Justine Calma | The Verge

The fossil fuel comeback is here

2026-10-19T17:48:01-04:00

The US is building more fossil fuel infrastructure than it has in a decade while renewable energy growth slows down, a new report shows.

Electricity demand is also climbing faster than it has in more than a decade with more data centers, factories, and electric vehicles plugging into the grid. Despite those growing demands, the Trump administration has sought to hobble the wind and solar industries in favor of fossil fuels.

America’s energy landscape is already shifting as a result, according to the report published last week by the nonprofit Environmental Defense Fund and research firm Atlas Public Policy. If all the gas projects in the development pipeline complete construction, they could lock in more pollution and higher electricity prices for decades, the authors warn.

“The U.S. power sector is at a new inflection point”

“The U.S. power sector is at a new inflection point, with decisions made today shaping the cost, reliability, and emissions profile of the grid for decades to come,” the report says.

While the US still generates 60 percent of its electricity from fossil fuels, renewables have quickly gained ground. Wind and solar produced more electricity together than coal for the first time in 2024. Solar was the fastest-growing source of new electricity generation that year, accounting for 81 percent of added capacity. Globally, solar has become the cheapest power source in most of the world.

Then Donald Trump stepped into his second term as president with a plan to “drill, baby, drill” for fossil fuels, his preferred power source for energy-hungry data centers. His administration has sought to kill renewable energy projects by issuing stop work orders, blocking permits, and removing tax incentives. We’re still seeing a rush of new solar and wind farms driven by relatively low costs, although many of them make up the last wave of projects that benefited from Biden-era incentives.

The winds are shifting in favor of gas-fired power plants. Plans for new fossil fuel infrastructure climbed from just 5 percent of all new capacity in 2024 to 30 percent by the second quarter of 2026, according to the new report, which analyzes data from the US Energy Information Administration. If the trend continues, fossil fuels could make up 70 percent of the development pipeline by the end of the decade. Planned fossil capacity would reach 68 gigawatts by 2030, reflecting a 45 percent jump from construction plans just six months ago.

Many of those plans could still fall through, particularly if forecasts for AI data center energy demand are overblown. But for now, renewable energy projects are being canceled at a faster clip than fossil fuels, at 4.9 percent compared to 2.9 percent of projects, respectively.

New utility-scale solar and wind farms still beat gas when it comes to cost, coming in at $68 to $69 per megawatt-hour on average compared to $90 per megawatt-hour for new combined-cycle natural gas plants, the report finds. Fossil fuels are also subject to increasing price volatility, particularly since the US instigated war with Iran. Gas and coal power plants also emit air pollutants and the greenhouse gases causing climate change, leading to more health costs. Electricity bills have risen across the nation as utilities upgrade power grids and build new energy infrastructure to meet rising demand.

Microsoft has a new machine learning tool to improve weather forecasts

2026-10-14T12:05:21-04:00

Old-school physics-based weather forecasting and even new AI models typically become less reliable beyond two weeks, a problem Microsoft claims to have solved using machine learning.

Researchers with the company say they’ve found a new, cost-effective way to make more accurate weather forecasts two to six weeks in advance — allowing farmers and emergency responders to plan ahead with greater certainty. The findings were published today in the journal Nature, and the tool is now open-source software.

Microsoft joins Google, Nvidia, and other companies developing AI tools for weather forecasting. Instead of creating a new standalone weather model, Microsoft’s latest research in this realm uses machine learning to correct bias in existing models. It’s a more accessible way for resource-strapped weather agencies to improve accuracy using AI, the researchers hope.

It’s a more accessible way for resource-strapped weather agencies to improve accuracy using AI

While physics-based modeling is a resource-intensive endeavor running on supercomputers and generative AI depends on powerful GPUs, Microsoft Research New England senior principal researcher Lester Mackey says the new tool can run just on CPUs. “I have a lot of hope both for the increased resource efficiency of weather forecasting and the extended reliability of forecasts, especially for extreme weather,” Mackey tells The Verge.

Physics-based and AI-based models run into similar problems when they start predicting the weather two weeks or more in advance. Traditional models attempt to simulate the complex physics of Earth’s atmosphere, while AI models look for patterns in historical data to make forecasts. In both approaches, models rely on approximations to fill data gaps, whether from regions with sparse weather stations or complex atmospheric phenomena that scientists are still working to understand. The ways in which these models make up for those gaps in data can become a type of systematic bias. The resulting errors compound when making forecasts further ahead in time, making these predictions less precise.

Microsoft calls its method Probabilistic Bias Correction (PBC). The tool uses machine learning to identify systematic bias in a model and make corrections. One example Mackey gives is to correct for changes in climate. Greenhouse gas emissions from fossil fuels are heating up the planet, but a weather model might make predictions based on historical records that can bias it toward colder temperatures than the world is actually experiencing today.

The researchers found that PBC improved the accuracy of a leading physics-based weather model by the European Centre for Medium-Range Weather Forecasts (ECMWF) for 98 percent of precipitation and 92 percent of temperature targets. The machine learning framework also doubled the accuracy of ECMWF’s new AI forecasting system in the study’s analysis. Before the paper was published, the ECMWF had already awarded the authors’ research on PBC in its AI Weather Quest competition.

The new paper on PBC published by researchers at Microsoft, MIT, ECMWF, and other institutions “is a good example of new ideas being developed for the benefit and consideration of the community,” Matthew Chantry, ECMWF’s strategic lead for machine learning, said in an emailed comment to The Verge.

“As always, turning ideas into reality [in] an operational process is a most challenging step,” Chantry added.

The amount of compute and infrastructure needed for weather forecasting has often left technological advances out of reach for many agencies in economically developing countries. The paper’s authors, however, think PBC can be different. “One of the things that’s really exciting to me about models like PBC is how lightweight they are, how easy to train, that scientists without huge pools of resources can work on trying to configure [these] models and make them perform better in specific localities,” says Genevieve Flaspohler, one of the authors of the paper.

Flaspohler worked on PBC as an intern at Microsoft with Mackey, and now she’s the executive director of the nonprofit Rhiza Research, which is working with the Kenya Meteorological Service Authority — with funding from the Gates Foundation — to see whether AI can help get more accurate early warnings out to people. PBC is one of the methods they’re testing ahead of what’s expected to be a more extreme rainy season during a record-breaking El Niño.

Google’s power-hungry data centers crave nuclear energy

2026-10-06T16:32:10-04:00

Google announced a new agreement to update six nuclear power plant sites across the US as the tech giant seeks to generate more electricity for its power-hungry data centers.

Google signed the 20-year deal with Constellation, the leading nuclear power plant operator in the US. The power purchase agreement is meant to guarantee the revenue needed to cover the cost of efficiency upgrades at existing nuclear plants and “represents more than $4.3 billion of new investment by Constellation,” according to the press release.

Over the next five years, the companies expect new equipment at each site to collectively lead to 890 megawatts of additional nuclear capacity. That’s roughly equivalent to the amount of electricity one large new nuclear reactor or three small modular reactors might generate, according to Google and Constellation.

PJM has struggled lately to meet growing demand from AI data centers

They plan to modernize 11 reactors at six sites in Illinois, Pennsylvania, and New Jersey. Each is connected to the PJM Interconnection grid, the largest power grid in the US, which spans 13 states and the District of Columbia — including the nation’s largest hub for data centers in Loudoun County, Virginia.

PJM has struggled lately to meet growing demand from AI data centers. An independent market monitor filed a complaint last year with the Federal Energy Regulatory Commission (FERC) saying that PJM was planning to allow huge data centers to connect to the grid that it could not reliably serve, which could lead to blackouts.

In January, the Trump administration and governors from all of the states where PJM operates issued a joint statement calling on PJM to hold an “emergency” power auction to support the buildout of new power plants through long-term contracts. PJM postponed its auction last week following a directive from FERC to revise its proposal in order to ensure stronger protections to prevent data centers from raising electricity bills for other customers. Electricity rates have skyrocketed in states connected to PJM, including New Jersey, where prices have climbed 54 percent over five years.

Along with the nuclear power deal, Google also inked a separate, much larger 15-year energy supply agreement with Constellation meant to guarantee continued revenue for other existing power plants Constellation operates within PJM, regardless of fuel type. Nuclear energy made up 70 percent of Constellation’s total generating capacity last year, and another 8 percent came from renewables. Gas and oil made up the remaining 22 percent, according to the company’s latest sustainability report. Gas also accounts for the largest chunk of PJM’s fuel mix. At 2,700MW, the second deal is much larger than the nuclear power purchase agreement.

Data centers are driving a surge in new gas infrastructure projects in the US, moving many tech companies further from their sustainability goals. Google’s carbon emissions rose by 18 percent last year along with a jump in electricity demand of 37 percent, according to the company’s latest sustainability report.

The planetary wave headed for California isn’t what you think it is

2026-10-06T00:33:57-04:00

An aerial view of equipment moving fresh sand during an emergency replenishment project along eroding shoreline at North Beach on August 15th, 2024, in San Clemente, California. | Photo: Getty Images

A rather strong “planetary wave” recently slammed into South America and is making its way toward California. In case you’re worried about headlines warning of the supposedly “freak” phenomenon “bringing trouble,” it’s worth understanding what this wave really is and isn’t.

Called a Kelvin wave, it’s carrying warm water to the Americas that’s typically trapped farther west. That’s raising sea levels along already vulnerable coastlines. These kinds of waves are common. When they do arrive, they can exacerbate existing problems. And this particular Kelvin wave could have more dramatic consequences than usual.

But “If you’re going to be out there looking for it, you’re not going to see it,” says Richard Seager, a research professor at Columbia University’s Lamont Doherty Earth Observatory.

“If you’re going to be out there looking for it, you’re not going to see it.”

A photo accompanying The New York Times’ reporting on the Kelvin wave, depicting a surface wave smashing into someone’s beachfront condominium, is misleading, he notes. The National Oceanic and Atmospheric Administration (NOAA) says a Kelvin wave is more like “waves in your bathtub, which slowly slosh around.” Water levels rise and fall in peaks and valleys, but they’re not breaking violently against a shore.

Kelvin waves are a typical signal of an El Niño, although they can also take place during other phases of the El Niño-Southern Oscillation (ENSO) pattern. During other phases, trade winds over the Pacific Ocean push warm waters near South America westward toward Asia. That causes an “upwelling,” pulling cooler water from below up toward the surface of the ocean along the equator. As a result, the eastern equatorial Pacific is typically cool, while the western equatorial Pacific is warm.

During an El Niño, those trade winds die down, allowing the warm water that piled up in the western Pacific to flow back east in the form of a Kelvin wave. It also slows down the upwelling, limiting how much cold water can reach the surface. Traveling east along the equator, the Kelvin wave hits Peru and Ecuador first before bouncing off the coast and moving north toward British Columbia and south toward Chile. The phenomenon causes temporary sea level rise, usually between 6 to 12 inches.

We’re also in the midst of a “super” El Niño that’s expected to be one of the strongest on record because of how warm sea surface temperatures are getting. Temperatures in the eastern Pacific reached 5 degrees Celsius above normal on October 4th. “That is whopping big,” Seager says. “As you would expect for a big El Niño event, it’s going to be a big, powerful Kelvin wave.”

Every inch of sea level rise can heighten the risk of flooding during storms and high tides. El Niño can push the North Pacific storm track southward, bringing more storms into Southern California, Seager explains. California Gov. Gavin Newsom declared a state of emergency in September to prepare for a particularly severe El Niño this season. Rising sea levels with climate change are already inundating seaside roads, railways, and neighborhoods in California.

“If you already don’t have a lot of infrastructure that is built to protect your community and your coastline, you’re going to feel it … because 6 to 12 inches is not trivial,” says Emily Yam, senior director of learning and public engagement at the Aquarium of the Pacific in Long Beach, California.

The extra warm water also has consequences for marine life. El Niño can be an exciting time to go fishing and whale watching, with many warm water-loving species traveling farther north than they typically would. NOAA notes that a 151-pound striped marlin broke a state record when it was caught in August off the coast of Washington. There have also been unusual sightings of hammerhead sharks and devil rays recently in Southern California.

Unfortunately, hotter temperatures also put stress on coastal ecosystems. Yam says scientists are bracing for a potential die off of California’s iconic kelp forests. They thrive in temperatures around 50 to 60 degrees Fahrenheit, but the water has reached well into the 70s lately. The Kelvin wave is piling hot water into an area already suffering from a marine heatwave since May of last year.

The temporary sea level rise is expected to linger through the start of winter. El Niño, which began in June, is forecast to weaken rapidly going into 2027.

If a data center is camouflaged in the woods, will anyone hate it?

2026-10-02T12:44:06-04:00

San Antonio City Councilmember Ric Galvan remembers when data centers first arrived in his city in the 2000s, looking like relatively unassuming office buildings. But that’s changed in recent years, as data centers have grown into massive “hyperscale” facilities spanning millions of square feet and housing the physical infrastructure for AI.

There are now more than a dozen data centers in Galvan’s roughly 55-square-mile district, and he expects that number to grow to around 20 over the next few years if proposed projects are completed. They’re clustered along two main roads, concentrating nuisances in the same neighborhoods: traffic and trees cut down during construction, followed by constant bright lights and noise when the data center is operating.

A lot of the area used to be green fields, he explains, allowing many residents to feel close to nature. Now, he says, “that [feeling] gets taken away pretty quickly when you make a gigantic data center.”

“That [feeling] gets taken away pretty quickly when you make a gigantic data center.”

Across the US, more and more communities are rising up against these hulking data warehouses — protesting the effects they could have on quality of life, pollution, water, and electricity bills. Tech companies have responded with promises to mitigate any potential harm by paying for infrastructure upgrades needed in local water and power systems.

Microsoft is also taking a different approach to soften the impact its data centers have. It announced plans to make its data centers less jarring to local communities and the environment through what it describes as “biomimicry.”

Biomimicry is sort of a catch-all term for many of the ways technologies take inspiration from nature, from “painless” needles that imitate a mosquito’s proboscis to naturally ventilated buildings modeled after termite mounds. In Microsoft’s case, biomimicry involves “blending our datacenters into the natural landscape” to “create resilient ecosystems that support biodiversity, strengthen community ties, and help our datacenters become better neighbors.”

Essentially, Microsoft wants to make its data centers greener, literally, with more gardens and native trees on-site. It’s also promising to invest in projects that repair damaged ecosystems where it has data centers. The company thinks these efforts will limit problems its facilities have caused for neighbors, like noise pollution and the loss of green space.

It’s a nice gesture, advocates and sustainability experts tell me. Microsoft has tried to set itself apart from its rivals by engaging more with communities where it builds data centers. The strategy has appeased some local concerns — depending on whom you ask.

But these data centers are expanding so fast that they’ve simultaneously razed thousands of trees, and Microsoft’s greenhouse gas emissions are growing with the energy demands of AI. Advocates and residents near these data centers want to see more from the company to limit its larger impact on communities and the environment.

Zooming in

“Maybe … they’re trying to just make sure that they’re covering their bases,” Galvan says. “But also, I think if that’s the case, that’s fine,” he adds — as long as Microsoft and other data center operators are actually addressing concerns residents have brought up for years around pollution, water use, and the destruction of natural green spaces.

Microsoft has operated data centers in the Greater San Antonio area for almost a decade, and is spending hundreds of millions of dollars to build several more. In 2022, the company upset residents in San Antonio when it cut down more than 2,600 trees across a 33-acre expanse to build a data center. That — along with Microsoft’s requests for a zoning change and tax incentive — was a big controversy, Galvan says.

“We’re very proud of our trees here,” Galvan says, pointing out that the tree canopy also helps to mitigate heat, which in Texas is a growing problem. “So it was very frustrating to see that completely decimated.” (In an email, Microsoft spokesperson Jennifer Lyons that the company is “investing in enhancement projects that includes 400+ acres of woodland and pollinator prairie restoration” in San Antonio with a goal of leaving “land more productive than we received it.”)

Microsoft says it has more than 300 data centers in more than 34 countries and started biomimicry projects at a data center in the Netherlands in 2022. Now it plans to bring “nature-inspired solutions to restore and enhance natural ecosystems” to more than 20 sites in the US and Germany, and says it will do the same for all new projects in the US and “a growing number around the world.”

“We’re actually customizing the solutions to fit in with the local habitat and enhance what’s already there and restore what might have been lost in the past,” Kaitlin Chuzi, director of global land strategy at Microsoft, says.

In its September announcement, Microsoft references successes it has had in Mount Pleasant, Wisconsin, where the company is repurposing the shuttered Foxconn LCD factory property for its own data centers. In 2023, Microsoft funded a $4.2 million project to restore the flow of a creek that runs through its property, working with the Root-Pike Watershed Initiative Network (Root-Pike WIN) to do so — an example of the company working with locals to restore ecosystems.

Root-Pike WIN executive director Dave Giordano was also pleasantly surprised to see Microsoft outfit stormwater ponds on its property with forebays to prevent sediment from escaping. “That’s rare around here,” he says. “Not all data center companies are the same. Not all of them think like Microsoft.”

Councilmember Galvan also acknowledges that Microsoft has made itself more approachable by employing a local representative for the company who responds to community concerns. “That’s very different than other data centers in our city,” he says. In January, Microsoft introduced a five-point plan called “Community-First AI Infrastructure” meant to prevent its data centers from overwhelming a locale’s ability to cope with the company’s energy and water demands.

“Not all data center companies are the same.”

In San Antonio, Microsoft is also considering planting pollinator gardens to help monarch butterflies during their annual migrations. Another tactic Microsoft is considering at more sites as part of its biomimicry initiative is to create buffers of plants around its data centers to muffle some of the noise from humming servers and generators and to “minimize the visual impact on communities near our data centers,” says Chuzi.

In other words, it seems Microsoft aims to make its data centers less ugly and intrusive. But the mismatch between Microsoft’s smaller-scale nature-based proposals and the company’s broader impact on the environment and already limited resources leaves some advocates unsatisfied.

“I have nothing bad to say about nature-based solutions that are sensitive to the native ecosystem,” says Adrian Shelley, Texas director at consumer advocacy group Public Citizen. “All of that is good and is positive. However, I think that it’s, you know, basically lipstick on a pig.”

Shelley is particularly upset about pollution from new gas plants built to power data centers. In June, for example, Microsoft inked a 20-year agreement with Chevron to power a new AI data center campus in West Texas. The project is expected to be “among the largest co-located natural gas power and data center developments” in the nation, according to Chevron.

“I cannot really see any level of biomimicry investment that makes a gas plant a desirable neighbor,” Shelley says.

Proposed data centers are driving up electricity demand, spurring the development of new gas infrastructure. But Shelley is afraid Texans will be left to foot the bill for abandoned energy projects if the AI bubble bursts.

With Public Citizen’s backing, Galvan proposed a moratorium in August on any new data center development in San Antonio. Since August, Texas Gov. Greg Abbott has also imposed statewide moratoriums on data center projects seeking to connect to the power grid pending audits on their electricity and water use.

“[Microsoft] is proposing to locate in regions of the state that just, from a larger perspective, cannot really handle this level of resource demand,” Shelley says, adding that the state ought to “severely restrict” these facilities’ access to drinkable water. Data centers often use water in cooling systems needed to keep servers from overheating. But the larger part of the facility’s water footprint typically stems from its electricity generation since fossil fuel and nuclear power plants need water for their own cooling systems.

Data centers in Texas consumed about 25 billion gallons of water last year, according to the Houston Advanced Research Center (HARC). That’s about 0.4 percent of total water use, but that figure is expected to climb. It’s difficult to predict a precise number because of the uncertainty around which proposed data center projects will ultimately materialize, and what kind of cooling systems they’ll use. But HARC estimates that data centers could consume between 0.5 to 2.7 percent of total annual water use by 2030.

Turning to renewable solar and wind energy can cut down that water use, according to Erin Kinney, manager of coastal resilience and communities and a senior research scientist at HARC. Kinney is also optimistic about the impact Microsoft’s biomimicry projects can have at the local level because the aim is to customize solutions to each location. “I actually think it’s a really exciting concept,” she says. The real test will be in how the company calculates and tracks the benefits and shares that with the local community. “Just as important as making something that fits the ecology, is fitting the community expectations as well,” she said.

Zooming out

Microsoft has a goal of becoming water positive by 2030, meaning it would restore more water than it uses. It similarly has a goal of becoming carbon negative, capturing more planet-heating pollution than it releases by the end of the decade.

Lately, however, the company’s focus on AI has jeopardized some of its sustainability goals. Its carbon pollution jumped 25 percent in 2025, according to its latest sustainability report. And in May, Bloomberg reported that Microsoft is considering abandoning its goal of matching 100 percent of its electricity use with renewable energy on an hourly basis — seen as one of the industry’s most ambitious strategies for encouraging renewable energy growth.

In 2024, after grappling with contradictions she saw between Microsoft’s sustainability goals and the company’s business deals, Holly Alpine left her role leading the company’s ecosystem restoration and climate resilience initiatives in dozens of communities near data centers.

While she was still at Microsoft, Alpine says, “as a skeptical and extremely passionate climate advocate, I was very excited about the biomimicry projects that I saw, they were extremely well designed to actually provide a positive benefit to the ecosystems where the data center is.” Even so, that was still not enough to outweigh the damage she sees stemming from Microsoft providing cloud services to oil and gas companies and developing AI tools that can be used to optimize fossil fuel production.

The company’s focus on AI has jeopardized some of its sustainability goals

She coauthored a research paper published in the journal Nature in August comparing potential AI-driven productivity gains for renewable energy and fossil fuel production, and found that AI ultimately causes more planet-heating emissions than it would help avoid. Big picture, she says, AI is enabling far more pollution than the company is addressing with even well-planned sustainability projects from data center to data center.

“In a period of rapid technological change, we remain focused on advancing our ambition to help address global sustainability challenges across carbon, water, waste, and ecosystems,” Microsoft’s Lyons said in an email to The Verge. “Delivering the infrastructure required to support AI at global scale requires a range of energy solutions in the near term. We continue to invest significantly in carbon free electricity and other decarbonization technologies to reduce emissions across our operations over time.”

The looming question I had for the company was if there was a point where Microsoft would say, “Maybe we just shouldn’t build a data center here” or “This isn’t the right place for this.” To that end, Chuzi was less clear in our interview.

“I think it is an ongoing learning process to understand what that means to be a good neighbor,” Chuzi says. “What it’s going to take to make sure that we are welcome in a community and to earn that place and that part in the community over time.”

Trump finalizes rule to make cars less fuel efficient

2026-09-28T14:39:50-04:00

President Donald Trump announces plans to weaken fuel efficiency and emissions rules for cars and trucks in the Oval Office at the White House in Washington, DC, on December 3rd, 2025. | Photo: Getty Images

The US Department of Transportation finalized its plans today to weaken fuel efficiency standards, calling it “among the largest deregulatory actions under the second Trump Administration.”

It’s a nail in the coffin for Biden-era standards that would have required fleet average fuel economy to reach 50.4 miles per gallon by model year 2031. President Donald Trump’s plan requires 34.9 miles per gallon, not much higher than the 30.1-mile-per-gallon target previously set for model year 2024.

Consumer advocacy, health, and environmental groups are incensed

The Trump administration claims its plan would shave $1,300 off the average cost of a new vehicle, saving $138 billion over the next five years — though those figures have been disputed.

Automakers are pretty happy with the new rules. But consumer advocacy, health, and environmental groups are incensed. They argue that weaker fuel economy standards will wind up making Americans pay more at the pump, on top of increased healthcare costs and damages from climate change due to air pollution. The Biden administration’s standards that Trump trashed were expected to save $23 billion in fuel costs, and a car owner more than $600 in gasoline over the lifetime of a vehicle.

“For decades, fuel economy standards have ensured that new cars, SUVs and pickup trucks are more efficient, which saves lives and money,” the American Lung Association’s president and CEO, Harold Wimmer, said in an emailed statement. “There is no reason to weaken standards that are technologically feasible and have a direct benefit for the quality of the air people breathe.”

Jensen Huang talks about AI and climate change like a supervillain

2026-09-24T14:04:44-04:00

As Jensen Huang puts it, AI can help fight climate change — but only if it inflicts “an enormous amount of pain and suffering” first.

The Nvidia CEO discussed the future of energy and AI’s impact on our planet in the latest episode of The Ezra Klein Show. But his comments boil down to the same accelerationist spin we’ve seen from a variety of tech leaders as well as President Donald Trump. They promise that AI will be such a gift to humanity that it’s worth the damage caused by data centers currently running on dirty energy.

Huang’s extreme level of privilege is glaring

Huang’s extreme level of privilege is glaring in the interview. He’s right that there will be pain. People are already suffering from fires and floods exacerbated by climate change. People are already dying earlier because of air pollution from fossil fuels. But Huang, with a net worth of around $192.6 billion, will not pay the price of the pollution his company creates through energy-hungry chip factories and data centers.

Huang tells Klein, after complaining that the US hasn’t built out enough fossil fuel plants, that:

It’s kind of like, in order to save you, they’ve got to hurt you first — that’s the nature of surgery. They’ve got to cut you open to save you. They’ve got to inflict an enormous amount of pain and suffering on you so that they can save you. And so I think AI’s kind of like that.

That’s the bit that gives Thanos energy. Resources are too limited and so after disappearing half of humanity, the Marvel supervillain believes, things will be much rosier. (“I will shred this universe down to its last atom, and then create a new one, teeming with life that knows not what it has lost but only what it has been given,” he tells the Avengers.)

Huang goes on to say:

Over the next several years we have to unfortunately use fossil fuels, because we just don’t have enough sustainable energy to make a difference. And then after that, hopefully, we can transition to that.

For context, nearly every country on Earth agreed to limit climate change when they adopted the Paris agreement in 2015. The accord sets a goal of cutting greenhouse gas emissions in half by 2030 and reaching net zero around 2050. That can only happen with swift transition to carbon-free energy like renewables and nuclear. Here’s just a short list of terrible things getting worse when we stall that progress to pursue advances in AI:

Earth is expected to lose 99 percent of its coral reefs. Those reefs support coastal economies, and also serve as a buffer protecting communities from tropical storms, flooding, and coastal erosion.
One in 10 people around the world, 770 million people, already face “acute risk” from sea level rise.
Since 2001, peak daily wildfire growth rates have more than doubled in the western US. Wildfires are faster, more frequent, and more intense.

These effects are showing up in Huang’s backyard, even if he’s insulated from them. He mentions in the interview that he doesn’t know what his own address is. However, he reportedly bought a $7.5 million home on Maui in 2004 — before wildfires devastated the island in 2023. That year, an Nvidia spokesperson confirmed with The Verge that the Huangs “have owned the same home in Maui for many years.”

Huang’s property reportedly escaped damage, and the Nvidia spokesperson told me that the Huangs “donated significantly to local relief efforts.” But despite the philanthropy from tech billionaires with residences on Maui, the inferno also inflamed tensions over luxury properties’ hold on Maui’s water supply.

“Billionaires coming to Maui, buying up large swaths of land as their, I don’t know, seventh vacation home, it’s painful for our community,” Maui County Councilmember Keani Rawlins-Fernandez told me at the time. “It underscores the stark disparity between Indigenous people who want to mālama ‘āina, care for our land, and those who see Hawaii as a vacation destination.”

“Billionaires coming to Maui, buying up large swaths of land as their, I don’t know, seventh vacation home, it’s painful for our community.”

While greenhouse gas pollution from fossil fuels makes wildfires more dangerous by exacerbating hot, dry conditions, climate change isn’t entirely to blame for the flames that razed much of Lahaina on Maui. I reported at the time that previous generations of wealthy land grabbers primed the islands to burn. The US annexed the Kingdom of Hawaii in 1898, egged on by American plantation owners who wanted to evade import taxes. They cleared native vegetation for their businesses, allowing invasive grasses to take hold and make the landscape much more flammable than it used to be.

So when I hear Huang make the case now that there will need to be sacrifices in the name of AI progress, it reminds me that we’ve seen this playbook before. An industry that wants to grow at any cost. And nearby communities and their descendants are just supposed to live with the aftermath.

Huang mentions growing community opposition to data centers in the interview with Klein, and scolds other companies for not spreading the good news with residents about how efficient data centers purportedly are before they start construction. “If they don’t want data centers to be built in their town or whatever it is, then so be it. But if you’re going to build in their town, be sure to go there and let them know what’s coming,” he says.

“I am inevitable,” Thanos notoriously said — and apparently, so are AI data centers.

Even so, data centers don’t actually have to be as polluting as they are turning out to be.

In the interview, Huang also (unfairly) attributes record renewable energy purchases to AI. He blames climate action for stalling the construction of energy infrastructure, presumably for fossil fuels.

I think we just have to acknowledge that we got ourselves really gummed up in climate change and sustainable energy, and as a result, we just didn’t plan enough energy production.

Then he proclaims that AI is “the best opportunity we have” to “turn a corner on climate change, if you want a future that’s sustainable.” He sees AI accelerating demand for clean energy in the long term. But AI already needs a lot of energy now, and fossil fuels are what we’ve got.

“A more polluting AI buildout is not inevitable.”

But renewables were already the fastest–growing sources of new electricity generation before the recent generative AI boom. It’s generally cheaper to generate power from a wind or solar farm than a new gas or coal plant. But Trump, who lined his campaign coffers with money from the fossil fuel industry, is incentivizing the development of gas and coal plants to power data centers.

And now, data centers are in turn spurring the development of far more fossil fuel infrastructure. Air pollution associated with AI could lead to up to 1,300 premature deaths and more than $20 billion in public health costs by 2028, according to a study from UC Riverside, Caltech, and the Rochester Institute of Technology.

Nevertheless, “A more polluting AI buildout is not inevitable. It depends on the choices government makes now,” Marc Boom, a senior director of the Environmental Protection Network, a group made up of EPA alumni, said in a recent press briefing on the growing health risks posed by data centers.

It also depends on the attitudes of industry leaders like Huang. If other people’s pain and suffering is the price you’re willing to pay for so-called progress, that says less about what’s possible and more about your own personal priorities.

Data centers are black boxes, but California wants to change that

2026-09-23T14:22:57-04:00

California Gov. Gavin Newsom signed a slate of bills on Monday that could finally give communities better data — and more say — on how data centers impact their electricity bills and water supply.

As data centers invade a growing number of communities across the US, they’ve triggered protests over how the infrastructure underpinning AI might stress power grids and local water systems. That’s fueled partly by concerns about how difficult it is to see just how much electricity or water they’re actually gobbling up. California’s new disclosure rules won’t offer a full picture, but starting next year, data center operators will have to start sharing some of these details. The transparency will help scientists and the public see whether data centers are raising bills, draining dwindling resources, and making good on sustainability commitments.

“It’s extremely frustrating to try to understand basic information,” says Mark Specht, senior manager for the climate and energy program at the Union of Concerned Scientists (UCS). “There’s a lot of hype about data centers right now … It doesn’t help that we just don’t have basic facts about what’s happening with them to try to understand how much of a problem they are.”

“It’s extremely frustrating”

The seven bills Newsom signed go beyond requiring disclosures. Senate Bill 886, Assembly Bill 2383, and Senate Bill 1168 direct the California Public Utilities Commission to create separate power rates for data centers to recoup the costs of connecting such power-hungry facilities to the grid. That’s supposed to prevent data centers from passing on the costs to other consumers of new infrastructure built to satisfy their energy demand. On top of that, AB 2383 also pushes data centers to use more renewable energy.

The rest of the legislation grapples with the enormous gaps the public has when it comes to how data centers impact surrounding communities and the environment. AB 1577 requires monthly reporting on data centers’ energy consumption. AB 2619 and AB 2469 require water disclosures, and the latter also makes data center operators responsible for covering the costs of infrastructure upgrades needed to serve them. And SB 887 eliminates categorical exemptions from the California Environmental Quality Act for data centers, ensuring that they’re subject to environmental review.

“As scientists, when we went to research state offers in California, I was kind of dismayed by the lack of information that there was to work with. And so I’m happy [AB 1577] passed,” Specht says.

Specht co-authored a fact sheet this year on how data centers could impact the power grid and utility rates in California. It describes one scenario where several factors drive up costs for Californians: ballooning electricity demand; a need for expensive new infrastructure to connect data centers to high-voltage transmission systems; and the risk of overbuilding if the AI bubble pops and data centers don’t end up using as much electricity as projected, leaving other utility customers to foot the bill .

A different scenario shows data centers could actually lower electricity costs

On the other hand, a different scenario shows data centers could actually lower electricity costs. Electricity rates include fixed costs for grid maintenance, so spreading those costs across more customers can bring bills down for everyone. The catch is that the potential savings depend on data centers contributing to costs across the distribution system that serves smaller customers at lower voltages. Large data centers often avoid these costs by connecting instead to the transmission system, which carries high-voltage electricity.

Without more information, it’s impossible to say which scenario is taking shape in California. “I wouldn’t say definitively that data centers have increased the cost [of electricity],” Specht says. “It’s possible they may even have brought them down some already, I just don’t think we have the numbers to demonstrate it.”

Even if data centers aren’t having much effect on electricity rates yet, that’s likely to change as their power needs grow, Specht adds — making it all the more important to see how much energy they’re using. “It’s just about getting ahead of it before we all start feeling it,” he tells The Verge.

Data centers’ water use has been just as opaque. While studying the impact of data centers on water-stressed communities in California, researchers from Santa Clara University recently contacted every water provider in districts housing data centers. Citing privacy regulations, the providers in every district refused to share data on how much water data centers use. The researchers also discovered that very few data centers in the state had environmental impact reports available to the public.

The laws still leave big gaps in what data centers need to report

That dearth of information is even more concerning considering what the researchers were able to find. As AI drives the development of hyperscale data centers, these facilities are moving farther from more affluent urban areas into more rural areas where small water systems and lower-income communities may have more difficulty coping with the added demand.

The laws still leave big gaps in what data centers need to report. Specifically, AB 2619 and AB 2469 only require companies to make water disclosures when applying for permits or getting a business license, so we won’t know how much water they’re actually using from year to year. The year in which they share their water use could be particularly wet or cool, not necessarily reflecting how much water the facility needs when it’s hotter and drier. It also makes it difficult to see if a data center’s consumption is growing or if a company is following through on its sustainability goals.

Still, “This is in many ways a step in the right direction. It’s signaling to the industry that people are paying attention, that lawmakers are paying attention,” says Iris Stewart-Frey, the lead author of that water report and a professor of environmental science at Santa Clara University.

She notes that Newsom actually vetoed a similar water disclosure bill last year, but that data centers have become a bigger concern for the public since then.

“The whole AI wave is not a force of nature, it’s not a hurricane,” she says. “This is something that we can actually regulate.”

Amazon wants to help the Colorado River, but we still don’t know how much water the company uses

2026-09-22T10:27:44-04:00

Amazon plans to spend $20 million on water conservation projects along the Colorado River, a crucial but dwindling water supply for 40 million people in the Western US.

The initiative comes as Amazon and other tech companies face mounting scrutiny over how much water their data centers use. Amazon has a goal of becoming water positive — replenishing more water than it uses — by the end of the decade. It recently disclosed some information about its water use, but we still have an incomplete picture of the company’s impact as a whole.

Amazon announced the launch of a Colorado River Basin Collaborative today, with a mission to “create one of the biggest corporate water conservation efforts for a single U.S. watershed.” It expects to raise $100 million over two years — including the $20 million investment from its own coffers — for water conservation efforts.

The Colorado River brings water to 40 million people

Stretching 1,400 miles, the Colorado River brings water to 40 million people across seven states in the Western US and and two states in Mexico. The waterway has faced “unprecedented” drought over the last 26 years, according to the US Bureau of Reclamation. Decades of overuse on top of human-caused climate change have exacerbated the problem. To cope, the US Department of the Interior finalized a plan in August imposing steep water cuts on Arizona, Nevada, and California.

As the first stop for most goods the US imports from Asia, California is home to more Amazon warehouses than any other state. In Orange County, California, Amazon pledged $2.2 million over the next three years to help detect and patch up leaks. That effort is ultimately supposed to conserve 189 million gallons of water each year, according to Amazon.

As part of the newly announced collaborative, other companies can contract independently with conservation projects they choose. The United Nations-backed Water Resilience Coalition will vet projects. Amazon is also joining the Water Resilience Coalition, which includes Microsoft, Meta, and other brands across a range of sectors.

Tech companies are facing waves of resistance against new AI data centers. AI was estimated to use between 312.5 billion and 764.6 billion liters of water in 2025, according to one estimate, about as much water as people consume globally in water bottles. In Arizona, Colorado River water cuts are expected to raise utility rates, and some residents fear data centers could make things worse.

In June, Amazon disclosed that its data centers used 2.5 billion gallons of water globally in 2025, including those it owns, leases, or shares. Data centers Amazon owns and operates directly reduced their water consumption by 2 percent in 2025 compared to the year prior, according to the company. It attributes gains to a 52 percent improvement in water efficiency since 2021. The company says it uses 0.12 liters of water per kilowatt-hour of electricity, claiming that’s about seven times more efficient than the industry average for data centers.

Amazon’s annual sustainability report shares water use per unit of power (l/kwh) for its data centers since 2021, but not the company’s total water use over that time period. Sustainability experts and environmental advocates caution against assessing a company’s sustainability by efficiency measurements alone. The more efficient a technology becomes, the more resources it might ultimately consume as companies start using it more, according to an economic concept called Jevons paradox. That makes it important to track how much water a company is using overall, as well as how much it needs in each place it operates during demand peaks when supply is most limited.

For comparison, Microsoft’s latest sustainability report discloses its data centers’ total water use since 2020, while Google’s report shares water use for global operations since 2021 with a breakdown by location. All three tech giants have had to abandon data center projects after facing community opposition, Reuters reports. Unlike other tech companies, Amazon’s warehouses also have a massive physical footprint that isn’t reflected in the company’s data center disclosures.

A Guardian investigation in 2025 alleges that Amazon obscured how much water the company uses by deciding not to account for how much it takes to generate electricity, considered a “secondary” water use. Concerned about “reputational risk,” executives decided to only address “primary” water use at its facilities when calculating progress toward sustainability goals. The numbers Amazon recently shared about its data center water use also exclude water consumption stemming from electricity generation.

While data centers often use water for cooling systems, a majority of a data center’s water footprint stems from this “secondary use.” Power plants need water to cool equipment and turn turbines using steam, and in 2021, Amazon used 10.5 billion gallons (39.7 billion liters) of water in 2021, according to the Guardian report citing an internal leaked document. That’s about as much as 95,000 US households consume.

In an email to The Verge, Amazon spokesperson Margaret Callahan says the Guardian report “relied on cherry-picked data from an outdated document that misrepresents our water strategy — making the conclusions fundamentally wrong and misleading” and that a “document’s existence doesn’t guarantee its accuracy or finality.”

Callahan didn’t provide more data on how much water Amazon has used in total over the years. She pointed out that the company is 75 percent toward its goal of becoming water positive, returning 3 gallons for every 4 that it used last year.

In its announcement today, Kara Hurst, chief sustainability officer at Amazon, compares the new Colorado River initiative to the company’s previous climate targets.

“Through the Climate Pledge, we’ve seen what happens when companies tackle a problem together: progress accelerates. We’re bringing that same model to the Colorado River because water and climate are deeply connected,” Hurst says.

Amazon’s carbon footprint, however, grew 16 percent in 2025 — a trend happening across the tech industry as companies race to make advances in AI.

Humans, not rogue AI, are still the biggest cybersecurity risk to energy systems

2026-09-19T16:28:15-04:00

Before recent high-profile hacks raised the specter of AI possibly “killing all humans,” our energy systems were already disturbingly vulnerable to cyberattack — and the risk is growing.

“We were always prey. We were just kind of surviving at the appetite of our predators,” Joshua Corman, executive in residence for public safety and resilience at the Institute for Security and Technology (IST), told me last year. At the time, I was preoccupied with a Department of Homeland Security warning that Iranian actors and sympathizers could target the US with cyberattacks.

Last week, I called Corman up to chat about recent incidents of rogue AI agents orchestrating their own complex cyberattacks. Even AI executives are talking about whether the technology they’re building could grow so out of control that it triggers an apocalypse. If there is now a 10 percent chance of artificial intelligence one day killing all humans, as some AI developers warn, surely there’s a chance it could knock our lights out in the meantime?

“Any sociopath that wants to [attack] is now more powerful than they used to be.”

But when I spoke to Corman and other cybersecurity experts, they were still more worried about generative AI in the hands of bad actors than they were about rogue agents. As tech companies race to build ever more powerful AI models, utilities will similarly have to shore up their defenses — no matter who or what initiated the attack.

“It’s literally any sociopath that wants to [attack] is now more powerful than they used to be,” Corman tells me. “This has been a force multiplier and continues to grow.”

Much of our critical energy infrastructure — keeping the lights on in our homes, food cold in our refrigerators, and life-saving devices working in hospitals — was never designed to connect to the internet. The lifespan of a power plant is typically decades long. The average age of a nuclear reactor in the US is about 44 years. They weren’t constructed with today’s cybersecurity risks in mind, making them easy targets for hackers.

Eventually much of this infrastructure did connect to the internet. It’s been difficult to fix any resulting cybersecurity vulnerabilities ever since. Some of the companies that originally designed the equipment still in use in the power sector have gone out of business, leaving no one behind to develop a software patch for those orphaned devices. Even when there is a patch available, applying it in a timely manner is another challenge. Unlike IT software upgrades, operational technology (OT) systems that control physical machinery for critical infrastructure might only be designed to apply updates once each quarter or year. Smaller utilities might also lack the resources, staffing, and know-how to use the latest defensive measures.

“The true difference from AI is that it’s letting adversaries move more quickly — but it’s very challenging for those defending the infrastructure to match that pace,” says Sophie McDowall, a research associate at the Foundation for Defense of Democracies’ Center on Cyber and Technology Innovation.

Intent is a key factor when assessing the risks posed by generative AI. When an OpenAI model managed to break out of the company’s training parameters to attack AI lab Hugging Face, “Some of the sophistication and the capabilities and just what we saw in that were really eye-opening and in a sense terrifying in terms of how effective they were,” says Rob Denaburg, cybersecurity program senior manager at the American Public Power Association, which represents community-owned utilities across 2,000 municipalities.

But Denaburg points out that even in the Hugging Face hack and similar instances of AI agents breaking into systems they were never supposed to target, the rogue agents remained focused on fulfilling their training goals. If someone was to train a model to carry out an attack on energy infrastructure and agents broke out of the sandbox in that scenario, it would probably be a bigger concern for a utility. Again, that involves human adversaries with malicious intent.

Historically, adversarial nation-states were largely considered the biggest cybersecurity threat to critical infrastructure. “They’re going to be more disciplined,” Corman says, and more capable of undertaking a sophisticated cyberattack. Now, AI is making it easier for less-skilled adversaries to launch an effective assault.

“A bad-actor human can use these tools to be better than they naturally would be to attack things they normally didn’t know how to … because whereas they may not know OT protocols and OT networks and OT strategies, the LLM has read the manuals and does know what to do,” Corman says.

Utilities have to be more prepared, and defensive strategies are similar regardless of who the bad guy is. “AI or not, it is at the end of the day, still a cyberattack,” Denaburg says. “Even though AI can help an adversary maybe chain vulnerabilities together and automate some of the process going from initial access to exploit … as long as you can stop them in one spot, they can’t carry out that attack.”

Power companies can follow a range of best practices to safeguard critical infrastructure. Some of them are non-cyber solutions, like ensuring systems can switch to manual operations when needed or in some cases pulling back on how interconnected this infrastructure is in the first place. Increasingly, “in the face of the AI stuff, they’re starting to realize if we can’t protect it, disconnect it,” Corman says.

“They’re offering support for a problem that they are partially causing.”

Governments and the companies developing advanced AI models hold responsibility, too, McDowall says. It was a positive step, she notes, that OpenAI CEO Sam Altman recently met with utilities to discuss securing power grids. But there’s a lot more they could do to prevent disaster, she says. “They’re offering support for a problem that they are partially causing,” while failing to adequately control their own technology advancements, McDowall tells The Verge.

And while there are regulatory guardrails for research and development when it comes to nuclear technologies and hazardous materials, there aren’t yet the same policy safeguards for AI. “This is a technological scientific development that could cause potential risk to critical infrastructure systems, that can cause potential threat to human life. And so there does need to be restrictions,” McDowall says.“I recognize that we also don’t want to limit development, but there’s no reason that we can’t drive research forward while also doing it responsibly.”

She points out that there’s a dearth of research into how AI might be used to improve cybersecurity for energy systems — particularly beyond just red teaming to discover vulnerabilities.

Earlier this month, OpenAI pledged $1 billion toward subsidizing training and access to new models that are supposed to help defend critical infrastructure.

“In the coming months, AI-enabled cyber attacks will become far more widespread and sophisticated as models around the world become increasingly capable,” OpenAI said in the September 3 announcement. “Frontier AI can help defenders move faster.”

But Corman cautions against relying on friendly AI agents to fight off the malicious ones. They could both be hard to control in the sensitive confines of an OT system for critical infrastructure.

“It’s also really dangerous to introduce too much change too fast in an OT environment,” Corman says. “Now we have an AI bull fighting another AI bull in an OT china shop.”