For two years, companies bought AI the way they bought a new SaaS tool. A team found something useful, put it on a credit card, and IT found out later.
Three stories this week suggest that phase is ending. AI is starting to be bought the way companies buy infrastructure, through existing contracts, audited vendors and trusted data providers.
Signal one: ChatGPT moves into the cloud budget
OpenAI announced that ChatGPT Enterprise is now available through AWS Marketplace. Organisations can buy it through their existing AWS account and, in OpenAI's words, apply eligible spend toward their AWS commitments.
That last detail is the whole story. Large companies sign multi-year cloud commitments worth millions. Every dollar spent through the marketplace counts toward a commitment they have already promised to spend. For a CFO, buying ChatGPT this way can feel almost free, because the budget exists already.
The effect on decision-making is significant. The person who owns the cloud commitment now has real influence over which AI tools the company uses. In many organisations, that is not the CMO. It is IT or procurement, optimising for the commitment rather than for the marketing team's workflow.
OpenAI says customers get the same models and features as buying directly. I believe that. The change is not in the product. It is in who signs, which budget pays, and how hard it becomes to switch when the spend is woven into a cloud contract.
Signal two: the vendor goes public
TechCrunch reported, citing The Wall Street Journal, that Anthropic plans to set a valuation range this week and could begin trading as soon as next week. Some institutional investors are pushing for a valuation below the $1.5 trillion target discussed recently, itself well below the $2 trillion floated a few months ago.
OpenAI was reportedly valued at $1.25 trillion last week after raising $11 billion, short of the $20 billion it initially sought.
I covered the earlier filing in Anthropic's filing reads like a lock-in manual. The IPO itself matters to buyers for a simpler reason. A public company reports every quarter. Its pricing, margins and growth targets become visible, and so does the pressure to hit them.
That is good and bad for customers. Good, because public vendors disclose more and tend to offer more predictable contracts.
Bad, because quarterly targets have a way of turning into price increases, tier changes and usage limits. If you rely heavily on one model provider, read their quarterly reports. Your next renewal is in there.
Signal three: the data vendor picks its AI
The third story is smaller and, to me, the most telling. GeniosAI, which builds AI software for financial analysis and modelling, came out of stealth with $19.1 million across seed and Series A rounds, reported by Axios. FactSet, one of the major financial data providers, led the Series A.
The company's pitch is analysis that financial institutions can "trace, defend and build on." That phrase is the buying criterion in regulated industries. A clever answer is worthless if an analyst cannot show where every number came from.
The investor matters as much as the product. When an incumbent data vendor backs an AI layer built on top of its data, it is choosing the AI its customers will meet inside tools they already trust and already pay for. Distribution comes through the existing relationship, not through a new sales motion.
Marketing will follow the same path. The AI your team uses for analysis will increasingly arrive inside your analytics platform, your CRM, your ad platform. The question will be less "which AI is best?" and more "which AI can show its sources?"
What ties the three together
All three signals move AI decisions away from individual teams and toward the people who manage contracts, audits and data. That is a normal stage for any technology that becomes essential. Cloud went through it. Analytics went through it.
For marketing leaders, it changes the job. Your AI stack will be shaped by commitments and contracts you did not negotiate. If you want a say, you need to be in the room when the cloud renewal, the CRM contract and the data vendor agreement get discussed.
I would bring three questions to that room. Which budget pays for this, and what does it lock us into? What happens to our data, prompts and workflows if we leave? Can the output be traced back to sources our finance team will accept?
The practical move this quarter
Make an inventory of every AI tool your marketing team pays for, and note which budget it comes from. Most teams discover tools on personal cards, tools inside other platforms and tools nobody uses anymore.
Then talk to procurement before they talk to you. The marketing teams that keep control of their AI stack will be the ones that showed up with requirements before the contract was signed.
AI stopped being a tool you try. It became infrastructure you commit to. Commit carefully.