Texas’ electrical grid made national headlines in the winter of 2021 when the state experienced statewide power outages. The Electric Reliability Council of Texas, or ERCOT—the state’s power grid operator—was caught completely off-guard when a winter storm exposed the organization’s lack of severe weather preparedness. Embarrassed, ERCOT developed a roadmap to increase the reliability of its energy delivery system.

But guaranteeing a reliable flow of energy from the state’s generating plants to the homes and businesses of Texan residents has proven more difficult than expected. ERCOT recently announced that if a comparable storm were to hit the Lone Star State this winter, there is an 80% chance that they would again experience blackouts during peak hours.

Failure to resolve Texas’ power grid bottlenecks is perhaps not entirely ERCOT’s fault. Demand for energy in the state has ballooned in recent years thanks, in part, to the explosion and hype around artificial intelligence.

Texas is a microcosm of the threat artificial intelligence poses to the world—lack of energy security for households, an accelerating climate crisis, and the consolidation of corporate power.

A significant expansion in the supply of data centers is needed to meet artificial intelligence demand because the systems rely on vast computational power. AI systems are energy hungry—for example, a query using ChatGPT takes 10 times the energy of a traditional Google search.

There are approximately 342 data centers currently operating in Texas. Running these systems non-stop, daily, for 24 hours, requires a gargantuan amount of electricity. As a result, ERCOT has identified data centers as presenting a potential energy emergency alert risk at night and during early morning hours this winter. Data centers are currently consuming close to 9% of the energy produced in Texas, and it is putting a significant strain on its power grid.

Texas is a microcosm of the threat artificial intelligence poses to the world—lack of energy security for households, an accelerating climate crisis, and the consolidation of corporate power.

It is estimated that new AI servers that will be sold in 2027 will consume between 85 and 134 terawatt-hours annually. This is comparable to the electricity consumption of 18 million people living in the Netherlands.

Data centers can be the size of multiple football fields, and they are dependent on energy-intensive cooling systems that prevent computer servers from overheating and crashing. Water is an important component for cooling towers, and lots of it is needed to bring down the temperature of server equipment.

An analysis conducted by The Washington Post and the University of California, Riverside found that generating a 100-word email with ChatGPT-4 requires the use of at least one water bottle. Multiply this by millions of queries that are inputted each day, and you can get an idea of the scale of the tech sector’s water consumption.

In regions where water is already scarce, the unquenchable thirst of Big Tech hits especially hard. In 2021, a Google-owned data center in The Dalles, Oregon consumed nearly one-third of the town’s water supply even as the community grappled with a prolonged drought. As one frustrated resident aptly put it, “Google has become a water vampire, basically.”

This surge in AI energy demand use has led utilities to build new gas plants and delay the retiring of current fossil fuel infrastructure, thereby forestalling progress........

© Common Dreams