Two announcements came out of Oracle's world on Tuesday. The one that got covered: a record fourth quarter, $19.2 billion in revenue and remaining performance obligations of $638 billion, up 363% in a year. The one that barely got covered ran four paragraphs on OpenAI's blog: Oracle customers will soon be able to pay for OpenAI's frontier models and Codex with their Oracle Universal Credits.
Read the last line of that post. “Contact your Oracle sales representative for details, timing, and availability.” The most coveted product in software ends its own announcement by sending buyers to another company's rep. If you sell for a living, read it twice.
Money inside the commit is already spent
I sold at Oracle, and later at Fastly, where revenue lived and died by committed-use contracts. A commit changes the physics of every purchase near it. Money inside the commit is already spent in the CFO's head. Money outside the commit is a new ask, with a fresh approval chain, a fresh security review, and a fresh chance for procurement to say no. At Fastly, an expansion that drew down an existing commitment could close in weeks while a similar-sized net-new deal took quarters. Same product both times. The difference was which bucket the money came from.
That's the lens for the OpenAI note. Enterprises have pre-paid for cloud on a scale that's hard to hold in your head: commitments to the big clouds crossed $500 billion last year. The marketplaces that let companies burn those commitments on third-party software were a $16 billion channel in 2023, and Canalys expects $85 billion by 2028. Tackle, which processes these transactions for software vendors, has stood by its prediction that $100 billion flows through this year. Among AWS's top 1,000 customers, 99% already have at least one active marketplace subscription. Snowflake has moved $7 billion through AWS Marketplace alone. Salesforce passed $5 billion.
AI made this the main event, because AI budgets are the least settled line item in the building. When nobody is sure which budget a new AI tool should come from, the path of least resistance is the budget that already exists. OpenAI signed a $300 billion compute contract with Oracle last September; this week's announcement closes the circuit. OpenAI buys Oracle's capacity, Oracle's customers buy OpenAI's models, and everyone's quota retires in the same currency.
Oracle's earnings whisper the same thing
One stat from Tuesday didn't make anyone's headline. Oracle grew revenue 21% last quarter while sales and marketing expense fell 10%. Read those two together. When $638 billion of future revenue is already under contract, you need fewer people creating demand and more people managing drawdown. Growth up, selling cost down is what a commit-based go-to-market looks like at scale, and every software board that saw Tuesday's print will want to know why their own company can't have that ratio.
Two things can be true
For buyers, this is mostly good. One bill, pre-cleared procurement, governance the risk committee already blessed. Forrester pegs the procurement time savings at 60%. And the convenience carries a cost that won't show up for a few quarters. A commit was never free money. Your company promised a cloud provider a set amount of spend, in exchange for better pricing, based on a forecast somebody made years ago. Once that promise is signed, the pressure runs one direction: use it all. So when two tools are on the table and only one can be paid for out of the commitment, the lesser tool can win without anyone arguing for it. Nobody says “let's buy the worse one” in a meeting. They say “this one counts against the cloud spend we already owe,” and the meeting moves on. Each purchase routed this way also makes an eventual exit harder, because by the time the contract comes up for renewal, the provider can read your dependence right off its own invoices.
The read
If you carry a number, three changes are worth making before the quarter ends.
Add one discovery question: where does the money for this live? Net-new budget, or inside a cloud commit, and which one? The answer tells you who your real competitor is, and a lot of the time the competitor is the drawdown itself, a vaguely comparable tool that happens to take your buyer's credits.
Get transactable. If your product can't be bought with money your buyer already committed, you're the expensive option at any price. Private offers on the marketplaces your accounts use are table stakes now, and “the AWS listing is on the roadmap” is becoming the new “SOC 2 in progress.”
Make friends with the cloud rep. Somewhere at Amazon, Microsoft, Google, or Oracle, a salesperson owns your buyer's cloud contract, and your deal counts toward that person's number when it runs through their marketplace. That person is inside the account every week and knows where the unspent money is. They also get paid when you win. Bring them in early and you've added a seller to your team for free. Leave them out and they'll point that budget at a competitor who brought them in.
The longer I sit with this announcement, the funnier the ending gets. The company whose product can write a rep's emails, and probably their forecast, decided that the fastest road into the enterprise is a human rep with a quota and an expense account. Software keeps eating the world, and the field seller keeps cashing the check. Ask where the money lives. The rest of the deal follows from the answer.