Every gamer knows how fragile a graphics card can be. Under sustained heavy use—the kind of multi-hour gaming marathons you’d see from a dedicated Fortnite teen—a GPU’s effective lifespan is typically only three to five years. This fact has had some investors shorting or making dire predictions about CoreWeave in recent months.
Up until five years ago, GPUs were really in the domain of recreational gamers; they’ve since become the backbone of the AI industry. The complicated computational exercises that chatbots like ChatGPT need to do turn out to be fairly easy for GPUs.
There have been a handful of companies that have also made a name for themselves by creating data centers full of GPUs to power the AI industry. Of course, there are the big players: AWS, Microsoft Azure, and then there’s the latest entrant to the fray, CoreWeave. The company, which ranked 45 on this year’s Inc. 5000 list, got its start building datacenters for cryptomining—another resource intensive activity that requires the sophisticated computing processing power of GPUs.
When the AI boom firmly materialized, CoreWeave was well-positioned to take advantage of the moment. As Inc. previously reported, the company closed out 2024 with nearly $2 billion in revenue, a 5,896 percent increase in growth over three years. It’s also been a stock-market darling IPO’ing to huge fanfare earlier this year. The company started selling for about $40 a share, but despite trading near $140 in October, today hovers around $70.
But while CoreWeave has arguably been one of the greatest beneficiaries of the AI hypercycle, it’s also drawing the skepticism of major short sellers.
Earlier this year, the investment management firm Kerrisdale Capital called CoreWeave “the poster child of the AI infrastructure bubble” after revealing a short position on its stock.
And earlier this month Michael Burry, the famed investor, railed against hyperscalers (another term for the big GPU datacenter companies) arguing that they were underselling the lifespan of the GPUs that make up the core of their product. For its part, CoreWeave measures the depreciation (or lifespan) of its technological assets (its GPUs) in its S-1 fillings as about six years . But short sellers are betting that those GPUs are going to last about as long as the GPU of a fortnite addicted teenager, considering how resource hungry artificial intelligence actually is. That wear and tear could significantly impact the company’s profitability, especially considering the company is already reporting more than $800 million in net losses.
Tekendra Parmar
Source link