Sam Altman challenged critics of A.I.’s water and electricity consumption. Photo by John MacDougall/AFP via Getty Images
Sam Altman is pushing back on mounting criticism over the environmental toll of A.I. The OpenAI chief has dismissed claims about A.I.’s water consumption as “fake” and drawn comparisons between the electricity required to power A.I. systems and the energy it takes to develop human intelligence.
Figures suggesting that tools like ChatGPT consume multiple gallons of water per query are “totally insane” and have “no connection to reality,” Altman said in a Feb. 20 interview with The Indian Express on the sidelines of the AI Impact Summit in New Delhi. Last year, Altman claimed that ChatGPT uses 0.000085 gallons of water per query—roughly one-fifteenth of a teaspoon—though he did not explain how he calculated that figure.
A.I.’s water footprint largely stems from the need for evaporative cooling systems used to keep data center hardware from overheating. But Altman argued that companies like OpenAI are no longer directly managing such cooling processes. Many A.I. developers, he noted, are shifting toward cooling systems that recirculate liquid rather than continually drawing fresh supplies. Meanwhile, tech giants like Microsoft, Meta, Google and Amazon have pledged to replenish more water than they withdraw by 2030.
Even so, data centers continue to drink up water at a rapid pace. Total A.I.-related water consumption for cooling reached 23.7 cubic kilometers in 2025, a 38 percent increase over 2020, and is expected to more than triple over the next 25 years, according to a January report from Xylem. Despite the industry’s pivot to alternative methods, the report found that 56 percent of data center capacity still relies on some form of evaporative cooling.
Altman was more measured when it came to electricity usage. “What is fair, though, is the energy consumption,” he said. “We need to move towards nuclear, wind, and solar very quickly.”
Last April, the International Energy Agency reported that data centers accounted for roughly 1.5 percent of global electricity consumption in 2024. Their power use is rising at a rate more than four times faster than overall electricity demand and is expected to more than double by 2030.
In response, major tech companies are pursuing data center agreements tied to alternative energy sources, including nuclear power, to ease pressure on grids. Altman, who previously led Y Combinator, has personally invested in nuclear ventures such as Oklo, which is developing small-scale nuclear plants, and Helion, which aims to commercialize nuclear fusion.
The OpenAI CEO also argued that critics overlook the energy required to develop human intelligence. “People talk about how much energy it takes to train an A.I. model relative to how much it costs a human to do one inference query,” he said. “But it also takes a lot of energy to train a human—it takes, like, 20 years of life and all the food you eat during that time before you get started.”
A more appropriate comparison, he suggested, would measure the energy used by a fully trained A.I. model to answer a question against that used by a human doing the same task. “Probably A.I. has already caught up on an energy efficiency basis measured that way.”
The remarks quickly sparked debate online over whether such comparisons are appropriate. “He’s saying a really big spreadsheet and a baby are morally equivalent,” wrote Matt Stoller, research director of the American Economic Liberties Project, in a post on X. Sridhar Vembu, founder and chief scientist of software firm Zoho Corporation, also took issue with the OpenAI chief’s statements. A.I. should “quietly recede into the background” instead of dominating our lives, said the billionaire on X. “I do not want to see a world where we equate a piece of technology to a human being.”
Shares in the company, which is designing so-called small modular nuclear reactors (SMRs), have surged nearly 140% over the past month on Big Tech’s growing interest in nuclear power. SMRs are designed to produce cheaper, faster, greener energy than traditional nuclear facilities.
“A nuclear power renaissance is underway with nuclear increasingly viewed as a solution which solves both the increased need for baseload power and the need to decarbonize,” wrote Craig-Hallum analyst Eric Stine in a recent note to investors. Baseload power refers to the day-to-day energy demand on an electrical grid.
Stine said Google and Amazon’s investments are “truly just the beginning of a multi-decade megatrend.”
Goldman Sachs estimates that global data center power consumption will grow 160% by 2030, driven by demand from artificial intelligence. Meanwhile, separate data from the International Atomic Energy Agency shows nuclear power production in North America potentially doubling by 2050.
Stocks of other firms making similar tech to Oklo’s, such as NuScale (SMƒR) and NANO Nuclear Energy (NNE), also surged following news of Google’s and Amazon’s investments on Oct. 14 and Oct. 16, respectively, before paring gains this week.
“The opportunity is so massive here in the market that there’s going to be a good number of folks that are successful,” Oklo CEO Jacob DeWitte told Yahoo Finance.
In fact, the SMR market could grow to $300 billion by 2040, according to research cited by Citi analysts.
Sam Altman owned a 2.6% stake in the company, according to a regulatory filing in June. He became chair of Oklo in 2024 after serving as its CEO for three years.
Sam Altman, co-founder and CEO of OpenAI as well as chairman of Oklo, speaks during the Italian Tech Week 2024. (Stefano Guidi/Getty Images) ·Stefano Guidi via Getty Images
While Oklo was founded in 2013, well ahead of the AI boom, the energy needs of artificial intelligence have been a boon to the firm as it builds its client book, DeWitte said.
“AI does make [up] the bulk of our order book that we’ve talked about,” DeWitte said in an interview. That includes clients in the chip sector as well as data centers, he added. Publicly, the company has said it’s working with hyperscalers Equinix and Prometheus Hyperscale. It has yet to disclose any deals with Big Tech firms.
But Wall Street analysts caution that the company and its competitors face regulatory and supply chain hurdles before they can produce anywhere near the amount of power needed to run artificial intelligence data centers.
Nuclear projects have been subject to stringent regulations in response to high-profile global nuclear meltdowns at Three Mile Island in 1979, Chernobyl in 1986, and Fukushima in 2011. On average, it takes the US Nuclear Regulatory Commission 80 months to approve nuclear plant construction in the US, compared to an average of 54 months in the UK, according to research cited by Canaccord Genuity.
While a number of companies are developing SMRs — including Bill Gates’ TerraPower — none have been deployed in the US to date, and long licensing processes mean profitability isn’t exactly on the horizon.
In Oklo’s first earnings report since going public on Aug. 13, the company reported a net loss of about $53 million for the first six months of the year, wider than the roughly $9 million loss it posted in the year-earlier period.
Mockup of Oklo’s Aurora powerhouse. (Courtesy of Oklo, Illustration by Gensler) ·(Gensler)
Fuel is also an issue for companies in the space: Many SMRs, such as those of Oklo, NuScale, and TerraPower, require a specific type called high-assay low-enriched uranium, also known as HALEU, which is imported from Russia. There’s little to no domestic supply since the West at large has avoided developing a HALEU supply chain. That’s because enriched uranium is used in nuclear weapons, according to Canaccord Genuity.
“A developed Western HALEU supply chain is absent, particularly given efforts to limit enrichment due to concerns over proliferation,” wrote Canaccord Genuity analyst George Gianarikas in a note to investors early this year.
Citing supply constraints and long and difficult licensing processes, Citi’s Vikram Bagri recently reduced his price target on Oklo stock from $11 to $10.
“There are too many ‘ifs’ in this process,” Bagri said. “The new technology and new nuclear installs, it seems like, will happen only after 2030. After 2030, it remains to be seen who gets successful and how successful they get, how many nuclear reactors we see.”
But Seaport analyst Jeffrey Campbell sees Oklo having advantages in the burgeoning market, pointing to its ability to use “much cheaper” recycled fuel.
Whereas a traditional nuclear plant can take billions of dollars to build, Oklo has said the cost of bringing one of its SMRs online is a few hundred million dollars.
For his part, DeWitte is optimistic about the future of the market — and Oklo’s place in it. DeWitte cited mounting bipartisan support to reduce regulatory hurdles and enhance the domestic supply of HALEU. He said Oklo is unique because the company is looking to own and operate its facilities and sell energy directly to customers, rather than the traditional model of licensing its technology to utilities. DeWitte believes this will enable the company to bring its reactors online faster.
Oklo CEO Jacob DeWitte, joined by Lynn Martin, NYSE group president, rings the Opening Bell after Oklo IPO. (Courtesy of Oklo) ·NYSE
“In the world today, where this opportunity is so massive — diverse customers that need different things, want different things … you’re going to see a diverse ecosystem evolve,” he said. He added, “You’re not going to see one winner take all.”
Laura Bratton is a reporter for Yahoo Finance. Follow her on X @LauraBratton5.
The Microsoft founder is ramping up his investments in nuclear power. Halil Sagirkaya/Anadolu via Getty Images
Despite the decades-long efforts of scientists around the world, the commercialization of nuclear fusion technology has not yet been achieved on Earth. However, Bill Gates, who has invested significantly in both nuclear fission and fusion startups, is betting on cutting-edge tech to provide a promising path toward green energy. “I’m a big believer that nuclear energy can help us solve the climate problem,” the Microsoft (MSFT) co-founder told The Verge in a wide-ranging interview published today (Sept. 5).
Gates has long been outspoken about his adventurous approach to climate technology. Such sentiments have become more pertinent in recent years as concerns about Big Tech’s energy use proliferate. The energy consumption of data centers that power A.I. computing, for example, is expected to potentially double to take up 9 percent of the nation’s electricity by 2030, according to the Electric Power Research Institute.
According to Gates, A.I. data centers will actually generate a less than 10 percent increase in energy use. Even so, Big Tech is exploring clean energy sources and will pioneer fission and fusion power “to help bootstrap that green energy generation,” he said. Microsoft, for example, last year signed a power purchase agreement with Helion Energy, a nuclear fusion company backed by Sam Altman, to buy electricity from the startup in 2028.
Lauded for its potential to provide mass amounts of affordable and clean energy, nuclear fusion is the same process that powers the sun and stars. It occurs when two light atoms combine to form a heavier one while releasing energy, a reaction that must take place in extremely high temperatures of around 10 million degrees Celsius, according to the International Atomic Energy.
Although the process has yet to be commercially harnessed, nuclear fusion technology has received an outburst of financial support in recent years. Of $7.1 billion in total funding since 1992, the sector received $900 million last year, according to a recent report from the Fusion Industry Association, which noted that 89 percent of private fusion companies believe the technology will be operational by the end of the 2030s.
The report identified 45 companies worldwide working to commercialize nuclear fusion. Of those startups, five are backed by Gates via Breakthrough Energy Ventures, his climate-focused investment fund. The billionaire has invested in the likes of Zap Energy, which is hoping to build a fusion power plant in the next few years, and Type One Energy, which uses magnets to help fuse atoms. Both Gates and Amazon (AMZN)’s Jeff Bezos have supported Commonwealth Fusion Systems, another startup aiming to make the commercialization of fusion power a possibility in the near future.
Despite skepticism over whether nuclear fusion—which doesn’t emit greenhouse gases or carbon dioxide—will actually come to fruition in the next few years or decades, Gates said he remains optimistic. “Although their timeframes are further out, I think the role of fusion over time will be very, very critical,” he told The Verge.
The billionaire has also invested in modern forms of nuclear fission energy, which produces energy when atoms are split apart. Gates is attempting to develop a cheaper form of fission via $1 billion worth of investments into TerraPower, a startup that recently broke ground on a nuclear power plant site in Kemmerer, Wyo. and aims to develop more affordable and safer forms of fission by using water to cool reactors instead of sodium. “People are appropriately skeptical because it’s never been done,” Gates told The Verge. “But they’ll get to see as we build that plant, and if so, it can make a contribution.”
Gates isn’t alone in his embrace of all things nuclear. Bezos, too, has become a prominent investor in fusion technology, having invested in Canadian startup General Fusion’s dreams of developing a pilot plant. OpenAI’s Altman has poured capital and time into the field as well, backing and chairing both Helion Energy and the nuclear energy startup Oklo.
Sam Altman is now chairman of a public company. But it’s not OpenAI.
On Friday, advanced nuclear fission company Oklo started trading on the New York Stock Exchange. The company, which has yet to generate any revenue, went public through a special purpose acquisition company (SPAC) called AltC Acquisition Corp., founded and led by Altman.
Under the ticker symbol “OKLO,” shares plummeted 54% on Friday to $8.45, valuing the company at about $364 million. Oklo received roughly $306 million in gross proceeds in the transaction, according to a release.
Oklo’s business model is based on commercializing nuclear fission, the reaction that fuels all nuclear power plants. Instead of conventional reactors, the company aims to use mini nuclear reactors housed in A-frame structures. Its goal is to sell the energy to end users such as the U.S. Air Force and big tech companies.
Oklo is currently working to build its first small-scale reactor in Idaho, which could eventually power the types of data centers that OpenAI and other artificial intelligence companies need to run their AI models and services.
Altman is co-founder and CEO of OpenAI, which has been valued at over $80 billion by private investors. He’s said that he sees nuclear energy as one of the best ways to solve the problem of growing demand for AI, and the energy that powers the technology, without relying on fossil fuels. Microsoft co-founder Bill Gates and Amazon founder Jeff Bezos have also invested in nuclear plants in recent years.
“I don’t see a way for us to get there without nuclear,” Altman told CNBC in 2023. “I mean, maybe we could get there just with solar and storage. But from my vantage point, I feel like this is the most likely and the best way to get there.”
In an interview with CNBC Thursday, Oklo CEO Jacob DeWitte confirmed that the company has yet to generate revenue and has no nuclear plants deployed at the moment. He said the company is targeting 2027 for its first plant to come online.
Going the SPAC route is risky. So-called reverse mergers became popular in the low-interest rate days of 2020 and 2021 when tech valuations were soaring and investors were looking for growth over profit. But the SPAC market collapsed in 2022 alongside rising rates and hasn’t recovered.
AI-related companies, on the other hand, are the new darlings of Wall Street.
“SPACs haven’t exactly had the best performances in the past couple of years, so for us to have sort of the outcome that we’ve had here is obviously a function of the work we put in, but also what we’re building and also the fact that the market sees the opportunity sets here,” said DeWitte, who co-founded the company in 2013. “I think it’s very promising on multiple fronts for [the] nuclear, AI, data center push, as well as the energy transition piece.”
The company has seen its fair share of regulatory setbacks. In 2022, the U.S. Nuclear Regulatory Commission denied Oklo’s application for an Idaho reactor. The company has been working on a new application, which it isn’t aiming to submit to the NRC until early next year, DeWitte said, adding that it’s currently in the “pre-application engagement” stage with the commission.
Altman got involved with Oklo while president of the startup incubator Y Combinator. Oklo went into the program in 2014 after an earlier meeting between Altman and DeWitte. In 2015, Altman invested in the company and became chairman.
It’s not Altman’s only foray into nuclear energy or other infrastructure that could power large-scale AI growth.
In 2021, Altman led a $500 million funding round in clean energy firm Helion, which is working to develop and commercialize nuclear fusion. Helion said in a blog post at the time that the capital would go toward its electricity demonstration generator, Polaris, “which we expect to demonstrate net electricity from fusion in 2024.”
Altman didn’t respond to a request for comment.
In recent years, Altman has also poured money into chip endeavors and investments that could help power the AI tools OpenAI builds.
Just before his brief ouster as OpenAI CEO in November, he was reportedly seeking billions of dollars for a chip venture codenamed “Tigris” to eventually compete with Nvidia.
Altman in 2018 invested in AI chip startup Rain Neuromorphics, based near OpenAI’s San Francisco headquarters. The next year, OpenAI signed a letter of intent to spend $51 million on Rain’s chips. In December, the U.S. compelled a Saudi Aramco-backed venture capital firm to sell its shares in Rain.
DeWitte told CNBC that the data center represents “a pretty exciting opportunity.”
“What we’ve seen is there’s a lot of interest with AI, specifically,” he said. “AI compute needs are significant. It opens the door for a lot of different approaches in terms of how people think about designing and developing AI infrastructure.”
