Steve Hou: GPU rents, token prices and the AI trade - Opening Bell Daily
Steve Hou, head of research at Silicon Data, says fears of fast GPU depreciation have been 'quite overblown', noting that even a six-year-old Nvidia A100 still commands significant rental rates.
Steve Hou, head of research at Silicon Data and a veteran quantitative investment strategist, says the fear of fast GPU depreciation has been 'quite overblown'. He argues that a six-year-old Nvidia A100 is still holding 'pretty significant rental rates'.
On Full Signal with Phil Rosen, in his fourth appearance on the show, Hou discussed the spread between token prices and GPU costs, what investors are pricing into AI stocks, and why he calls the bubble label a 'tiresome, oversimplistic characterization.'
Token-price squeeze
Rosen noted that GPU rental prices are 'right near a record high' and have been rising all year, even as token prices appear to have 'fallen off a cliff since about May this year.' Hou explained that Silicon Data tracks GPU rental and token pricing as AI fundamentals.
The key figure to watch, according to Hou, is the spread between what tokens sell for and the cost of renting a GPU, which he likened to the squeeze between CPI and PPI. He described how a shift toward cheaper flash models could deflate the effective price of a million tokens, yet still be profitable due to increased demand—a dynamic he called Jevons paradox.
Market pricing
Hou suggested that worries about labs' economics helped drive a local peak in AI-related stocks. He mentioned that while the S&P 500 was only a couple of percentage points from its highs, a median semiconductor name might have been off 50% before the most recent rebound.
Stocks are pricing the second derivative of growth, not a halt in the AI buildout.
He explained that spending is still growing, but prices may be anticipating that it will not keep accelerating from 2028 onward. When asked about valuations, Hou declined to make a recommendation but noted fundamentals have improved, pointing to Anthropic's S-1 filing which showed good margins, a low burn rate, and a cash-flow-positive quarter.
GPU values
Hou addressed how depreciation affects hyperscaler earnings. Using on-demand rental rates and market-implied long-term rental curves, he calculated that a six-year-old A100 (first announced in May 2020) has a 22% fair value, a chip he said should already be dead under three-year straight-line depreciation.
He put the H100 at about 50% to 60% of its original sale price and estimated the B200's fair value at about 180%, 'well over 100%'. This, he said, is good news for Nvidia and its customers because the chips can remain productive for extended periods.
Regarding the AI bubble label, Hou called it a 'tiresome, oversimplistic characterization.' He dismissed fears of an overbuild of unused compute like the dark fiber in the dot-com era, stating 'that's certainly not the case today, because all the GPUs are being used.' Hou noted that on-demand rental rates have been 'going vertical' in recent weeks, 'like the GPUs are melting.'
Related
