Azure OpenAI vs OpenAI API vs Bedrock: Where to Buy GPT-6
- Bedrock vs Vertex AI vs Foundry: Your Cloud Commit Decides
- OpenAI Starts Billing for ChatGPT Enterprise Dots on November 18
- LibreChat and Open WebUI offer ChatGPT Enterprise alternatives
- AWS and Azure DMA Duties Land Midway Through a 3-Year Commit
- AI Bundles Push Microsoft, Google and Salesforce to Raise Prices