No. of Recommendations: 1
"I think a useful analogy might be all those nice older aircraft parked in the desert."
I'm not an expert on GPUs nor older aircraft, although I travelled and worked in 3rd world countries over the decades, and flown on a number of very "dated" airplanes still in use by the local airlines. Flying on "Garuda" and "Egypt Air" with my family made me nervous, but they took us from point A to point B.
This is a post from Dwarkesh Patel, a SillyCon Valley based podcaster and author, who recently riffed on why compute may get more expensive, rather than less expensive.
dwarkesh.com - Why compute might get 10x more expensiveI believe he makes some good points. For instance:
"To look at how crazy the compute market is in that tranche, consider the price at which Google and Anthropic are renting compute from SpaceX. Google is reportedly paying $900 million a month for 110K GPUs that are a blend of GB200s and GB300s. That’s roughly 2x the spot price per hour for those GPUs. And the current spot price is itself 40% higher than it was in February.
I want to emphasize the key conclusion here: as AI models become smarter, they’ll better monetize the same amount of compute. If a true human-level software engineer that could run on an H100 (6 year old chip)equivalent, at current market rates for software engineers, that H100 should rent for over $250k a year. That’s 15x today’s spot prices." 