AI’s Power Needs Will Destroy the Renewable Energy Revolution

In 2015, I introduced a concept I called the “solar singularity”: the inflection point where solar energy becomes cheaper than fossil fuels and therefore the default choice for new electricity generation.

Every time the global capacity of installed solar panels doubled, prices fell by about 20 percent (this is called Swanson’s Law). I argued that we were halfway to nearly 100% of the world’s energy coming from solar and other renewables, even though we were then technically only at 1%. By 2020, I suggested, renewable energy would dominate the world’s energy infrastructure, not for environmental reasons or political mandates, but for economic reasons. Solar power would take over.

And Bill McKibben’s recent book Here is the sun confirms that the solar singularity occurred largely as I predicted. Solar power has become cheaper than fossil fuels in most countries, and by 2024, renewable energy would account for 92.5% of all new electricity generation worldwide. McKibben writes: “Solar power is growing faster around the world, not only than anything else right now, but than anything else ever has been. »

On supporting science journalism

If you enjoy this article, please consider supporting our award-winning journalism by subscription. By purchasing a subscription, you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.

California regularly generates more than 100% of its electricity from renewable energy during peak hours of the day. China has become a electrostate, manufacturing solar panels and batteries on an unprecedented scale.

But McKibben’s book overlooks a looming threat amid the celebration of renewable energy and carbon-based energy reduction: electricity demand from artificial intelligence data centers. The collision between the conflicting exponential curves of renewable energy supply and AI energy demand will determine not only the trajectory of the global energy system, but also how successful the transition to renewable energy can be before another technological consumption frenzy derails it completely.

We need to slow or stop the rush to build new AI data centers everywhere, regardless of the inconvenience or cost. We have maybe three to five years to make serious changes. The question is whether AI’s exponential energy appetite will increase faster than we can deploy renewable energy, forcing us back to fossil fuels or creating an energy shortage that will limit both technologies.

AI is already a huge energy drain. U.S. data centers consumed 183 terawatt hours (TWh) of electricity in 2024, equivalent to Pakistan’s entire electricity demand. With demand growing by 16% in 2025 alone, consumption is expected to reach 325 to 580 TWh by 2028, with AI driving most of this growth. AI-optimized servers are growing 30% annually, and by 2028, more than half of all data center electricity will be for AI, enough to power 22% of U.S. homes.

But there is more to come. In January 2025, OpenAI, SoftBank, Oracle, and MGX announced Project Stargate, a four-year, $500 billion AI infrastructure initiative (although recent funding disputes between the partners may delay the timeline). Additionally, Meta, Amazon, Alphabet, and Microsoft have collectively committed $320 billion to AI infrastructure in 2025 alone, representing a 39% increase from 2024. These are not incremental expansions; they represent the largest private investments in infrastructure in modern history.

The upcoming power demand for AI is staggering: Microsoft and Constellation Energy announced a 20-year deal to restart Three Mile Island Unit 1 exclusively to power AI’s data centers — the first commercial operation at the site since its 1979 accident. Amazon is building a 30-data center facility in Indiana consuming 2.2 gigawatts. The Meta campus in Louisiana will increase the utility’s energy demand by 30 percent. The Stargate facilities alone have a capacity of 10 gigawatts, which is equivalent to powering nearly 10 million homes.

Ireland serves as the striking canary in the coal mine. Data centers now consume 21% of the country’s electricity, and projections are expected to reach 30% by 2030, meaning nearly a third of an entire country’s energy budget dedicated to server farms. In the Mid-Atlantic and Midwestern United States, where PJM Interconnection manages the network, data center demand accounts for 30 of the projected 32 gigawatts of growth in peak load through 2030. Rates are increasing accordingly.

According to REN21 Global Situation Report 2025Data centers and cryptocurrency mining have led to a 20% increase in electricity demand (111 TWh) in 2024 alone, contributing 0.4% to total global electricity growth. Goldman Sachs predicts that AI will drive a 160% to 165% increase in data center energy demand by 2030 compared to 2022 levels.

Several jurisdictions (Amsterdam, China, Germany, Ireland, Singapore and parts of the United States) have already introduced restrictions on new data center connections due to network capacity constraints.

When an AI called DeepSeek was able to do the same job as GPT-4 with 11 times less computational effort, it seemed at first that it would lead to more energy-efficient AI. Instead, it incentivized AI developers to simply create larger, more complex systems.

Projections from the International Energy Agency (IEA) illustrate the gap between institutional forecasts. The IEA predicts that data centers will account for around 3% of global electricity consumption by 2030, a “relatively modest” increase. But these projections assume gradual and distributed expansion. They do not take into account the scaled functions represented by the megaprojects currently under construction.

Stargate alone – 10 gigawatts by 2025 – represents the electricity consumption of 7.5 million homes. It is a consortium project. When you add Meta’s $10 billion Louisiana facility, Microsoft’s $80 billion spending, and Amazon’s $100 billion commitment, you get a concentrated energy demand that dwarfs assumptions of distributed growth. If even half of the announced megaprojects come to fruition, I calculated that data center consumption in the United States could reach 15 to 20 percent of electricity demand by 2030. Globally, concentrated AI infrastructure could push data center consumption above 5 to 6 percent.

But the most immediate limitation is not capital or chips, but power infrastructure. Grid operators are managing demand that is growing faster than transmission capacity can be built. Delivery times for electrical equipment extend over two years. While global percentages may seem manageable, the concentration creates acute stress: Ireland hits 21 percent and growing, utilities build power plants for single data center projects, network operators implement emergency curtailment procedures.

Despite evidence that exponential development of AI is materializing, policy responses in the United States remain inadequate. There are no mandatory efficiency standards for AI training, no hard caps on calculated energy consumption, and no coordination between AI development and network planning. Investment in fundamentally different IT approaches is minimal.

Instead, we’re seeing a reactive rush: utilities building gas plants to meet data center demand, grid operators implementing emergency procedures, regulators approving massive rate increases for infrastructure upgrades, and ratepayers panicking over rising rates.

A little over 10 years after predicting the solar singularity in my book Solar: why our energy future is so promising, it arrived on time. But it is interrupted before it fully blossoms. The question is not whether solar storage, along with batteries, will become dominant – the economics are now irrefutable.

We face three potential pathways: a fundamental shift in AI development, moving away from scaling computation; conflicts over energy resources as AI competes with human needs; or immediate political intervention to prevent the crisis. Exponential mathematics suggests that we don’t have long before the collision of these curves forces radical changes.

What McKibben celebrated in 2025 – and what I predicted a decade ago – may not survive the coming decade unless we confront the AI ​​energy tsunami with the same urgency that we ultimately brought to the climate crisis.

This is an opinion and analysis article, and the opinions expressed by the author(s) are not necessarily those of Scientific American.