OpenAI now sells ChatGPT Enterprise through AWS Marketplace, so a company can buy it on its existing AWS account and apply eligible spend toward its AWS commitments. A day later Hugging Face let organizations move their Hub subscription and compute charges onto a Google Cloud bill. Two AI vendors in two days chose the hyperscaler invoice as the way into the enterprise, and the procurement, finance and governance teams who sign for AI now have a different set of questions to ask.
What OpenAI announced
In a November 9 post, OpenAI said the AWS Marketplace listing gives organizations a way to discover, buy and manage ChatGPT Enterprise using their existing AWS purchasing and billing relationship, and to apply eligible spend toward their AWS commitments. The package covers ChatGPT, ChatGPT Work, Codex and dots. OpenAI says AWS customers get the same models, features and functionality available directly from OpenAI, with new features released at the same time.
The company also says eligible customers can buy ChatGPT for Healthcare and ChatGPT for Financial Services the same way. It positions the listing next to the OpenAI models already offered in Amazon Bedrock: developers build applications and agents on the models through Bedrock, while business teams use ChatGPT Enterprise for daily work. Customers are told to contact their OpenAI or AWS account team to confirm eligibility.
Matt Yanchyshyn, vice president of AWS Marketplace and Partner Services, said in the post that customers “can buy it through their AWS account, govern it alongside the rest of their software spend, and move from evaluation to rollout faster.”
Hugging Face made the same move on Google Cloud
On November 10, Hugging Face said organizations can link their Hub organization to a Google Cloud account so that Hugging Face usage appears on the Google Cloud invoice instead of a corporate credit card. The scope is wide. It includes upgrades to a Team or Enterprise subscription, pay-as-you-go compute (Jobs, Inference Endpoints, Inference Providers, GPU Spaces and ZeroGPU quota extensions) and additional storage, all at public Hugging Face pricing. Hugging Face says purchases go through the existing Google Cloud billing account and procurement process, with no new vendor to onboard. Only an admin of the Hugging Face organization can link it.
The two announcements differ in one detail a finance team will notice. OpenAI’s post names commitment drawdown directly: eligible spend can count toward AWS commitments. Hugging Face’s post talks about one bill and no new vendor, and does not say whether its charges count toward a Google Cloud commitment. That is a question for the Google Cloud account team before anyone assumes it.
This is the second route for frontier models through the cloud providers’ own catalogues that AITech has covered in three weeks. In October, Google Cloud put OpenAI’s GPT-6 models in Model Garden in private preview, for developers who build on the API. The AWS Marketplace listing extends the same idea to the seat-based product that business teams log into.
What it means for the buying committee
Commitment drawdown changes the comparison
For a company that has signed an AWS spend commitment, an AI subscription that counts toward it draws on money already promised to AWS, and a CFO may see the purchase as using a prepaid balance. That can move a ChatGPT Enterprise rollout ahead of a rival tool sold only on direct paper, even when the rival scores better in an evaluation. Buying committees should score the product first and look at the billing route second, and write down which order they used.
OpenAI’s wording is “eligible spend”. Its post does not define which ChatGPT Enterprise charges qualify, and it sends customers to the account teams for that answer. Finance should get the eligibility rule in writing, with the share of the contract value that counts, before the drawdown is used to justify the deal.
Contract terms still need a line-by-line check
OpenAI says the product, models and release timing match what it sells directly. Its post does not say whether price, data retention, support level or usage allowances are identical on the two routes. Ask for the marketplace terms and the direct order form and compare them. OpenAI is also changing how usage is metered: from December it will pool ChatGPT and Codex usage into one allowance, so check how that pooled allowance appears in the marketplace terms.
A single invoice line hides which team used the tokens
Moving AI charges onto the cloud bill puts them in one place, but a single line item does not show which team or workload used the tokens. AWS published a reference build on November 10 that shows the level of detail cost owners will want. It collects Amazon Bedrock model invocation logs from every linked account into one central account, stores them as Parquet in Amazon S3 and exposes them in Amazon Athena. Each record carries input and output token counts, the model, the stop reason, the account, the timestamp and the IAM role that sent the prompt. Joining that table to the AWS Cost and Usage Report 2.0 on the IAM ARN gives cost per team or account per day.
The AWS post also says tokenomics, the management of AI token spend, was announced as a formal FinOps domain at the 2026 FinOps Foundation conference. It warns that prompts collected this way may contain personally identifiable information, and that IAM and S3 bucket policies need review before the data lake goes live. That caveat applies to any central store of prompts, whichever vendor’s models produced them.
Token prices are also moving fast. Ramp’s index showed the blended OpenAI and Anthropic token price fell 26.8% in four weeks. A commitment sized against this year’s prices can buy far more usage next year, or run out early if usage climbs faster than prices fall.
How to evaluate a marketplace AI purchase
Start with eligibility. Get the vendor and the cloud account team to state in writing which charges count toward the commitment and at what rate.
Next, compare the paper. Put the marketplace terms next to the direct order form and check price, data retention, support, usage allowances and renewal terms.
Then check who can buy. Hugging Face lets any admin of the Hugging Face organization link billing, and that admin may sit outside procurement. Procurement should decide whether marketplace subscriptions for AI tools need the same approval as a direct contract.
Finally, plan the cost attribution before the rollout. If the bill will arrive as one line, decide now how usage will be tied to teams, whether through per-team accounts, invocation logs joined to the cost report, or the vendor’s own admin analytics.
Source: OpenAI, ChatGPT Enterprise now available in AWS Marketplace. Additional sources: Hugging Face, Pay for Hugging Face Hub with your Google Cloud Account; AWS Machine Learning Blog, Amazon Bedrock tokenomics data lake.
