OpenAI Projects ~$750 Billion in Computing Spending by 2030
nashnova research
OpenAI expects to spend roughly $750 billion on computing resources by 2030 — a figure that signals the AI race is shifting from model superiority to infrastructure dominance, with GPUs, data centers, and power grids as the defining bottlenecks.
How big is $750 billion?
OpenAI projects cumulative compute spending of roughly $750 billion through 2030.
In plain terms = this is not a one-off purchase but a sustained, multi-year buying spree — chips, data centers, and power infrastructure, year after year.
This reflects OpenAI's core bet: AI compute demand is not peaking — it is just getting started.
Where does the money go?
Three pillars: GPU chips (the core engine of AI computation), data centers (the physical space housing those chips), and power infrastructure (electricity and cooling to keep them running).
This means → the competition among AI companies is no longer just "whose model is smarter" — it is who can secure enough chips and electricity.
In plain terms = models can be open-sourced and replicated, but server farms and power grids are hard assets — whoever builds first holds the moat.
What does this mean for the market?
A $750 billion procurement pipeline provides long-term demand support for the upstream supply chain — GPU makers like Nvidia, data-center builders, and power-equipment suppliers all sit on the beneficiary list.
This means → AI infrastructure is not a short-cycle capex burst but a multi-year sustained investment wave.
This signals a broader shift: when the leading AI company bets its future on raw compute, certainty across the entire compute supply chain rises.
市场有风险，内容仅供研究参考，不构成投资建议。