All news

CoreWeave's cut-price IPO tests public appetite for AI infrastructure

A graphite server-like block on a white plinth with a silk cover just pulled off

CoreWeave listed at $40, below its marketed range, and raised $1.5 billion. Public investors will fund AI compute but charge for debt and concentration.

Public investors bought CoreWeave at $40 a share on 27 March, below the $47 to $55 range the company had marketed, and only after it cut the deal from 49 million to 37.5 million shares. The listing still raised about $1.5 billion, the largest US tech listing since 2021, and the stock began trading on Nasdaq the next day.

This was the first large public test of a business that rents GPUs on multiyear contracts and funds them with debt, and the discount is the verdict. Private markets have paid for AI compute on the promise of demand, lately on the scale of project finance. Public investors priced it on credit as much as on growth, and the price they set becomes the reference for every AI infrastructure company weighing a listing and for every lender that takes GPUs as collateral.

The prospectus shows why the pricing was hard. Revenue rose to $1.9 billion in 2024 from $229 million a year earlier, with a net loss of $863 million. Microsoft alone accounted for 62 percent of revenue, and debt principal stood at about $8 billion at year end, much of it borrowed against GPUs. The discount maps onto three exposures: one dominant customer, borrowing at interest rates mostly above 10 percent, and the pace at which today's chips lose value. If specialist clouds keep trading at a discount to their growth, capital for compute will flow through debt and strategic partners rather than public equity.

Any team with a large training or inference bill gains from the same facts. Compute supply is broadening, and specialist clouds now compete with the hyperscalers for AI workloads, offering capacity, terms and support a startup can negotiate. That is real leverage. The exit lesson runs the other way. Public investors will buy AI companies with fast revenue growth but charge for concentration and leverage, so founders who plan to list should build a broad customer base early and fund growth with contracts and equity before borrowing against hardware.

Whether the discount was too harsh is a question the coming quarters answer. CoreWeave's first quarterly report as a public company will show whether new contracts, including the multiyear agreement with OpenAI signed in March, loosen its reliance on Microsoft. Other AI infrastructure companies filing to list this year, or choosing to wait, will reveal how the market read this deal. The variable that matters most sits in the market for used chips: rental rates and resale prices for previous-generation GPUs decide whether debt secured by them is safe.

Credit, as much as growth, now sets the terms on which AI infrastructure enters public markets. Between the balance sheets that carry the hardware and the companies that use this capacity to build products with many customers, Fanaura, which counts public markets as well as private companies in its remit, prefers to own the second.

For founders

Building AI products that turn rented compute into revenue from many customers? Send us the project.

Fanaura invests its own capital and the capital of its limited partners (LPs), usually through a separate SPV for each deal. If we see a fit, we get in touch.

Apply in five minutes
Subscribe

Get the news by email

One short note each weekday. No marketing, and you can unsubscribe at any time.