Definition
Gigawatt is a gigawatt is one billion watts of electric power, and AI companies use it to size data centers and power deals: a one-gigawatt campus draws about as much power as a large nuclear reactor produces.
A gigawatt (GW) is a unit of power equal to one billion watts, or 1,000 megawatts. In AI, it has become the standard way to say how big a data center, a power deal or a build-out plan is.
Watts measure how fast electricity is used at a moment. Energy is power over time, measured in watt-hours. A facility rated at one gigawatt can draw up to a billion watts at any instant. Over a full year of nonstop use, it would consume roughly 8.76 terawatt-hours.
Think of a faucet. The gigawatt rating is how wide the faucet can open. The watt-hours are how much water ends up in the bucket.
A common rule of thumb is that one gigawatt of continuous supply serves roughly 700,000 to 750,000 average US homes. The exact figure moves with assumptions about how hard the site runs, so treat it as an estimate. It is also often compared to the output of a single large nuclear reactor.
Early data centers were sized in tens of megawatts. Plans for AI campuses are now announced in gigawatts, because training and serving large models takes enormous numbers of GPUs, and each one draws hundreds of watts before cooling is counted.
Chips are no longer the only limit. Getting a gigawatt of power connected can take years, because it needs generation, transmission lines and permits. That is why labs now talk about power the way they talk about compute: as a scarce input they have to secure ahead of time.
When a company says it has contracted "1.7 gigawatts" of anything, it is describing peak capacity, not a finished data center. Equipment still has to be installed, connected and filled with hardware.
Ask three questions. Is it capacity or actual consumption? Is it built, under construction or only planned? And is it power the company generates itself or buys from the grid? The answers change what the number means for inference capacity you can actually use.
A gigawatt describes the size of an AI data center and its appetite for electricity. How much of that power turns into useful AI work depends on the chips, the cooling and the software. Capacity announcements are plans, and plans lag the hardware by months or years.
Estimates cluster around 700,000 to 750,000 average US homes if the supply runs continuously. The figure varies with how much of the time a site actually runs.
A gigawatt measures power, the rate of use at a moment. A gigawatt-hour measures energy: one gigawatt sustained for one hour.
Microsoft says Maia 200 is in production and Maia 300 starts multi-gigawatt deployments in 2027, making its own chip the primary inference system for its products. GitHub Copilot and Foundry move to Maia by year end, so your requests will soon run on Microsoft silicon.
The Nasdaq fell 5.5% on Nov. 5, its worst day since April 2025, and Nvidia dropped more than 9% as the midterms turned against data centers. The market is betting on less compute, and compute is what you pay for.
Meta launched about $35 billion of stock and convertible preferred offerings on Oct. 26 and plans a $60 billion stock program from 2027, with Nvidia, AMD and Broadcom indicating up to $5 billion. The AI buildout has outgrown even Meta's cash flow.
Aramco closed its purchase of 49% of Saudi AI company Humain on Nov. 1, putting in about $2.6 billion in cash plus AI and compute assets, while PIF keeps 51%. Saudi Arabia now has one AI champion with an oil giant's balance sheet behind it.
Anthropic struck a deal worth about $3.4 billion for 1.7 gigawatts of gas-fired generators from Texas supplier OnePWR Solutions, the Financial Times reported Sunday. The lab behind Claude is now buying electricity the way it buys chips: in bulk and ahead of time.