The market for AI accelerator chips such as GPUs is on fire, having grown from $10bn in 2022 to $78bn this year. Can it last? Omdia’s latest research note, Reconciling AI hardware spending with applications revenue growth, examines whether the AI applications that are meant to justify this massive investment are generating enough revenue to sustain it.
Omdia’s Principal Analyst for Advanced Computing, Alexander Harrowell said: “At least in the short term, infrastructure investment is consuming a very large percentage of the revenue from AI applications. Our forecasting suggests that there is likely to be a squeeze next year as growth in infrastructure spending peaks, when nearly all the revenue on the application side will be going out to the chip makers.”
“In the long run, AI infrastructure spending growth has to converge with the growth rate of AI applications; it’s the applications that pay the bills. The question is whether this is achieved by the applications increasing their monetization or the infrastructure spending slowing down. There is also a twist in that now the race for size in AI models has cooled off, comparable models are getting more efficient from generation to generation. This is helpful for the applications’ unit economics but a headwind for chip demand.”
