The race for power generation is gaining momentum beyond China, with Silicon Valley tech giants leading the charge. At the forefront of this movement is Sam Altman, co-founder of OpenAI and a major investor in the nuclear fission startup, Oklo. Following its stock market debut on May 11 of last year, Oklo has seen its shares skyrocket nearly 900%, quickly establishing itself as one of the most significant players in the US stock market.

Alongside Altman, a slew of influential figures from the American tech and investment sectors are pouring resources into nuclear energy. This includes heavyweights such as Bill Gates, Jeff Bezos, Peter Thiel, Larry Ellison, and Cathie Wood. These industry leaders are well aware that the electrification of artificial intelligence is creating a massive demand for energy.

Reports indicate that each query processed by ChatGPT consumes a staggering amount of electricity—almost ten times that of a traditional Google search. This exponential increase in energy requirements has prompted tech giants to rethink their energy investments. NVIDIA's CEO, Jensen Huang, raised eyebrows with his comments, stating that “without improved computational speed, we may need 14 planets, three galaxies, and four suns to fuel everything.”

This worrying energy consumption discussion revolves primarily around supply-side electricity concerns. However, a complete electricity system encompasses more than just power generation; it incorporates intricate networks of transmission, distribution, generators, transformers, and other critical electrical infrastructure. Alongside renewable energies such as nuclear, wind, and solar power, which require uranium fuels, advanced nuclear hardware, and storage facilities, a comprehensive understanding of energy systems is vital.

Unfortunately, much of the electrical infrastructure in Europe and North America is outdated, with many grid lines needing urgent repairs. In the US, nearly all transformers are imported, as the nation grapples with an electricity investment deficit that will require unprecedented capital to rectify.

Driven by the artificial intelligence revolution, the world has entered a historic cycle of electricity demand. Yet, this surge places a unique constraint on the development of artificial general intelligence (AGI), where energy, rather than technological advancements, may prove to be the limiting factor.

Prominent figures within the tech sector have sounded alarms about the impending energy crisis faced by AI development. Altman has posited that the future of AI hinges on breakthroughs in energy supply, as its consumption will exceed previous expectations. Meanwhile, Elon Musk has made it clear that after the chip shortage comes an electricity shortage, predicting that by 2025, power will not meet the demands of all chip production.

The electricity demands for generative AI are immense. For instance, ChatGPT reportedly handles approximately 200 million requests daily, consuming over half a million kilowatt-hours of energy—effectively 17,000 times the electricity used by an average American household. According to Guosheng Securities, a single training session for GPT-3 costs around $1.4 million, with larger language models ranging from $2 million to $12 million. An alarming 60% of these costs are attributed to electricity consumption.

Data centers themselves represent a massive draw on electricity, with Amazon’s data center consuming as much energy as a medium-sized city each year. The U.S. Department of Energy reports that electricity use by data centers has surged from 58 TWh in 2014 to 176 TWh in 2023, now accounting for 4.4% of total U.S. energy consumption. Projections suggest this could rise to between 325 TWh and 580 TWh by 2028, potentially increasing its share to between 6.7% and 12% of the national total.

As general AI still finds itself in its infancy, the future will likely see an influx of supermodels akin to DeepSeek and ChatGPT. However, it remains challenging to predict the long-term electricity requirements that will follow the widespread adoption of AI technologies.

Currently, the global electricity supply is heavily reliant on traditional fossil fuels, incurring high costs and struggling to meet low-carbon transition demands. Renewable resources like solar and wind, while crucial, are not consistently reliable, making nuclear energy an increasingly attractive option for tech giants.

Particularly, small modular reactors (SMR) that require less time and investment, and offer enhanced safety and adaptability, are viewed as optimal investments by AI companies. In the U.S., SMR investments often come with government subsidies, making them more financially appealing than those for larger nuclear plants. According to the Department of Energy, SMR power costs about $180 per MWh, but with subsidies, this can drop to roughly $100 per MWh—less than wind and solar energy.

Since last October, tech giants in Silicon Valley have been aggressively acquiring nuclear power stakes:

• Microsoft partnered with Constellation Energy, investing $1.6 billion to revive the Three Mile Island Unit 1 reactor in hopes of supplying power by 2028.
• Google contracted with the startup Kairos Power to buy electricity from seven reactors, totaling 500 MWh.
• Amazon signed three agreements, investing $500 million with Dominion Energy to develop an SMR project, while also securing funding for four advanced SMRs and setting up a data center next to Talen Energy's nuclear facilities in Pennsylvania.
• Oracle’s Larry Ellison announced plans to create a nuclear-powered data center driven by three SMR reactors, producing over 1,000 megawatts of electricity.

Despite these ambitious plans, the output from the proposed small-scale nuclear energy initiatives remains minimal compared to the soaring overall electricity demands, underscoring the need for a massive global electricity investment.

In the face of this energy-driven era, many countries and large enterprises are outlining ambitious renewable energy plans, yet outdated grids and electrical gear hinder these transformative efforts.

The UK government committed in 2021 to achieving 100% clean electricity by 2035. Despite heavy investments pouring into wind and solar projects, the reality is that the generated electricity often cannot be fed into the grid. The antiquated grid system leaves renewable energy underutilized while awaiting connection—a process that can take up to 10 to 15 years.

According to the UK’s National Grid, by 2030, the high-voltage transmissions needed will exceed the total of the past three decades combined. A staggering amount of solar and wind capacity—596 GW—awaits integration into European grids, where the average wait time for new projects is estimated between 3 to 7 years.

Much of Europe’s high-voltage infrastructure dates back to the 1950s through the 1980s, with certain grid systems operating for over 70 years. Similarly, most U.S. electrical systems and transformers were installed between the 1950s and 1970s, with 70% functioning for over 25 years, and 60% of circuit breakers exceeding 30 years of operation. In essence, many countries’ electrical grids are nearing or have already surpassed their operational lifespans, contributing to the challenge of integrating new renewable resources.

The urgency for grid modernization cannot be overstated. The U.S. Department of Energy (DOE) has initiated the GRIP program, projecting an $10.5 billion investment in grid transformation over five years. The European Union is pledging €584 billion by 2030 for electrical upgrades, while the UK’s National Grid plans to spend £60 billion over the next five years to revamp its infrastructure—almost double the previous five-year allocation.

Such massive investments are already showing up in the performance and stock prices of electrical companies, with traditional power giants seeing growth rivaling that of AI companies.

General Electric (GE), despite being a shadow of its former self, has seen its spin-off, GE Vernova, thrive post-independence, with stock prices rising over 165% since its April launch. In 2024, GE Vernova reported a net profit of $1.552 billion, an impressive increase of 454.34% year-over-year.

Siemens Energy, born from the German powerhouse, has seen its stock price soar 331% over the past year, prompting comparisons to AI-stocks. It holds a record backlog of orders amounting to €123 billion, with its capabilities for the next two years already fully booked. This reflects the high demand at the downstream level.

Established companies are also adapting to industry changes. For example, Hitachi’s stock has jumped 67.96%, while Schneider Electric and ABB have increased by 42.66% and 40.69%, respectively. Comparatively, NVIDIA, a leader in AI chips, has experienced stock price growth of “only” 171% and profit growth of 190.64% in Q3 2024.

This surge mirrors how industrialization and urbanization have previously amplified demands on the electrical grid, as AI investment paves the way for a new epoch of energy expenditure. "We are in the initial phase of a super cycle in power investments," stated Scott, CEO of GE Vernova.

Elon Musk has also raised significant concerns over transformer shortages, asserting that, “the parallel growth of electric vehicles and AI will generate immense demand for power equipment.”

A fully functioning electrical system encompasses ample supply, an organized grid, and necessary equipment including transformers, meters, storage systems, and sophisticated control centers. Currently, transformers are particularly critical components in advancing the electrical systems across the U.S. It mirrors the challenges associated with the semiconductor industry, as the U.S. relies heavily on imports of crucial transformer materials, specifically "oriented silicon steel."

This material was first discovered in the 1930s in a laboratory by N.P. Goss; however, its lack of domestic production capacity has severely delayed grid upgrades in the U.S., risking the progress of AI and its journey into the future.

Silicon steel, an essential alloy for transformer cores, represents about a quarter of the total transformer cost. If the U.S. cannot secure domestic sources of this essential material, the country's dependence on imports will grow. Between 2023 and the present, the U.S. has imported transformers worth approximately $5.8 billion—an increase of 48.7% year-on-year.

Moreover, the increasing costs associated with oriented silicon steel and copper, compounded by the shipping disruptions caused by recent global events, serve to escalate transformer prices sharply, driving growth of 60%-70% between 2020 and 2023. Delivery cycles have lengthened dramatically, with waits stretching from 50 to 150 weeks—some even subsequently reaching five years.

China is a primary transformer-producing country, yet with rising trade restrictions, the U.S. has turned to imports from Mexico, South Korea, Brazil, and Canada. Most supplies derive from companies like Hitachi, GE Vernova, Siemens Energy, S&J Electric, and Hyundai Electric.

It's vital to note that exports from China contribute only about 8% of the U.S. transformer market; thus, claims that “China is blocking the United States' electrical equipment” lack factual basis. The electrical equipment market largely operates on competitive principles, differing fundamentally from the high-tech semiconductor sector, which is hindered by monopolistic practices and trade barriers.

However, transformers are just one small piece of this energy transition puzzle. There remains an urgent requirement for upgrades to smart meters, relays, capacitors, inverters, generators, and other elements, alongside the digitization of smart grids. The pressures experienced now arise from decades of underinvestment and deindustrialization.

The idea that "artificial intelligence is both embodied and material, composed of natural resources, energy, human input, infrastructure, logistics, history, and categorization,” serves to ground AI in a more tangible context. As Caitlin Crawford articulated, AI has become the “oil of the new century.” Nonetheless, the aging energy systems in Europe and North America haven’t prepared themselves for this seismic shift in energy demands.

In this evolving narrative of electricity investment, each region has distinct storylines—China, the U.S., and Europe all tread diverging paths. While China has consistently increased its investment in energy, pioneering advancements in ultra-high voltage technology and digital grids, countries in the West find themselves grappling with the need to modernize outdated infrastructures.

The anticipated rise in electrical demand driven by AI, household needs, and industrial demands presents a reality where traditional fossil fuel systems will increasingly fall short of meeting societal needs. Viewed through a broader historical lens, the AI boom may merely represent the beginning of a transformative era for global energy systems.