---
title: "OpenAI and Anthropic still look hard to float"
author: "Sloane Carrington"
datePublished: 2026-07-06T07:15:00.000Z
canonical: "https://scramnews.com/post/00thq800o06j4/openai-anthropic-struggle-to-float-2026"
---

Anthropic’s march toward investor meetings and OpenAI’s apparent willingness to wait until 2027 have turned a private-market fantasy into a public-market stress test. [Anthropic](https://www.anthropic.com/) and [OpenAI](https://openai.com/) still sit near the centre of the AI boom, but a float would force both groups to answer questions that private capital has largely let them postpone. The issue is not whether investors want the deals. It is whether public investors would underwrite frontier-model economics, unusual governance and still-blurry disclosure on anything close to today’s private valuations.

From an analyst’s vantage, the timing gap already tells part of the story. [CNBC’s reporting on Anthropic’s investor meetings](https://www.cnbc.com/2026/07/15/anthropic-ipo-banks-investor-meetings.html) and [Morningstar’s analysis of an OpenAI delay](https://www.morningstar.com/stocks/what-delayed-openai-ipo-would-tell-investors) point to a simple divide: Anthropic looks closer to testing the market, while OpenAI looks closer to testing how long the market will indulge a scarcity premium. That first move matters. The company that lists first is likely to set the comp set, the valuation language and the early tolerance for frontier-AI volatility.

Skeptics arrive just as quickly. Governance is not a side note here; it is part of the valuation case. [Wired’s reporting on Anthropic’s confidential filing](https://www.wired.com/story/anthropic-files-s1-ipo-sec/) highlighted how much attention investors are already paying to structure, while [Anthropic’s own description of its Long-Term Benefit Trust](https://www.anthropic.com/news/the-long-term-benefit-trust) makes plain that financially disinterested trustees will gain growing influence over board selection over time. OpenAI comes with a different complication. The [Financial Times analysis of its listing hurdles](https://www.ft.com/content/7bff5ad3-a7dc-4641-be97-7f383446ff75?syn-25a6b1a6=1) framed the company’s nonprofit-to-public-benefit-company evolution as a capital-markets question, not just a corporate-history footnote.

Plenty of bankers see obvious demand for a marquee AI listing. Morningstar’s report on the sector’s backlog quoted Piper Sandler analyst Brian White putting the case bluntly:

> “We need OpenAI and Anthropic to go public so we can start to see other software IPOs come to market.”
>
> — Brian White, Piper Sandler, via [Morningstar](https://www.morningstar.com/stocks/what-happens-if-openai-delays-its-ipo-2027)

Investor appetite is real. So is the reason it may not be enough.

## The first float will set the comp

Private valuations cited in recent coverage, roughly [$965 billion for Anthropic and $852 billion for OpenAI](https://www.cnbc.com/2026/07/15/anthropic-ipo-banks-investor-meetings.html), mean neither company would arrive as a routine software float. They would arrive as tests of whether public markets want to price frontier AI like elite SaaS, like infrastructure-heavy compute resellers, or like a scarcity asset that deserves a premium precisely because comparable names do not yet trade.

![Smartphone displaying AI apps in front of a financial data screen, reflecting the market's attempt to turn AI leaders into tradable securities.](https://images.pexels.com/photos/33955927/pexels-photo-33955927.jpeg?auto=compress&cs=tinysrgb&dpr=2&h=650&w=940)

Morningstar’s first-mover logic matters more than the usual bragging rights around timing. The publication quoted Class V Group’s Lise Buyer arguing that the company that goes first gets to define the playing field, from comparables to key metrics. In other words, the first S-1 does not just raise money; it teaches the market how to value the whole cohort.

> “The company that goes first gets to define ‘the playing field’—which companies are the comparables, what is the valuation, what are the key metrics and measures of success.”
>
> — Lise Buyer, Class V Group, via [Morningstar](https://www.morningstar.com/stocks/what-happens-if-openai-delays-its-ipo-2027)

Even so, public investors are unlikely to stop at hype. [PitchBook’s valuation analysis of OpenAI](https://pitchbook.com/news/articles/openai-business-quality-valuation-analysis) argued that OpenAI screens as the most expensive AI name in its peer set, while [Morningstar’s comparison of the two companies](https://www.morningstar.com/stocks/what-delayed-openai-ipo-would-tell-investors) cited a gap of roughly 34 times sales for OpenAI against 20 times for Anthropic. The same work pointed to PitchBook AIBQ scores of 4.53 for OpenAI and 8.20 for Anthropic. That partly answers the analyst’s first question about comps. If public buyers lean on business quality rather than narrative scarcity, Anthropic looks closer to a defensible software multiple and OpenAI looks closer to a valuation that still assumes perfect execution.

Elsewhere in the IPO tape, [CNBC reported that SpaceX shares slipped below their $135 IPO price](https://www.cnbc.com/2026/07/15/spacex-spcx-stock-ipo-price.html) just as anticipation built around other mega-listings. That does not make SpaceX a clean comp for either AI lab. It does show how quickly post-listing trading can puncture the idea that scarcity alone will hold up a premium once daily price discovery begins.

## Disclosure is where the story hardens

The insider case for these companies is easier to tell in private than it will be in an S-1. Both labs can point to rapid revenue growth, deep enterprise interest and a market hungry for exposure to generative AI. Yet the filing stage will force harder disclosures on gross margins, revenue mix, customer concentration and the true cost of staying at the frontier. That is the point where a growth story turns into a public-market operating model.

![Detailed view of a stock market screen showing numbers and data, symbolising the disclosures investors will demand from any AI IPO.](https://images.pexels.com/photos/534216/pexels-photo-534216.jpeg?auto=compress&cs=tinysrgb&dpr=2&h=650&w=940)

Inside the enterprise pitch, the important question is not whether demand exists. It is whether that demand behaves like durable software revenue once quarterly scrutiny arrives. [CNBC’s reporting on Anthropic’s IPO push](https://www.cnbc.com/2026/07/15/anthropic-ipo-banks-investor-meetings.html) tied part of its public-market appeal to faster revenue scaling, and [StoneX’s analysis of AI IPO economics](https://www.stonex.com/en/insights/what-openai-and-anthropic-ipos-could-tell-us-about-ai-s-true-economics/) argued that a listing could finally show how much of that revenue remains after compute bills and infrastructure tolls. The same question hangs over developer demand for coding tools such as Claude Code: is it recurring software revenue, or metered usage that looks sticky only while the model cycle is hot?

Public investors can at least test that question. If Anthropic can show that enterprise customers renew, expand and tolerate pricing without a corresponding explosion in compute cost, the software case gets stronger. If OpenAI eventually files with similar clarity, its delay will look tactical rather than ominous. If the disclosures show thin margins, heavy concentration or aggressive accounting choices, the trillion-dollar dreams begin to look less like delayed inevitability and more like private-market optimism meeting public-market discipline.

Board structure magnifies that pressure because it narrows the room for investor trust. OpenAI’s corporate evolution already asks buyers to assess how mission control, commercial urgency and shareholder rights coexist inside one structure. Anthropic asks a different version of the same question. The trust model described by the company and analysed by [Harvard Law School Forum’s note on the Long-Term Benefit Trust](https://corpgov.law.harvard.edu/2023/10/28/anthropic-long-term-benefit-trust/) may please those who want safety-minded oversight, but public investors usually prefer governance they can map cleanly. A mission-protecting structure can be defensible. It can also carry a discount.

Price talk is where even bulls become cautious. Morningstar quoted DataPower Capital’s David Yakobovitch saying a $1 trillion valuation would be “the toughest pill to swallow right now.” That is less a dismissal of frontier AI than an admission that public investors will want proof before they pay up. In private markets, ambition can stand in for disclosure for longer than it can after a listing.

For scramnews readers, the cleanest read is that these are not conventional IPO candidates waiting for a better window. They are capital-intensive labs trying to discover whether the public market shares the private market’s tolerance for opacity, governance experimentation and frontier burn. Anthropic may still go first. OpenAI may still command the bigger narrative. But the first filing that matters will not prove AI is hot. It will show whether the sector’s two flagship labs can explain, in public-company terms, why their economics and control structures deserve to trade like software rather than like faith.