If we can get one of these big AIs on line, can somebody ask it what the best energy path forward is? Inquiring minds want to know.
Above, recent interview with Matt Garman, the CEO of Amazon’s cloud-computing unit.
Big money from Big Tech going into the nuclear sector, and I wish them all the luck in the world. The record suggests they are going to need it.
We’ll know the answer in the 2030s some time, but worth bearing in mind that there is no serious scenario for decarbonization that does not involve a LOT of solar, wind, and battery storage – all things that we know how to do, quickly and economically, right now. Whether you are pro, anti, or agnostic nuclear, it should be clear that we need to get cracking.
Amazon.com AMZN 1.26%increase; green up pointing triangle, Google and Microsoft MSFT -0.08%decrease; red down pointing triangle each struck recent deals meant to bring more nuclear power online to satiate burgeoning energy demand tied to AI. They are betting billions of dollars that nuclear power can help curb surging carbon emissions tied to the data centers, which threaten their climate pledges.
Some of the projects depend on unproven next-generation technology, and each is slated to take years to complete, in part because of financial and technological challenges that have stymied the growth of the U.S. nuclear industry for decades.
In the meantime, utility companies and power providers need to burn fossil fuels to supply the power tech companies need. Some have proposed new natural-gas plants, which are faster and cheaper to build than nuclear reactors, while others have contemplated extending the lives of coal plants that had been slated for closure.
The tech giants years ago set ambitious goals to slash their carbon emissions by purchasing clean energy such as wind and solar power. The development of ChatGPT and other large language AI models, which consume enormous amounts of electricity, has upended those promises and set off a scramble among the companies to find more clean power. Large data centers can consume roughly the same amount of electricity as a midsize city.
M.V. Ramana PhD for Utility Dive:
This past November, the Utah Associated Municipal Power Systems, or UAMPS, terminated what was to be “the first NuScale Power small modular reactor plant to begin operation in the United States.” This was a death foretold; the red flags have been obvious for years now. Although there were problems specific to that project, the financial challenges and cost trends witnessed in this case will afflict any small modular nuclear reactor project. In a rational world, no utility or government would invest another dime on these theoretical reactor concepts.
As announced in 2015, the UAMPS project initially involved constructing 12 reactor modules capable of generating 600 MW, with the aim of starting operations “around 2023,” and at an “overnight cost” of $3 billion. In 2018, NuScale announced a design modification with each module now producing 60 MW of electricity, or 720 MW for the whole plant, claiming this would lower the cost “on a per kilowatt basis from an expected $5,000 to approximately $4,200.”
The estimated costs of the project rose to $4.2 billion in 2018, then $6.1 billion in 2020, and finally $9.3 billion in 2023, after it was scaled down to 462 MW in 2021. In the end, the costs were clearly too high for UAMPS members to bear.
The engineers and accountants on the project were evidently unrealistic about the likely costs, or perhaps NuScale preferred to gently break the bad news about how immense the bill would be. Or both. NuScale also managed to retain members by claiming an unreasonably low cost of power from the project once operational, a cost derived using an opaque economic methodology without clarity over its assumptions. There is a lesson here: one just cannot trust initial cost estimates for nuclear reactors and their electricity.
—
Below, GE Executive Jon Ball on the economics of Small Modular reactors. TLDR, by mid century, they hope to have it down to double (or triple) what solar and wind are today.