The AI IPO Race Is the Wrong Race to Win

Anthropic just confidentially filed for an IPO. OpenAI is targeting a fall listing. Together, these two companies are chasing valuations approaching $2 trillion.

The market is cheering — I'm not.

This isn't skepticism and “AI Bubble” fearmongering for its own sake. Anthropic's revenue run rate crossed $47 billion this year; that represents real growth. OpenAI has over 900 million weekly active users. These are not vaporware companies. The technology is real and the demand is real.

But going public right now, at these valuations, at this stage of the technology's development, into this market structure, is the wrong move. Not just for the companies, but for the market — and primarily, the people who'll own shares in their 401(k) without ever choosing to.

Here's why:

The Mission isn’t compatible With the Market

Anthropic was founded in 2021 by former OpenAI researchers who left because they believed AI was being developed too fast, with too little regard for safety. That wasn't a PR positioning decision. It was the founding thesis; the entire reason the company exists is to prove that you can build frontier AI responsibly, that safety and capability don't have to be in opposition.

That's a hard thing to hold onto when you're answering to public shareholders every 90 days.

I've worked with enough growth-stage leaders to know how this goes. The first earnings call feels manageable. The second one, you start making small compromises. By the third, the mission has quietly become a talking point rather than a guardrail. Public markets don't fund missions; they fund quarters.

The moment Anthropic rings the bell, the pressure begins — pressure to monetize faster, ship more aggressively, cut the research that doesn't have a near-term revenue line.

Leadership attention is a finite resource, and earnings calls, analyst expectations, and activist investors all compete for it voraciously.

We've watched this dynamic play out across every transformative technology sector. The companies that changed the world were almost always the ones that had the runway to stay patient; the ones that could make long-term bets without justifying every dollar to a shareholder base looking for next quarter's return.

AI is still early. The safety frameworks aren't locked in. The alignment research isn't finished. The regulatory environment is still being written. This is exactly the wrong moment to hand the steering wheel to public markets.

Freezing isn't a strategy. But sprinting before you know where you're going isn't either.

The S&P 500 Doesn't Need More Tech

Let's talk about what adding trillion-dollar AI pure-plays to public markets actually does to the broader economy; because most people haven't thought through the downstream consequences.

The S&P 500 has a concentration problem, and it's not a minor one. Tech-oriented stocks now represent 44% of the index, up from 26% a decade ago and just 15% twenty years ago. Nvidia, Microsoft, and Apple alone account for nearly 20% of the entire index; the top 10 holdings make up over 38% of total weight. The "broad market" benchmark is behaving less like a diversified index and more like a concentrated tech fund.

This isn't just an academic concern. Evercore ISI has warned that concentration is now at record highs, with a handful of AI-focused names driving over 40% of 2026 earnings revisions. When seven stocks drive a third of an index, a single regulatory setback — an antitrust ruling, a production snag, a missed earnings print — can cascade across the entire market as index-tracking funds are forced to liquidate.

Now imagine adding OpenAI and Anthropic to that mix at trillion-dollar valuations.

AI-focused sectors already represent roughly 60% of the index when you include tech, communication services, and consumer discretionary. Every dollar that flows into an S&P 500 index fund, including the retirement savings of people who have never heard of Claude or ChatGPT, becomes further exposed to the fortunes of a technology that is still finding its business model.

My parents have an index fund. They are not AI investors. They shouldn't have to be.

Concentration risk doesn't announce itself; it accumulates quietly, and then it doesn't.

The Numbers Don't Hold Up in Daylight

This is the part that keeps me up at night.

A significant portion of AI's eye-watering valuations isn't built on revenue; it's built on circular funding, a web of companies investing in each other, inflating each other's worth, creating the appearance of demand where there is really just interdependence.

Consider the architecture. When Oracle announced the $500 billion Stargate Project with OpenAI, Nvidia was positioned as the primary hardware supplier. Nvidia also invested $100 billion of its own capital in OpenAI; Nvidia also holds a stake in CoreWeave, which counts Microsoft as a major partner, and Microsoft is itself one of OpenAI's largest backers. Microsoft and Nvidia have also invested in Anthropic, which has committed its primary cloud infrastructure to Amazon and Google; both of whom are major Anthropic investors.

The circle keeps closing on itself.

GMO analysts have described these arrangements as "reminiscent of the circular financing of the internet bubble"; structures that create the appearance of commercial demand while masking an underlying financial circularity that inflates all parties' valuations at once. Singapore's Monetary Authority put it more plainly, warning in its Financial Stability Review that some large tech firms are using financing arrangements that "could mask leverage and increase funding dependencies."

These aren't fringe voices. These are serious institutions flagging a serious structural risk.

And the gap between valuation and fundamentals is hard to ignore. OpenAI has raised $180 billion from investors across 13 funding rounds; in 2025, it generated $13.1 billion in revenue. HSBC analysis projects OpenAI will require $207 billion in additional capital and will remain unprofitable through at least 2030. Anthropic's growth is more impressive — revenue run rate crossed $30 billion this year — but the company has raised over $64 billion, most of which has been consumed by compute costs.

Here's what an IPO does to that math: it forces it into the open. S-1 documents don't let you tell the story the way a private funding pitch does. The revenue figures, the burn rate, the assumptions baked into those sky-high valuations; all of it becomes public, scrutinized by analysts whose job is to find the gap between the narrative and the numbers.

That scrutiny is healthy in the long run. But if the gap is as wide as some analysts fear, the revaluation that follows won't be quiet. And unlike a down round in private markets, where a handful of institutional investors absorb the hit, a public revaluation lands on everyone.

Including your parents' index fund.

The Companies Most Likely to Win Are the Ones That Stay Patient

None of this means AI isn't real. It is. None of this means Anthropic or OpenAI are fraudulent; they're not. The technology is genuinely transformative and the long-term opportunity is enormous.

But the companies most likely to define that long-term opportunity are the ones disciplined enough to build something durable; not the ones that sprint to liquidity before the foundation is set.

Going public at this moment rewards the investors who got in early. It transfers the risk to everyone else. That's not a growth strategy; that's an exit dressed up as a milestone.

The AI IPO race is being framed as a sign of maturity, proof that these companies have arrived. I'd argue it's the opposite. The companies that will matter in ten years are the ones that resist the pressure to perform for public markets right now and stay focused on getting the technology right. Getting it safe. Getting it durable.

There's a reason the most enduring tech companies spent years building before they ever faced a shareholder meeting. Patience isn't weakness; it's the strategy.

In this case, the race to ring the bell first isn't the race worth winning.