Thursday, July 2, 2026

Anthropic's IPO Arrives Before Its Business Model Does

Three AI giants racing to public markets could test institutional appetite and expose a deeper question: does anyone in AI actually have a sustainable business?

Jun 3, 2026 · 17 Minutes

The Starting Gun Nobody Expected

Anthropic was supposed to wait until fall. Instead, the company has confidentially filed its S1 with the SEC, quietly pulling the trigger on what could become the most consequential IPO season in market history. Coming off a $65 billion private round that valued the company at just under a trillion dollars, Anthropic is now expected to cross that threshold on public markets. For a company that did not exist until 2021, that trajectory is, to use a precise financial term, completely wild.

The move reshuffles the entire race. SpaceX is targeting a reported $75 billion raise on June 12th. OpenAI, which has publicly struggled with missed revenue targets and reported internal friction between its CEO and CFO, now finds itself watching a former rival set the pace. Sam Altman told CNBC he is not focused on IPO timing, which is exactly what someone focused on IPO timing would say.

Who Actually Wins the Race to Market

The banker pitch circulating in the business press is that the company going public first gets to define the market. Neeta pushes back on that framing, and the pushback holds. The company with the stronger underlying business is the more durable bet, and on that score Anthropic has a genuine argument to make.

Unlike OpenAI, which spread its attention across consumer products, robotics, and video generation before pulling back, Anthropic spent five years in focused pursuit of enterprise customers. That discipline appears to be paying off. The caveat, one that the S1 will either confirm or complicate, is that questions hang over Anthropic's accounting. OpenAI executives have publicly questioned the legitimacy of its revenue methodology, and separate reporting suggests a cost-sharing arrangement with another major player may have flattered its recent profitability figures. The S1 will be clarifying.

There is also the governance question. Anthropic is structured as a public benefit corporation and has created something called a long-term benefit trust, an entity to which the company holds partial accountability. It is an untested structure, and institutional investors will want to understand exactly what it means for shareholder rights. The comparison to Novo Nordisk, where a nonprofit foundation sits atop a durable commercial enterprise, is instructive and worth taking seriously, even if the structures are not identical.

Retail Investors Are the Last Stop on the Train

Here is where the IPO season gets genuinely uncomfortable. Nasdaq has changed its index inclusion rules so that newly listed companies can be absorbed into index funds within 15 days of listing. That means ordinary investors holding passive index products will own SpaceX, Anthropic, and potentially OpenAI almost immediately after they list, regardless of whether those companies are profitable or even have coherent long-term business models.

If these companies perform, that is fine. If they do not, retail investors will have had no meaningful choice in the matter.

The Token Cost Problem Is Not a Footnote

Buried beneath the IPO drama is a structural problem that deserves its own headline. Anthropic shifted enterprise customers from subsidized subscriptions to direct token billing earlier this year. The results have been jarring. One unnamed enterprise customer reportedly received a $500 million token bill. Microsoft has reportedly scaled back Claude Code subscriptions. Reports from Tom's Hardware suggest AI agents consume roughly a thousand times more tokens than standard AI queries, a figure that, if accurate, makes current cost structures untenable at scale.

Uber's leadership has been candid that the link between AI coding adoption metrics and actual product output is not yet clear. When one of the most technologically sophisticated consumer companies in the world says it cannot draw a straight line between AI spend and shipped features, the broader enterprise picture deserves scrutiny.

The Zombie Unicorn Warning

The AI IPO wave is also cresting at a moment when the prior vintage of venture-backed companies is quietly deteriorating. PitchBook data shows that nearly half of the 857 startups that reached unicorn status have not raised capital in three years. Companies last funded in 2021 are worth 68% less on average. The 2022 cohort is down 55%. These are not abstractions. They represent employees, limited partners, and venture funds carrying losses that have not yet been fully realized or acknowledged.

The concern Neeta raises is timing. If the zombie unicorn reckoning accelerates at the same moment the AI IPO wave meets market resistance, the collision could be significant. AI spending is currently propping up broader economic activity. A simultaneous contraction in private-market valuations and public-market appetite for unprofitable AI companies would be an uncomfortable place to be holding a passive index fund.

What Comes Next

The summer will be revealing. The Anthropic S1, when it becomes public, will answer questions about revenue quality, governance mechanics, and the true health of the enterprise business. Whether institutional investors have the appetite to absorb SpaceX, Anthropic, and OpenAI in close succession is genuinely unknown. What is clear is that the race is on, the rules have changed, and the bill for a great deal of optimistic accounting may be arriving shortly.

Sources & Further Reading
Watch Next
Subscribe Free →