AI's Appetite and a Nuclear Revival

Hello, CipherTalk Readers!

Artificial intelligence is reshaping our lives in countless ways. But behind the chatbots and breakthroughs is a staggering fact: AI is driving a surge in electricity demand. It’s threatening how we generate power and the systems we rely on to deliver it.

One unexpected answer? Nuclear energy. Once dismissed as too complicated or risky, it’s now re-emerging as a key solution to meet the growing need for reliable electricity. But the stakes are high. The nations with abundant energy (and state-of-the-art infrastructure) will lead in the AI era and shape the global order.

In this issue, we’ll break down how energy is consumed today, why AI is such an energy guzzler, and what’s fueling the comeback of nuclear power.

AI is supercharging energy demand

Training a large AI model can use as much energy as 130 U.S. homes for a year. Once these models are live, they need constant power to answer questions or drive chatbots… every second of every day. Today, this amounts to the energy of a small country. It’s expected to match that of the entire country of Japan by 2026.

In the developed world, most electricity—from bedroom lights to the data centers behind websites—comes from the electric grid. The grid is essentially a massive web of power plants and transmission lines, supplying electricity to billions of people worldwide. It’s the backbone of modern life for billions of people.

AI’s rapid growth is pushing the electric grid to its limits. These systems, many designed decades ago, weren’t built to handle the massive energy demands of training and running advanced AI models. We need to ensure the grid can handle this demand reliably and efficiently, as usage continues to soar.

The grid itself is powered by a mix of sources:

• Fossil fuels, like coal and natural gas, still dominate in many places.

• Renewables, like solar and wind, are expanding quickly. But they depend on weather conditions and aren’t always reliable.

• Nuclear energy provides steady, carbon-free power.

  It hasn’t been widely adopted in recent decades due to high costs and safety concerns. But that’s starting to change. Advances in nuclear technology promise safer, more reliable, scalable solutions. Big tech needs energy that can meet the growing demand of AI systems—and it’s betting on nuclear.

Why is AI so energy-intensive?

A helpful stat: If generative AI were applied to every Google search, it would add the equivalent of 10 terawatt-hours to the world’s energy demand each year. That’s roughly the same amount consumed by 1.5 million people.

• Training involves running complex algorithms on thousands of powerful chips for weeks or months, consuming massive amounts of electricity.

• Usage, like answering your ChatGPT questions, requires constant data processing and storage, which keeps data centers running around the clock.

By 2026, energy use from AI and data centers is expected to double, further straining global power grids.

The 30,000-foot view

While compute is becoming an increasingly critical resource, the landscape is deeply uneven. Nearly 60% of the world’s servers are concentrated in Europe and North America—regions with just 17% of the global population. Meanwhile, sub-Saharan Africa faces daily power outages that cripple digital development, and countries like the U.K. are bogged down by planning delays. This imbalance highlights the stark reality of a new digital divide. Access to energy and infrastructure defines who leads, and who lags, in the intelligence era.

The pace of innovation leaves little room for error. Compute ecosystems are now a frontier for economic growth, but staying ahead is getting difficult. AI hardware performance doubles every 2.3 years as investors pour money into high-performance supercomputers.

But these advancements put pressure on the energy systems that back them. AI already consumes as much energy as a small country. The U.S. and China dominate this race, leveraging compute capacity as a benchmark of global power. Emerging players like Malaysia, Israel, and Rwanda are attempting to catch up through strategic investments in clean energy, talent pipelines, and regional partnerships.

The stakes couldn’t be higher. Nations with abundant energy and state-of-the-art infrastructure are racing ahead, treating compute as the foundation of both economic growth and geopolitical clout. The countries and companies that master this transition will lead not only in AI innovation, but in shaping the global order—where intelligence itself becomes the ultimate currency.

Why nuclear power is back

Nuclear power works by splitting atoms in a process called fission, which releases heat to produce electricity. (This is different from fusion, which combines atoms but is still in the experimental stage.) Nuclear fission has been around for decades, but safety concerns, high construction costs, and public skepticism have kept it on the sidelines in recent years.

So, why the renewed interest? There are three main reasons:

1. Reliability: Unlike solar or wind, nuclear plants make electricity around the clock.

2. Low Emissions: Nuclear energy is carbon-free, making it an essential tool in the fight against climate change.

3. New Technology: Small Modular Reactors (SMRs) are changing the game. These reactors are smaller, faster to build, and safer than traditional nuclear plants, making them ideal for powering data centers or small communities.

Sam Altman, CEO of OpenAI, believes nuclear energy is critical to AI’s future, calling “abundant energy” the foundation for progress. (Altman is chairman of recently-public Oklo, a nuclear fission company, and key investor in nuclear fusion co Helion Energy.)

But nuclear isn’t an overnight fix. A traditional plant can take decades to build, and even SMRs, with their streamlined design, may take about nine years to become operational due to regulatory and technical hurdles. Still, these timelines feel increasingly worth it as demand for reliable energy grows.

Tech giants are betting on nuclear

The biggest names in tech are taking nuclear power seriously. Here’s how they’re stepping into the space:

• Amazon: Amazon’s cloud computing arm AWS uses energy equivalent to a small country, much of which comes from renewables. In October, Amazon announced a new partnership with X-energy (Series C) to develop SMRs near Pennsylvania’s Susquehanna nuclear plant. Amazon also works with Dominion Energy and Talen Energy to supply its nuclear power plants.

• Google: Google offsets much of its grid usage with renewable energy credits. Now, it’s partnering with Kairos Power (which got approved yesterday to build two new reactors) to develop numerous SMRs. These could generate 500 megawatts of clean electricity by 2035.

• Microsoft: Azure, Microsoft’s cloud platform, is a major energy user. To meet this demand sustainably, Microsoft signed a 20-year deal with Constellation Energy (Public) to source power from the Three Mile Island nuclear plant. It’s also collaborating with Helion Energy (Series E) to explore fusion power.

What’s next?

The AI revolution isn’t just about smarter tools—it’s also about rethinking how we power them. Nuclear energy is stepping up as a key part of the solution. While it won’t happen overnight, the investments being made today could shape the energy of tomorrow.

As always, we’ll keep you updated on how technology is transforming the world.

Until next time,
/m