Construction on Fresno’s high-speed rail station could begin as early as 2027. Rendering of the planned Fresno station courtesy of the High Speed Rail Authority

What's at stake?

High-speed rail executives are looking for more public-private partnerships to increase the financial viability of the $120 billion train. AI data centers could be a part of the solution.

But communities across the US are fighting back against the water and power-thirsty centers, saying they take more than the economic benefits they provide.

Could Fresno soon be home to thirsty AI data centers? Recent remarks from California’s top high-speed rail executive suggest that it’s on the table.

Ian Choudri, the CEO of the California High-Speed Rail Authority, came into his position last fall with a tall task: make the beleaguered project profitable. Just a few months after he started, a new federal administration, hostile to California’s high-speed rail, signaled an end to federal funding. 

The Trump Administration made good on that promise this July, officially yanking $4 billion of funds pledged to the project. Last week, the House Committee on Oversight and Government Reform announced an investigation into the project.

A new project update, released last Friday, provides a glimpse into the authority’s new vision to regain financial viability through more private-sector partnerships.

And while the report didn’t say it outright, Choudri has hinted that data centers fueling the growing artificial intelligence industry – could be coming to the Fresno area.

“We’re getting contacted by Silicon Valley investors now asking, ‘Hey, can we move data centers into Fresno and plug into your power grid that’s renewable?’” he told Forbes, in an Aug. 13 story.

AI data centers have been exploding across the US, but there’s currently none in the San Joaquin Valley, according to a Bloomberg analysis from May.

That could change soon. A PG&E executive told Reuters in May that they’re seeing a 40% increase in demand for data centers, with proposed developments in the Fresno area.

They’re controversial because of the massive water and power demand they put on local infrastructure.

Large data centers can consume up to five million gallons of water per day, according to researchers at the Lawrence Berkeley National Laboratory. Medium data centers can consume around 300,000 gallons of water per day, or around the same consumption as 1,000 households.

Fresno’s daily water consumption is around 108 million gallons of water per day, based on 2020 data.

A rail authority spokesperson did not comment on how they’re planning to address any potential water supply challenges that come with AI data centers – but noted that more details on any data center strategy will come in the authority’s next business plan, to be released in 2026.

Fresno Mayor Jerry Dyer was not available for comment, nor was Assemblymember Joaquin Arambula, who also sits as an ex-officio member on the high-speed rail authority board.

Assemblymember Esmeralda Soria acknowledged the impact of data centers on water supplies in an emailed statement to Fresnoland, but said more information was needed.

Finding additional water to support thirsty data centers in the greater Fresno area and San Joaquin Valley could be challenging, as nearby farmers and rural communities are struggling to keep groundwater aquifers replenished, under the state’s groundwater law, in effect since 2014.

Several nearby communities still rely on water trucks and bottled water deliveries.

And Fresno County well owners will soon have to register their wells so regulators can start tracking their water consumption.

Some communities have started to push back as AI data centers creep into their periphery, from Santa Clara to Chandler, Arizona. In rural Georgia, residents reported water shortages and dry wells after a Meta data center was built nearby. 

But others welcome the economic benefits

In some places, they’re built with union labor. And they can come with a few hundred jobs to keep the machines running. Eastern Oregon leaders approved a controversial $1 billion tax incentive deal to lure five new Amazon data centers in 2023.

But data centers aren’t the only possibility the rail authority is scheming around to improve the train’s financial viability.

Eliminating the Merced stop – due to the expense of creating a new spur off the mainline – was suggested in the authority’s recent report as one way to cut costs, a move that Soria called “shocking,” in a statement to Fresnoland.

“This latest proposal from the HSR is nothing more than a broken promise, and a slap in the face to those in the Central Valley,” she added.

The latest report estimates that the Merced to Bakersfield segment will be complete by 2032 – unless legislators decide to switch priorities.

According to their projections, the project’s ability to make money lies in its ability to connect to the mega regions of Southern California, via Palmdale, and Silicon Valley, via Gilroy.

But the Central Valley segment, despite lower ridership projections, still contains value for the authority, through its now-massive land holdings.

The report lists a few ideas to monetize that land, some that have been around for awhile: fees from HSR-owned parking structures next to stations; retail development at the station itself; station naming rights.

Up to 10% of the authority’s land holdings could be sold off to help pay for the project, according to the report.

In the meantime, Gov. Gavin Newsom and some California legislators have been angling to replace the lost federal funding by pledging a $1 billion annually from the state’s cap and trade program to the project.

Go back

Your message has been sent

Warning

Go back

Your message has been sent

Warning
Warning
Warning
Warning

Warning.

I created Fresnoland so we can make policy public for everyone.

2 replies on “Could Fresno soon be home to thirsty AI data centers?”

Comments are closed.