Big dreams and tiny revenue are the new norm for AI IPOs

Investors are used to valuing companies with no earnings. They have little choice, because so many large companies today bleed red ink. SpaceX, Intel, Snowflake and Moderna all notched up a loss over the past year. But what about when revenue, not just profit, is in short supply?

Businesses with no top line are rare on public markets, but they do exist. Of the 6,883 companies with primary listings in the US, just 243 are projected to make no revenue over the next year, according to S&P Capital IQ. Europe seems less tolerant, with just 59 turnover-free companies. Almost all are in biotech or mining.

Like those sectors, AI offers very little revenue now and potentially a lot in the future. But the nature of the AI boom and the absolute size of valuations mean that the contrast between ambition and income is even more glaring.

Take SB Energy and Nscale, two data centre builders preparing for initial public offerings in New York. Both do have revenue, though not much: $140mn in the last half year, in both cases. Yet their hoped-for valuations of $50bn and $35bn, respectively, would make them outliers compared with the rest of the market. The biggest US-listed company with annual revenue below $500mn, excluding biotechs and bitcoin treasuries, is satellite company AST SpaceMobile, valued at $27bn.

Both have done what they can to reassure incoming investors that revenue is coming. They have signed contracts with companies such as OpenAI, Anthropic and ByteDance, committing the future tenants to pay or find someone else who will. There’s still risk, though, because their generous “backlogs” depend on the companies building their data centres on time and on budget.

Companies with little or even no revenue are, of course, perfectly at home in the portfolios of venture capitalists, who are adept at putting a price on moonshots. Safe Superintelligence, an AI model maker with no revenue — and no current desire to make any — was valued last year at $32bn, and is currently raising $5bn in new funding, according to PitchBook.

But public markets aren’t such a good match for those businesses. Look at the rush of start-ups, in futuristic sectors like quantum computing and vertical take-off, that merged with “special-purpose acquisition companies” about five years ago. It generally didn’t go well. The median Spac is down nearly 90 per cent since merging with its target, according to ListingTrack data.

Investors’ imaginations will be tested by the potential listings of Anthropic and OpenAI. Both have revenue in the tens of billions, but their hoped-for valuations — potentially $2tn in Anthropic’s case — will be many times the size of their sales, with little certainty as to how much will convert to profit.

When markets are ebullient, this is less of a problem. But concerns over the uncertain nature of AI investment are mounting. Meanwhile, the listings of OpenAI, Anthropic and SB Energy are already taking longer than expected. When risk tolerance recedes, those with least to show in relative terms tend to feel the chill most.

john.foley@ft.com

Lex sends a newsletter every Wednesday with additional insights, the pick of the week’s columns and links to what we’re reading. Premium subscribers to the FT can sign up