Three money stories landed this week, and individually they read as separate headlines. Together they describe one shift. The AI capital market is starting to move from private, where a valuation is whatever the last investor agreed to, toward public, where a valuation has to survive an audit, a prospectus, and a few thousand skeptical analysts. That move is where the claims meet the proof.
I am not calling a top. I am reading what the smart money is doing with its exits, because that tells you more than any keynote about where the cycle actually is.
OpenAI Filed to Go Public
OpenAI confirmed it submitted a confidential draft S-1 to the SEC, the first formal step toward an IPO, while stressing it has not set timing for anything further. Confidential filing is the standard way large companies start the process quietly, testing the path without committing to a date.
The signal is the direction, not the paperwork. The most watched private company in the world is opening the door to public markets, and public markets ask questions private rounds never did. Real revenue, real margins, real disclosure, all of it printed and legally binding. I wrote about what the 852 billion dollar valuation actually means, and an S-1 is the moment that number stops being a story investors tell each other and starts being a claim a regulator and a public float get to test.
For operators the read is simple. When your most important AI vendor becomes a public company, its incentives shift toward quarterly numbers and investor expectations. That changes pricing, roadmap, and support in ways worth planning for now, not after the first earnings call.
Perplexity Set a Date Anyway
In an interview with CNBC, Perplexity CEO Aravind Srinivas said the company is planning its own IPO for 2028, and notably, regardless of how the OpenAI and Anthropic offerings perform. That last part is the interesting bit.
Naming a date three years out, and decoupling it from how the giants do, is a confidence play. It says we intend to be a durable public company on our own schedule, not a passenger on someone else's exit. Whether that holds is unknowable. The posture itself is the data point.
There is a caution under the bravado. A 2028 target assumes the window stays open that long, and AI windows have a habit of closing fast. I described the narrow exit window AI startups are working inside, and a confident multi-year IPO plan is a bet that this cycle is the exception. Maybe. The companies saying it most loudly are the ones who most need it to be true.
A Founder Called the Valuations Fake
The third story is the honest one. Mercor co-founder Brendan Foody publicly accused Sequoia of dual pricing, investing a large amount at a low valuation and a small amount at a much higher one to manufacture an impressive headline number. He said he had seen roughly half a dozen such rounds in six months.
The example he pointed to is brutal. One company announced a one billion dollar Series B days after being valued at under 400 million in a Series A extension. TechCrunch reported it, Sequoia's partner denied any intent to deceive and called it a response to market pricing, and the gap sat there in public for everyone to see.
This is the proof problem in its purest form. A valuation is a claim until something forces it to be earned. Private rounds let the claim float, and the headline number becomes a recruiting tool, an angel-investor lure, a press release. Public markets are the thing that eventually forces the question, which is exactly why the first two stories matter. The IPO window opening is also the audit window opening.
The thread tying all three together is the one I keep landing on this week. In private, you can be valued on a story. In public, you have to be believed on the numbers. The AI money is starting to cross that line, and the crossing is going to separate the companies with a business from the companies with a narrative.
For an operator picking vendors, this is a usable filter. A company priced on a story will defend the story, which shows up as aggressive sales, constant roadmap promises, and pricing built to fund the next round rather than serve you. A company that can survive public scrutiny tends to behave more calmly, because it does not need your contract to prop up a number. You cannot read a private valuation, but you can read how a vendor sells, and the tell is usually there.
Watch the exits, not the announcements. The announcements are claims. The exits are where the claims finally get priced by people who do not care about the pitch.