A doubling every nine months
Around 20 million AI accelerator chips are deployed in data centres worldwide. That stock is doubling roughly every nine months, according to Epoch AI research highlighted by The New York Times, putting the world on a path to about 200 million chips by the end of 2028. Hundreds of data centre projects are under construction across the United States, Europe and Asia to house them.
The measure is deliberately physical rather than financial. Analysts count H100-class accelerators and their equivalents, then multiply each model by its computing throughput to track how much usable capacity exists. Nvidia remains the dominant supplier, and the largest hyperscale buyers control the majority of global AI computing power.
Power, memory and long lead times
The bottleneck is no longer only chip supply. Data centres already consume roughly as much electricity as Germany, and demand is expected to quadruple by 2030. Building capacity costs on the order of $100 billion per gigawatt once land, power distribution, cooling and networking are included. Global AI infrastructure spending is projected to top $1 trillion a year by 2029, up from $318 billion last year, with Amazon, Google, Microsoft, Meta and Oracle alone expected to spend hundreds of billions annually.
That demand ripples outward. Power equipment has multi-year lead times. High-bandwidth memory and advanced processors stay tightly allocated. Suppliers of specialty materials, passive components and printed circuit board assemblies report pricing pressure and harder allocation decisions, and makers of precision machined and vacuum components used in chip fabrication are seeing orders stretch. Semiconductor manufacturing capacity is being expanded, but fabs take years to bring online, so the gap between demand and supply is expected to persist.
Regulators push back
Not everyone welcomes the buildout. New York State has imposed a one-year moratorium on new data centres, with Governor Kathy Hochul taking an aggressive line on the sector and some state lawmakers pushing for stricter limits. Local opposition has focused on electricity prices, water use and land, even as operators argue that the facilities bring tax revenue and grid investment. Utilities and grid operators say the timing mismatch is the hard part: projects want power in two or three years, while new generation and transmission lines can take a decade.
---
Reliable Ferrofluid Vacuum Rotary Feedthrough for Semiconductor Manufacturing
In semiconductor manufacturing, maintaining ultra-high vacuum integrity during rotary motion is critical — from single silicon crystal growth and CVD to sputtering and plasma etching. Ferrofluid Vacuum Rotary Shaft Feedthrough from FerrofluidTec delivers hermetically sealed rotary feedthroughs with leakage rates as low as 10⁻¹² Pa·m³·s⁻¹, vacuum degree up to 10⁻⁶ Pa, and customizable shaft diameters from 10mm to 100mm. Built with non-magnetic stainless steel housings and magnetic permeability stainless steel shafts, these feedthroughs provide reliable, maintenance-free sealing for demanding vacuum environments. Visit www.ferrofluidtec.com to learn more.