CoreWeave: Climbing on Nvidia’s supply constraints

Key Points

• CoreWeave has been a retail investor favourite since its recent IPO

• The company gets preferential access to Nvidia GPUs and then leases this capacity

• Nvidia’s supply constraints may act as a tailwind

One thing we learnt from the Nvidia earnings report was that AI infrastructure capital expenditures are not slowing. CEO Jensen Huang forecasts $3 trillion to $4 trillion in AI infrastructure spend by the end of the decade.

That might seem fanciful – who knows whether Google, Microsoft or Meta can sustain the current pace of spending. A decade is a long time in tech investing. But certainly for now Nvidia’s biggest barrier to growing revenues faster seems to be its own supply constraints; it just can’t build the GPUs fast enough to keep pace with demand. 

This could be good news for CoreWeave, an AI company that has been a favourite among retail investors since it went public earlier this year.

What Does CoreWeave Do?

CoreWeave has preferential access to Nvidia’s GPUs – Nvidia owns a big chunk of the company. It has about 250,000 of the GPUs that AI hyperscalers are desperate to get their hands on, spread across 33 data centres.

Essentially it leases this capacity out to the highest bidders – providing overflow capacity when companies that are building the next ChatGPTs of tomorrow need it. 

Q1 results showed a good performance. It posted $982 million in revenue — up 420% — and a huge backlog of sales worth $25.9bn, including a $4bn deal with OpenAI.

Why Does this Matter?

Nvidia’s results showed that its chief barrier to growing revenues faster is down to supply, not demand. Its miss on data centre revenues was mainly a supply constraints problem. 

There are risks, of course. All of CoreWeave’s GPUs come from Nvidia, while 62% of its revenues last year came from Microsoft. There are clear concentration risks (mind you, 39% of Nvidia revenues come from two customers...it’s the nature of the beast as only a very few large cap stocks with very broad shoulders have the cash flow to spend on AI buildout). It also has a lot of debt – about $8bn as of the end of 2024 due to the massive capex involved. 

And the stock has been very volatile – since its IPO it has traded up or down 5% or more on half its trading days. After a post-IPO high above $183 it’s shed about $80 to trade around $103 as of Friday, 29 August despite the approximately 7% gain on Thursday.

Analyst View

Wall Street analysts have a mixed view on the stock with 8 Buy ratings, 15 Holds and three Sell equivalents.