Trump Actions and Statements on California Water Dismay State Experts
Last week, President Trump made statements and took actions regarding California’s water that dismayed state officials and experts. First, on Sunday, January 26, the new president released an order (dated two days earlier) that directed federal authorities to maximize “water deliveries” in the state and override what he called disastrous policies of California officials, some of which were meant to preserve wildlife. He ordered that federal officials take steps to ensure that state agencies do not interfere with the Bureau of Reclamation’s operations to maximize water delivery to high-need communities.
The next day, he claimed on social media that the U.S. military had entered California and “turned on the water” under emergency powers. However, state authorities quickly responded, saying that there was no invasion of California by the military, and what actually occurred was that federal water pumps were restarted after being offline for maintenance for three days. Maintenance is carried out routinely.
Then, late last week, Trump administration officials began releasing water from two dams in central California that the president falsely said would have prevented the Los Angeles wildfires. According to the Washington Post, Gene Pawlik, the spokesperson for the Army Corps of Engineers, said in a statement that the releases were intended to ensure California has water available to respond to the wildfires.
However, the newly released water will never make its way to L.A. because it would have to travel over a mountain range for which there is no path. Most of that water, which Trump falsely takes credit for now delivering to stop fires, will flow to agriculture in the San Joaquin Valley, and as the Public Policy Institute of California explains, there is no physical connection for that water to be put to use fighting fires in Southern California.
Still, the president insisted on his social media platform that, if people had listened to him six years ago about releasing water, there would not have been fires in L.A. His executive order of January 24 referred to empty reservoirs. However, the wildfires in Southern California were not a result of dry reservoirs. Instead, as has been widely reported, fire hydrants ran dry because municipal systems weren’t built to fight so many flare ups at once, not because of a lack of water. Southern California’s reservoirs were quite a bit above their historic average when the wildfires began, according to the Public Policy Institute of California.
Moreover, the water releases started last week could have created dangerous conditions. The New York Times reports they would threaten the stability of levees and flooded farms. The release dismayed water experts, who said it would not benefit farmers since they have little need for it now, given they irrigate in the spring and summer.
As local water managers learned of the releases, they rushed to prepare for gushing water they had not requested. Authorities moved to warn people of possible flooding and to move equipment after they got only a one-hour warning from the federal government, according to Politico. Officials were relieved when the Army Corps of Engineers scaled back the decision to dump massive amounts of water from the two dams, local news outlet, SJV Water reports.
The president’s statements inaccurately make a complicated water picture look simple. He claimed that his executive order to open up the pumps and the valves in the north part of the state would make water pour down to Southern California. A spokesperson for Governor Gavin Newsom told E&E News that Trump’s attempt to connect water management in Northern California to firefighting in Los Angeles has zero factual basis.
A Deep Dive into DeepSeek—Is It Really Better for the Environment?
President Trump started his term by touting a $500 billion joint venture among AI companies. The so-called Stargate Project would build up to 20 large data centers around the country to advance U.S. AI dominance.
Most of us know AI through tools like OpenAI’s ChatGPT and Google’s Gemini, but up until last week we probably didn’t know about DeepSeek—a small Chinese startup that has upstaged the large language models (LLMs) built by U.S. tech giants. DeepSeek released its model R1 on January 20 and was in first place in the Apple App Store, bumping ChatGPT down to second place. Its sudden dominance sent tech stocks plummeting because it was allegedly built with a fraction of the cost and computing power—qualities that could mean a significantly lower impact on the environment.
Artificial intelligence has a voracious appetite for energy, as racks of computers in data centers gobble up large amounts of electricity—much of which comes from fossil fuels. AI is also very thirsty. The microchips needed for the calculations require a lot of water to manufacture—and not just tap water but “ultrapure” water, which is thousands of times cleaner than drinking water to remove any impurities that could damage chips. According to the World Economic Forum, it takes roughly 1,400 to 1,600 gallons of municipal water to make 1,000 gallons of ultrapure water. An average chip manufacturing facility uses ten million gallons of ultrapure water per day—equal to the daily drinking water use of 33,000 U.S. households.
Those AI chips emit heat as they respond to our burning questions and queries, so data centers need water to keep the servers and the buildings cool. According to a study by University of California, Riverside, ChatGPT gulps down roughly a half-liter (around 17 ounces) of water every time it responds to between 20 and 50 prompts. Increase that number if the facility is running on hydroelectric power or nuclear plants that require large cooling towers and by design are located next to rivers or water bodies.
Last, the extra energy generation needed by data centers can impact communities living near—and not so near—power plants. A recent study by Caltech and University of California, Riverside, said as tech companies race to provide AI services, the resulting air pollution in the form of lung-penetrating fine particles—those smaller than 2.5 micrometers—and other federally regulated pollutants, such as nitrogen oxides, is expected to steeply increase. The result could be as many as 1,300 premature deaths per year by 2030 in the United States. Total public health costs from cancers, asthma, and other diseases, as well as from missed work and school days are approaching an estimated $20 billion per year.
With that as background, what could be the environmental impact of DeepSeek when compared to U.S. models? H2O Radio spoke to Shaolei Ren of UC, Riverside, a researcher focused on AI and sustainability. Ren explained that DeepSeek uses less energy—and consequently less water—because it’s smaller and piggybacks on larger models in a method called “knowledge distillation” where it learns from their data sets without having to start from scratch.
Furthermore, instead of using the powerful NVIDIA chips that ChatGPT relies on, DeepSeek used an older generation chip that requires less energy. The model also uses a mixture-of-experts (MoE) architecture. Put simply, it deploys multiple smaller models (called “experts”) that are only active when they are needed, reducing computational costs. (A good explainer video is available here.) Also, DeepSeek claims to have developed its latest model for less than $6 million, compared to billions spent by U.S. tech companies.
Ren adds that DeepSeek is open source, which means it could be replicated by researchers to vouch for its lower environmental impact. But that impact will only be lower if people actually use it over other models. As the New York TImes reports, researchers testing its capabilities have expressed concerns about DeepSeek's ties to the Chinese government and its being susceptible to censorship.
Ren also points to a phenomenon known as the "Jevons paradox” in which the environmental benefits of DeepSeek could shrink as the number of people using the tool grows, in the same way owning a fuel efficient gas car could lead a person to drive more, but he says that DeepSeek’s smaller size would still make its impact lower than U.S. models.
Every product has an environmental cost, but the DeepSeek-type models offer the chance to counter the argument made by some tech companies that we must build more natural gas plants to power data centers and keep the U.S. competitive in the AI race. Many experts suggest DeepSeek’s modest size could mean it would eventually run on smartphones.
Ren suggests that AI could be deployed sustainably to offset its impacts, such as using recycled or recirculating water to cool facilities and use renewable energy sources. He notes that tech companies have claimed their facilities are carbon neutral when in truth they are buying wind or solar offsets and taking their electricity from the grid which is at least in part fossil-fuel based. He adds that AI can—and should—be used responsibly to address the problems it creates and advance society by learning from past mistakes.