- Jim Chanos challenges NVIDIA‘s “rentable” compute thesis, questioning why the company isn’t charging more for its scarce GPUs
- Nvidia increasingly pitches its chips as productive infrastructure that generates revenue rather than hardware that simply depreciates
- The debate highlights growing scrutiny over who ultimately captures the economics of the AI infrastructure boom
- Rental prices for three-year-old H100 GPUs rose 22% in one month, defying traditional depreciation assumptions
Legendary short seller Jim Chanos questioned Nvidia CEO Jensen Huang’s characterization of the company’s GPUs as “fungible, durable and highly rentable” assets. The founder of Kynikos Associates, known for predicting Enron’s collapse, responded by asking why Nvidia doesn’t rent the chips out itself or simply keep raising prices.
Chanos clarified his skepticism targets companies buying Nvidia hardware specifically to rent it out, not Nvidia itself. He argues that if the chips remain scarce and highly utilized, third-party GPU clouds and neoclouds could capture economics that Nvidia leaves on the table by selling hardware upfront.
Nvidia has increasingly positioned its GPUs as productive infrastructure rather than semiconductors that depreciate over time. The company has partnered with major financial firms to create financing platforms around AI compute, describing it as an “investable asset.”
Huang reshared a post claiming rental prices for three-year-old H100 GPUs increased 22% over one month to $3.28 per hour, defying traditional assumptions that aging hardware steadily loses economic value. However, Chanos questions whether GPU renters can sustain attractive returns.
NVDA shares dipped 2% on Tuesday amid a market-wide selloff, though they recovered slightly in overnight trading. On Stocktwits, retail sentiment for NVDA remained bearish, unchanged from the previous week.
✅ Follow BITNEWSBOT on Telegram, Facebook, LinkedIn, X.com, and Google News for instant updates.
