DRAM, $110bn in 2021, 10% CAGR, $284bn in 2030 (analyst projections)
HBM, $2.5bn in 2024, 25% CAGR, $8bn in 2029 (analyst projections)
My projections are that 25% CAGR is an underestimate, I estimate something close to 50%, to get to $150bn by 2030. The assumptions for the 25% I think underestimate the application demand materially. I’ve seen forecasts all over the place from 25-50% so mine is on the higher range but that reflects my broader view that demand is AI application demand is being systematically underforecast.
I also agree that AI application demand is being systematically underforecast -- the way I'd put it is "At some point this decade the entire US tech industry + executive branch will be in agreement that 'winning' the AGI race is the #1 national priority, and will be throwing around trillion-dollar-bills to achieve that end. And the main thing they'll be buying will be compute." Insofar as this prediction comes true, do you think it'll still be only about $150bn, or would that itself be an underestimate?
Can you say more about where the $150B bullish scenario number is coming from? Why not say $300B? Why not $1T?
DRAM, $110bn in 2021, 10% CAGR, $284bn in 2030 (analyst projections)
HBM, $2.5bn in 2024, 25% CAGR, $8bn in 2029 (analyst projections)
My projections are that 25% CAGR is an underestimate, I estimate something close to 50%, to get to $150bn by 2030. The assumptions for the 25% I think underestimate the application demand materially. I’ve seen forecasts all over the place from 25-50% so mine is on the higher range but that reflects my broader view that demand is AI application demand is being systematically underforecast.
I also agree that AI application demand is being systematically underforecast -- the way I'd put it is "At some point this decade the entire US tech industry + executive branch will be in agreement that 'winning' the AGI race is the #1 national priority, and will be throwing around trillion-dollar-bills to achieve that end. And the main thing they'll be buying will be compute." Insofar as this prediction comes true, do you think it'll still be only about $150bn, or would that itself be an underestimate?