OpenAI's Marketplace Turns Your AI Commit Into Budget for HubSpot and Salesforce
OpenAI will let enterprises spend part of their AI commitment on partner software, the same day Axios put its run rate near $70 billion. Listed vendors skip the new purchase order.
September 30, 2026
OpenAI opened a marketplace at its developer conference on September 29 with 32 launch partners. The feature that matters to GTM teams is in the fine print. According to Decrypt's DevDay recap, eligible enterprise customers can apply part of their existing OpenAI spending commitment to software from approved partners. The partner list includes Adobe, Figma, Salesforce, HubSpot, ServiceNow, Harvey, Legora, Palo Alto Networks and CrowdStrike. HubSpot co-founder Dharmesh Shah posted that HubSpot is a launch partner and called ChatGPT a super app people will use to reach systems like HubSpot. Runway, the video-model company, told customers they can now put part of their OpenAI commitment toward Runway purchases.
The same day, Axios reported that OpenAI's annualized revenue is nearing $70 billion. That’s up more than 70% since the start of the third quarter, and business revenue has more than doubled since July. The Decoder, citing the Axios scoop and OpenAI's own DevDay figures, adds 2.5 million businesses and more than 35 million weekly Codex and ChatGPT Work users.
OpenAI is trying to capture enterprise budget, first by growing it and then by making it spendable elsewhere. A revenue number and a product launch are different kinds of news. Here they make the same argument.
On October 8, TechCrunch relayed a Financial Times report that corrects the figure above: OpenAI told investors its annualized revenue is approaching $50 billion, about $20 billion below the Axios number. Per the FT, the higher figure came from investors trying to compare OpenAI with Anthropic, which counts sales made through its cloud partners while OpenAI doesn't. The budget-pool argument rests on enterprise commitments growing faster than model spend, and the FT report doesn't touch business revenue, so the mechanism holds even as the headline number shrinks, and what that gap means for how you report run rate gets its own piece. OpenAI's chief financial officer, Sarah Friar, confirmed the figure on October 9: an annualized revenue run rate of about $50 billion at the end of September, up 77% in three months.
The commit becomes a budget pool
The mechanics come from cloud. AWS has let customers count AWS Marketplace purchases against their enterprise discount commitment for a long time. Sellers cite a cap of about 25% of the annual commit, as Suger's guide for marketplace sellers explains. That rule turned AWS Marketplace into a procurement shortcut. A CIO who had already signed a large AWS commitment could buy Datadog or Snowflake through it without a new purchase order or new vendor review. There was no fight with finance, because the money was already approved. Software sellers listed there because their deals closed faster, and many now treat a marketplace listing as a standard part of the enterprise sales playbook.
OpenAI is running the same play a decade into the cloud era, on a much shorter clock. AWS needed years of cloud spend before its commitments were big enough to act as a budget. OpenAI's commitments are big enough now, which is what the Axios number shows. When business revenue doubles in a quarter, customers are likely signing larger commitments than they can spend on models alone. A marketplace gives that surplus somewhere to go.
For SaaS sellers, that splits the market in two. Partners get deal velocity, and HubSpot saw it early. Shah noted HubSpot was the first CRM to ship a native ChatGPT connector, more than a year before this program existed. When a HubSpot or Salesforce rep sells into an OpenAI customer, part of the price can come out of money the buyer already committed. The deal skips the new-budget conversation. Sellers outside the program compete against pre-approved spend. A procurement team choosing between a partner tool drawn from an existing commit and a rival that needs a fresh purchase order will rarely pick the fresh purchase order, even at a better price. Finance sees one line drawing down against a contract it already approved, which is about as easy as software approvals get.
Partners aren't limited to applications either. Philip Kiely at Baseten, one of the launch partners, posted that enterprise teams can now run open models in Codex and count that spend against their OpenAI commit. OpenAI is letting its commitment pay for models that compete with its own. Owning the budget seems to matter more to OpenAI than owning every token.
This is the lock-in Anthropic has to answer
Anthropic is the obvious target. Parameter's summary of the two labs' revenue race cites preliminary filing that puts Anthropic’s run rate at around $65 billion as of late July. The two companies are close in size and competing for the same enterprise logos. Anthropic has built distribution of its own. Our earlier piece, "Apollo, G2, and Zuora Ship Agents as Anthropic Opens Its Connector Marketplace," covered its directory of more than 2,000 Claude connectors. But a connector directory tells a buyer what Claude can reach. OpenAI's marketplace tells the buyer's CFO where committed money can go, and that's harder to walk away from at renewal.
The capital behind the move is getting larger too. OpenAI is reportedly seeking at least $30 billion at about a $1.4 trillion valuation, according to TechCrunch's summary of Bloomberg's reporting. The money is a bridge to a public listing now expected in 2027. A company raising at that price needs enterprise commitments to grow. Making those commitments more useful is a direct way to grow them.
The strongest objection is OpenAI's own history. Shah acknowledged in his post that OpenAI has tried marketplaces before without them taking hold, and he's betting this version is different. The economics are also unknown. Every's DevDay review describes the program as a beta catalog of approved partner software. None of the coverage we read discloses a minimum commitment, a cap on the share that can be applied or a revenue share. If the cap turns out to be small, the budget pool is a puddle. If partners have to discount to be listed, the deal velocity costs them margin. AWS Marketplace sellers already make that trade through listing fees.
These objections limit how much this matters, and they don’t reverse it. A capped pool still tilts procurement toward partners at the margin. OpenAI’s earlier stores sold to consumers and developers with no commitment behind them, while this one is tied to a contract the CFO already signed.
What this means for enterprise GTM operations leaders
What a GTM leader does with this depends on which side of the list they're on. If you sell software to enterprises, find out which of your top 50 accounts carry an OpenAI commitment. Then decide whether to apply to the marketplace before your competitor does. If you buy software, ask your OpenAI account team for the cap and the list of eligible partners before your next renewal. And treat that list as a price input in every tool evaluation that follows.
My bet is that Anthropic answers with its own commitment-portability program within two quarters. A connector directory can't compete with money that's already been approved.
OpenAI's marketplace tells the buyer's CFO where committed money can go, and that's harder to walk away from at renewal.
Before your next OpenAI renewal, ask your account team for the cap and the list of eligible partners, and treat that list as a price input in every tool evaluation that follows.
