Small Modular Reactor, Credit Westinghouse.
In the last couple of months, Microsoft, Google, and Amazon, in that order, made announcements about using nuclear power for their energy needs. Describing nuclear energy using questionable adjectives like “reliable,” “safe,” “clean,” and “affordable,” all of which are belied by the technology’s seventy-year history, these tech behemoths were clearly interested in hyping up their environmental credentials and nuclear power, which is being kept alive mostly using public subsidies.
Both these business conglomerations—the nuclear industry and its friends and these ultra-wealthy corporations and their friends—have their own interests in such hype. In the aftermath of catastrophic accidents like Chernobyl and Fukushima, and in the face of its inability to demonstrate a safe solution to the radioactive wastes produced in all reactors, the nuclear industry has been using its political and economic clout to mount public relations campaigns to persuade the public that nuclear energy is an environmentally friendly source of power.
Tech giants like Microsoft, Amazon, and Google, too, have attempted to convince the public they genuinely cared for the environment and really wanted to do their bit to mitigate climate change. In 2020, for example, Amazon pledged to reach net zero by 2040. Google went one better when its CEO declared that “Google is aiming to run our business on carbon-free energy everywhere, at all times” by 2030. Not that they are on any actual trajectory to meeting these targets.
Why are they making such announcements?
Greenwashing environmental impacts
The reasons underlying these companies investing in such PR campaigns is not hard to discern. There is growing awareness of the tremendous environmental impacts of the insatiable appetite for data from these companies, as well as the threat they pose to already inadequate efforts to mitigate climate change.
Earlier this year, the Wall Street company Morgan Stanley estimated that data centers will “produce about 2.5 billion metric tons of carbon dioxide-equivalent emissions through the end of the decade”. Climate scientists have warned that unless global emissions decline sharply by 2030, we are unlikely to limit global temperature rise to 1.5 degrees Celsius, a widely shared target. Even without the additional carbon dioxide emitted into the air as a result of data centers and their energy demand, the gap between current emissions and what is required is yawning.
But it is not just the climate. As calculated by a group of academic researchers, the exorbitant amounts of water required in the United States “to operate data centers, both directly for liquid cooling and indirectly to produce electricity” contribute to water scarcity in many parts of the country. This is the case elsewhere, too, and communities in countries ranging from Ireland to Spain to Chile are fighting plans to site data centers.
Then, there are the indirect impacts on the climate. Greenpeace documented, for example, that “Microsoft, Google, and Amazon all have connections to some of the world’s dirtiest oil companies for the explicit purpose of getting more oil and gas out of the ground and onto the market faster and cheaper.” In other words, the business models adopted by these tech behemoths depend on fossil fuels being used for longer and in greater quantities.
In addition to the increasing awareness about the impacts of data centers, one more possible reason for cloud companies to become interested in nuclear power might be what happened to cryptocurrency companies. Earlier this decade, these companies, too, found themselves getting a lot of bad publicity due to their energy demands and resulting emissions. Even Elon Musk, not exactly known as an environmentalist, talked about the “great cost to the environment” from cryptocurrency.
The environmental impacts of cryptocurrency played some part in efforts to regulate these. In September 2022, the White House put out a fact sheet on the climate and energy implications of Crypto-assets, highlighting President Biden’s executive order that called on these companies to reduce harmful climate impacts and environmental pollution. China even went as far as to banning cryptocurrency, and its aspirations to reducing its carbon emissions was one factor in this decision.
Crypto bros, for their part, did what cloud companies are doing now: make announcements about using nuclear power. Amazon, Google, and Microsoft are now following that strategy to pretend to be good citizens. However, the nuclear industry has its reasons for welcoming these announcements and playing them up.
The state of nuclear power
Strange as it might seem to folks basing their perception of the health of the nuclear industry on mainstream media, that technology is actually in decline. The share of global electricity produced by nuclear reactors has decreased from 17.5% in 1996 to 9.15% in 2023, largely due to the high costs of and delays in building and operating nuclear reactors.
A good illustration is the Vogtle nuclear power plant in the state of Georgia. When the utility company building the reactor sought permission from the Nuclear Regulatory Commission in 2011, it projected a total cost of $14 billion, and “in-service dates of 2016 and 2017” for the two units. The plant became operational only this year, after the second unit came online in March 2024, at a total cost of at least $36.85 billion.
Given this record, it is not surprising that there are no orders for any more nuclear plants.
As it has been in the past, the nuclear industry’s answer to this predicament is to advance the argument that new nuclear reactor designs would address all these concerns. But that has, yet again, proved not to be the case. In November 2023, the flagship project of NuScale, the small modular reactor design promoted as the leading one of its kind, collapsed because of high costs.
Supporters of nuclear power are now using another time-tested tactic to promote the technology: projecting that energy demand will grow so much that no other source of power will be able to meet these needs. For example, UK energy secretary Ed Davey resorted to this gambit in 2013 when he said that the Hinkley Point C nuclear plant was essential to “keep the lights on” in the country.
Likewise, when South Carolina Electric & Gas Company made its case to the state’s Public Service Commission about the need to build two AP1000 reactors at its V.C. Summer site—this project was subsequently abandoned after over $9 billion was spent—it forecast in its “2006 Integrated Resource Plan” that the company’s energy sales would increase by 22 percent between 2006 and 2016, and by nearly 30 percent by 2019.
This is the argument that the growth in data centres, propped up in part by the hype about generative artificial intelligence, has allowed proponents of nuclear energy to put forward. It remains to be seen whether this hype about generative AI actually materializes into a long-term sustainable business: see, for example, Ed Zitron’s meticulously documented argument for why OpenAI and Microsoft are simply burning billions of dollars and why their business model might “simply not be viable”.
In the case of the V.C. Summer project, South Carolina Electric & Gas found that its energy sales actually declined by 3 percent compared to 2006 by the time 2016 rolled around. Of course, that did not matter, because shareholders had already received over $2.5 billion in dividends and company executives had received millions of dollars in compensation, according to Nuclear Intelligence Weekly, a trade publication.
One wonders which executives and shareholders are going to receive a bounty from this round of nuclear hype.
What about emissions?
Will the investments in nuclear power by companies like Google, Microsoft, and Amazon help reduce emissions anytime soon?
The project expected to have the shortest timeline is the restart of the Three Mile Island Unit 1 reactor, which Constellation Energy projects will be ready in 2028. But if the history of reactor commissioning is anything to go by, that deadline will come and go without any power flowing from it.
Restarting a nuclear plant that has been shutdown has never been done before. In the case of the Diablo Canyon nuclear plant in California, which hasn’t been shut down but was slated for decommissioning in 2024-25 till Governor Gavin Newsom did a volte-face, the Chair of the Diablo Canyon Independent Safety Committee explained why doing so was very difficult: “so many different programs and projects and so on have been put in place over the last half a dozen years predicated on that closure in 2024-25 and each one of those would have to be evaluated and some of them are okay, and some of them won’t be and some are going to be a real stretch and some are going to cost money and some of them aren’t going to be able to be done maybe”.
The cost of keeping Diablo Canyon open has been estimated by the plant’s owner at $8.3 billion and by independent environmental groups at nearly $12 billion. There are no reliable cost estimates for reopening Three Mile Island, but Constellation Energy, the plant’s owner, is already seeking a taxpayer-subsidized loan that would likely save the company $122 million in borrowing costs.
One must also remember that Microsoft already announced an agreement with Helion Energy, a company backed by billionaire Peter Thiele, to get nuclear fusion power by 2028. The chances of that happening are slim at best. In 2021, Helion announced that it had raised $500 million to build its fusion generation facility that would demonstrate “net electricity production” in three years, i.e., “in 2024”. That hasn’t happened so far. But going back further, one can see a similar and unfulfilled claim from 2014: then, the company’s chief executive had told the Wall Street Journal that the company hoped that its product would generate more energy than it would use “in the next three years” (i.e., in 2017). It is quite likely that Microsoft’s decision-makers knew of how unlikely it is that Helion will be able to supply nuclear fusion power by 2028. The publicity value is the most likely reason for announcing an agreement with Helion.
What about the small modular nuclear reactor designs—X-energy and Kairos—that Amazon and Google are betting on? Don’t hold your breath.
X-energy is an example of a high-temperature gas-cooled reactor design that dates back to the 1940s. There have been four reactors based on similar concepts that were operated commercially, two in Germany and two in the United States, respectively, and test reactors in the United Kingdom, Japan, and China. Each of these reactors proved problematic, suffering a variety of failures and unplanned shutdowns. The latest reactor with a similar design was built in China. Its performance leaves much to be desired: within about a year of being connected to the grid, its power output was reduced by 25 percent of the design power capacity, and even at this lowered capacity, it operated in 2023 with a load factor of just 8.5 percent.
Kairos, on the other hand, will be challenged by its choice of molten salts as coolant. These are chemically corrosive, and decades of search have identified no materials that can survive for long periods in such an environment without losing their integrity. The one empirical example of a reactor that used molten salts dates back to the 1960s, and this experience proved very problematic, both when the reactor operated and in the half-century thereafter, because managing the radioactive wastes produced before 1970 continued to be challenging.
Simply throwing money will not overcome these problems that have to do with fundamental physics and chemistry.
Just a dangerous distraction
Although Amazon, Google, and Microsoft claim to be investing in nuclear energy to meet the needs of AI, the evidence suggests that their real motive is to greenwash themselves.
Their investments are small and completely inadequate with relation to how much is needed to build a reactor. But their investments are also very small compared to the bloated revenues of these corporations. So, from the viewpoint of top executives, investing in nuclear power must seem a cheap way to reduce bad publicity about their environmental footprints. Unfortunately, “cheap” for them does not translate to cheap for the rest of us, not to mention the burden to future generations of human beings from worsening climate change and, possibly, increased production of radioactive waste that will stay hazardous for hundreds of thousands of years.
Because nuclear power has been portrayed as clean and a solution to climate change, announcements about it serve as a flashy distraction to focus public attention on. Meanwhile, these companies continue to expand their use of water and draw on coal and especially natural gas plants for their electricity. This is the magician’s strategy: misdirecting the audience’s attention while the real trick happens elsewhere. Their talk about investing in nuclear power also distracts from the conversations we should be having about whether these data centers and generative AI are socially desirable in the first place.
There are many reasons to oppose and organize against the wealth and power exercised by these massive corporations, such as their appropriation of user data to engage in what has been described as surveillance capitalism, their contracts with the Pentagon, and their support for Israel’s genocide and apartheid. Their investment into nuclear technology, and more importantly, hyping it up, offers one more reason. It is also a chance to establish coalitions between groups involved in very different fights.