infrastructureScore 70/100Research
Sam Altman: AI inference demand growing 10x+ annually; compute shortage permanent through scaling
“I would guess that worldwide demand for inference so a kind of subjective ly grows 10x a year for the next many years... we will sort of never be out of the compute shortage. Uh I've never seen any commodity quite like this one but it seems to me like the demand for sufficiently high quality intelligence at a sufficiently low price is effectively uncapped.”
Why it matters
Altman explicitly signals perpetual compute shortage despite exponential infrastructure expansion, indicating sustained capex acceleration and that inference workloads—not just training—will drive sustained GPU/chip demand and power infrastructure investment for years.
Investment implication
Suppliers of data center GPUs, HBM, interconnect, power delivery (e.g., NVIDIA, TSMC, advanced packaging vendors, power management firms), and grid/power infrastructure companies will face sustained upside from inference scaling. Inference cost reduction will not curb demand; it will expand TAM.