To power their data centers, American tech giants are building a "shadow energy system"
In the USA, amid the artificial intelligence boom, a parallel energy system is beginning to emerge: tech companies are building data centers along with their own power plants, effectively bypassing the general power grid, writes Devby.io based on a publication by The Washington Post.
Illustrative image. Photo: freepik.com
One of the illustrative cases is the GW Ranch project in Texas: dozens of hangars with computing hardware that will consume more energy than all of Chicago, but will not connect to external lines at all. Electricity there will be generated on-site using gas and partially solar energy. Such projects are becoming part of a rapidly growing "shadow power grid" across the country.
The reason is that the classic infrastructure cannot keep up with demand. The growth of data centers has put pressure on grids and faced resistance from authorities and local residents, so big tech bypassed the issue by building its own capacities.
According to industry analysts, dozens of such projects are planned in Texas, New Mexico, Pennsylvania, Wyoming, Utah, Ohio, and Tennessee, with some already under construction. The race involves major players — Meta, OpenAI, Oracle, and even energy companies like Chevron. Projects already approved are collectively capable of generating energy comparable to New York's consumption — and not just once.
The key nuance is that almost all these capacities are built on natural gas. Renewable sources like solar and wind are too unstable without grid support, so the focus is on gas generation, which automatically increases carbon dioxide and other pollutant emissions. According to experts, this could seriously undermine the climate goals that tech companies themselves talk about. At the same time, some projects are being implemented with less efficient and more polluting equipment, as modern gas turbines are booked for years in advance.
There are also purely engineering issues. Data centers operate around the clock, while gas power plants can be down for maintenance for a third of the year or more. Ensuring a stable 24/7 energy supply without a grid is a task that has not been fully resolved even with virtually unlimited budgets. In addition, such demand from big tech could drive up equipment prices and indirectly increase tariffs for those who remain on the general power system.
A separate line of tension is local communities. In several states, regulations have already been softened for such projects, effectively depriving local authorities of the ability to block construction. As a result, residents learn about future large-scale gas power plants only after the fact and protest. Critics speak of a "gold rush," where decisions are made opaquely, and the consequences — environmental and infrastructural — fall on the regions.
For the industry, however, this is a forced step. Connecting to the grid can take years, and in the AI race, that's too long. A telling example is Elon Musk's xAI data center project in Memphis. It was launched in 2024 on the tightest schedule, and to avoid waiting for a grid connection, the site was partially powered by mobile gas generators. This allowed the infrastructure to be set up in months. Within the industry, this is perceived as a signal: whoever secures their energy faster wins the race.
As a result, the largest technology companies are gradually transforming into energy companies as well. Proponents of the idea believe that such a model will reduce the burden on the grid and protect ordinary consumers from rising tariffs. Skeptics, however, warn that this creates a parallel infrastructure with high emissions, technical risks, and long-term consequences for the entire US energy system.