Power Users: Silicon Valley's Atomic?Ambition

In this edition of The Elite Edge, we bring you an interesting read on how Silicon Valley is turning to nuclear power to meet its AI needs.? Source: the New York Times “Hungry for Energy, Amazon, Google and Microsoft turn to nuclear power.”?

Fission chips: Silicon Valley’s New Power Play?—?How artificial intelligence is forcing tech giants to think nuclear

As artificial intelligence devours electricity, tech giants are betting their future on nuclear power

In the rolling hills of Pennsylvania sits Three Mile Island, site of America’s worst nuclear accident. Now, in an ironic twist, Microsoft plans to breathe new life into this dormant symbol of nuclear fear. It is not alone. Google has inked a deal with Kairos Power for small modular reactors, while Amazon is pouring $500 million into a partnership with Dominion Energy. Silicon Valley, it seems, has gone nuclear.

The driving force behind this atomic renaissance is artificial intelligence’s insatiable appetite for power. A single rack of AI computing chips demands over 100 kilowatts of electricity?—?more than ten times that of conventional servers. The numbers are staggering, global data center consumption is projected to more than double to 1,000 terawatt-hours by 2026, equivalent to Japan’s entire electricity usage.?

Even ChatGPT’s water consumption raises eyebrows, using half a liter for every few dozen queries. Researchers at the University of California, Riverside, published a study in April last year that found ChatGPT consumes 500 milliliters of water for every 10 to 50 prompts. That equates to roughly the amount of water in a standard 16-ounce bottle.

An order of magnitude solution

This surge in demand has forced tech giants to reconsider their energy strategy. While they previously favored wind and solar power, these renewable sources’ intermittency poses problems for data centers that must run continuously. Nuclear power offers what Raul Martynek of DataBank calls “an order of magnitude more intensive” solution. The tech industry’s capital expenditure reflects this urgency?—?the five largest companies spent $59 billion in the last quarter alone, up 63% year-on-year.

Yet skeptics abound. Arnie Gundersen of Fairewinds Energy Education points to a sobering statistic: of 250 American reactor projects attempted since 1960, more than half were cancelled before generating any power, and none met budget or schedule. Greenpeace maintains that nuclear power remains “incredibly expensive, hazardous and slow to build.”

Small is beautiful

The industry’s response is to think smaller. Rather than traditional behemoth reactors, companies are betting on small modular reactors (SMRs). “The key with nuclear is you have to pick something and build a lot of it to make it cheap,” argues Rich Powell of the Clean Energy Buyers Association. This assembly-line approach could slash costs and construction times, though the technology remains unproven. Both Amazon and Google announced that they were focusing on a new generation of small modular reactors.

Meanwhile, data centers increasingly strain local power grids, particularly in hubs like Northern Virginia. The tech industry’s nuclear pivot arrives alongside broader electrification trends, from electric vehicles to heat pumps. As Rosanne Kincaid-Smith of Northern Data Group observes, “I just don’t see traditional grids being the sustainable power that’s ongoing in the development of AI.”

Chain reactions?—?when big tech goes nuclear

The convergence of Big Tech and nuclear power represents a remarkable shift in American industrial policy. With bipartisan support and the Biden administration’s blessing, the nuclear sector may finally achieve its long-promised renaissance?—?powered not by utilities but by Silicon Valley’s deep pockets and urgent need for reliable, carbon-free electricity.

Atomic ambition

For an industry built on disrupting the status quo, Big Tech’s nuclear gambit may be its most ambitious yet. The question remains: can Silicon Valley’s optimism and capital overcome nuclear power’s troubled history? The answer may determine not just the future of AI, but of America’s energy landscape itself.