BofA: China's AI Capital Expenditure to Triple by 2030
Alina Collins
Bank of America forecasts China's AI capital spending will triple by 2030, while global AI capex grows fivefold; power-equipment lead times have stretched to three years, emerging as the binding constraint on buildout pace.
China's AI capex tripling — where does the money go?
BofA analyst Matty Zhao projects China's AI capital expenditure will reach three times its current level by 2030.
This means → China's AI spending ramp is still in a steep climb, far from peaking.
Global AI capex is expected to grow fivefold over the same period — China's pace trails the world average, but the absolute scale remains enormous.
One-third of spending has left IT — what is it funding?
BofA notes that one-third of global AI capex has shifted to non-IT infrastructure, away from traditional server and chip procurement.
In plain terms = building an AI data center used to mean buying GPUs; now, constructing power plants and installing cooling systems takes a growing share of the bill.
This reflects a pivot in the AI supply chain's center of gravity — from compute hardware toward energy and physical infrastructure.
Three-year lead times on power gear — how bad is the bottleneck?
Beyond semiconductors, Zhao identifies power supply and cooling systems as the critical bottleneck in AI infrastructure buildout.
Transformers — the devices that step high-voltage grid power down to voltages a data center can use — now carry lead times of three years.
This means → even when funding is in place, power-equipment capacity cannot keep up, throttling the pace of AI data-center construction — a scenario where capital is available but unspendable.
Content is for reference only, not financial advice.