Hyperscale AI data centers
Hyperscale data centers are now powering AI models with a revolutionary architecture—at a staggering energy cost.
In sprawling stretches of farmland and industrial parks, supersized buildings packed with racks of computers are springing up to fuel the AI race. These engineering marvels are a new species of infrastructure: supercomputers designed to train and run large language models at mind-bending scale, complete with their own specialized chips, cooling systems, and even energy supplies.
Hyperscale AI data centers bundle hundreds of thousands of specialized computer chips called graphics processing units (GPUs), such as Nvidia’s H100s, into synchronized clusters that work like one giant supercomputer. These chips excel at processing massive amounts of data in parallel. Hundreds of thousands of miles of fiber-optic cables connect the chips like a nervous system, letting them communicate at lightning speed. Enormous storage systems continuously feed data to the chips as the facilities hum and whir around the clock.
Tech companies like OpenAI, Google, Amazon, Microsoft, and Meta are pouring hundreds of billions of dollars into this infrastructure. Governments are spending big too.
But the impressive computing power comes at a cost. The densely packed chips run so hot that air-conditioning can’t cool them. Instead, they’re mounted to cold water plates or dunked in baths of cooling fluid. Dipping them in seawater may be next.
The largest data centers being built can devour more than a gigawatt of electricity—enough to power entire cities. Over half of that electricity comes from fossil fuels, while renewables meet just over a quarter of the demand. Some AI giants are turning to nuclear power. Google is dreaming of building solar-powered data centers in space.
The frenzied buildout of data centers is driven by the scaling laws of AI and by exploding demand as the technology gets wedged into everything from anime girlfriends to fitness apps. But the public may shoulder the costs of all this construction for years to come, as communities hosting the power-hungry facilities grapple with soaring energy bills, water shortages, droning noise, and air pollution.
Deep Dive
Artificial intelligence
10 Things That Matter in AI Right Now
MIT Technology Review's authoritative overview of the 10 technologies, emerging trends, bold ideas, and powerful movements in AI in 2026.
A new US phone network for Christians aims to block porn and gender-related content
Launching next week on T-Mobile's network, the cell plan takes a nuclear approach to online safety.
Musk v. Altman week 1: Elon Musk says he was duped, warns AI could kill us all, and admits that xAI distills OpenAI’s models
Musk kept his cool, and OpenAI’s lawyer bulldozed him with piercing questions about his motivations for suing the company.
Three reasons why DeepSeek’s new model matters
The long-awaited V4 is more efficient and a win for Chinese chipmakers.
Stay connected
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.