Your AI suppliers are heading to the public markets
If your company runs on AI, two or three model makers now sit in your supply chain as firmly as your cloud host or your payroll provider. On September 18, two reports made their economics hard to ignore.
What was reported
Bloomberg reported that OpenAI expects negative free cash flow of $278 billion from 2026 through the end of 2030, while revenue grows from $36 billion this year to $350 billion in 2030. The figures come from a person familiar with materials shared in July, and the Financial Times reported them first. Those are the company’s own projections as reported, not audited results. The same article said Bloomberg had reported earlier that OpenAI is in early talks about funding at a valuation above $1.2 trillion.
The same day, Investing.com, citing The Wall Street Journal, reported that Anthropic had moved its expected IPO timing from October to November. The report said public-market participants anticipate a valuation of about $2 trillion, and that early backers project annualized revenue to clear $110 billion by year-end.
Four days later, on September 22, both companies released new models. Simon Willison’s write-up of the launches called it a new price war.
This post is not about whether either company is a good investment. That is a different question, and nothing here is investment advice. It is about what these numbers mean for a company that buys from them.
Why it lands on the CEO’s desk
A supplier that answers to public markets, or to investors funding years of heavy losses, is under pressure to change its terms. Prices can fall to win share; this week’s launches were framed as a price war. They can also rise, and usage limits, bundles and plan names can change with little notice. Neither direction is a prediction; both are normal for a supplier in this position.
For a lean company, that is a strategic exposure, not an IT detail. If your support, sales follow-up and internal reporting all depend on one model from one lab, a pricing change is a margin change, and an outage is a service outage.
Expect a board member to ask about it. A CEO who can answer in one page looks prepared. A CEO who says “the team handles it” looks exposed.
The one-page supplier memo
Write this once, then update it each quarter. Keep it to a page.
- Who we depend on. Every AI model and platform the company pays for, with monthly spend and the share of total AI spend each one takes.
- What stops if it stops. For each supplier, the workflows that would halt or degrade if it went down or changed terms tomorrow.
- How hard it is to switch. For each workflow: easy (the prompt and tests move to another model in a day), medium (a week of rework and testing) or hard (built on features only this supplier offers).
- Price sensitivity. What happens to gross margin if this supplier’s price doubles, and what happens if it halves. Both numbers matter, because a cut is a chance to move more work over.
- Data terms. Where each supplier stores your data, how long it keeps it, and who in the company last read those terms.
- The next review. A date, and a named owner for the memo.
A decision rule helps with item 3: any workflow rated “hard” that touches customers or revenue gets a tested fallback on a second supplier within the quarter. A tested fallback means the prompts, evaluation set and access are already in place, not just an account.
A worked example
Take a hypothetical 25-person company spending on two model suppliers. Filling in the memo shows that customer support drafts, sales call summaries and the weekly revenue report all run on one supplier, and that the support workflow uses a feature only that supplier offers. That makes support the “hard” item that touches customers.
The action is small: build the same support workflow against a second model, run both on last month’s tickets, and keep the second one ready. The memo to the board then says what depends on whom, what a price change would do to margin, and which fallback is tested. It fits in one paragraph.
What to do this week
- Ask finance for a list of every AI charge on the company card and every AI invoice from the last three months.
- Fill in items 1 and 2 of the memo yourself. It will take less than an hour and will show you where the concentration is.
- Pick the single most customer-facing workflow and ask its owner how hard it would be to move.
- Put the memo on the agenda for your next board or advisory meeting, as a standing quarterly item.
Your suppliers will have a busy few months. The memo makes sure that whatever they change, you have already decided what you will do.
Sources
- OpenAI Projects Burning Through $278 Billion by 2030 · Yahoo Finance (Bloomberg) · 2026-09-18
- Anthropic delays IPO staging to November amid AI fears, WSJ says · Yahoo Finance UK (Investing.com) · 2026-09-18
- Claude Opus 5.5, GPT-6 Sol, GPT-6 Luna, and a new price war · Simon Willison · 2026-09-22
Researched and drafted with AI assistance, checked against the sources above.
See how this could work in your day.
See a worked example →