Artificial Intelligence Needs Always Available Electricity
A gloomy Bill Gates, writing in his book “How to Avoid a Climate Disaster: The Solutions We Have and the Breakthroughs We Need,” warned as recently as 2021 that humanity’s only salvation was “net zero.”
In a new GatesNotes essay called “Three Tough Truths About Climate Change,” the Microsoft founder now says that while we’re going to miss global temperature goals, climate change won’t “end humanity.” According to The Guardian (Oct. 28, 2025), Gates believes global efforts need a “pivot” toward “reducing human suffering” instead of obsessing over temperature charts.
Gates may not be rejecting climate science. However, he rejects climate catastrophizing and the speculative “worst-case” scenarios used to justify eliminating reliable energy sources before alternatives exist. His reassessment reflects a hard-headed conclusion: policies that deprioritize affordability, growth, and national competitiveness will fail. And in an era when artificial intelligence will define economic power, energy-constrained nations will fall behind.
Artificial intelligence runs continuously. Training and inference workloads do not pause for clouds, wind lulls, or politically determined decarbonization timetables. Today, U.S. data centers consume roughly 3.5%–4% of national electricity. By 2030, projections indicate they could exceed 8%–9%, representing nearly half of all incremental electric-load growth.
This is baseload demand — not discretionary consumption. It requires dispatchable power, delivered on demand, every hour of the year. The physics is straightforward:
- Dispatchable power, always available: natural gas, nuclear, hydro, and transitional coal.
- Intermittent power, sometimes available: wind and solar, conditioned by weather and daylight.
The policy disconnect is equally straightforward. Net-zero frameworks elevate intermittent supply while restricting dispatchable generation. The result is a structural conflict: America cannot lead the AI century if it devalues reliable electricity.
A megawatt of installed solar capacity is not a megawatt of dependable capacity. California, Germany, and the U.K. have demonstrated that heavy renewable penetration without firm backup increases costs and erodes reliability. Germany’s power output from renewables has fallen below 5% of demand during periods of low wind and sun. Their experience is a caution: intermittent energy requires parallel grid infrastructure to prevent outages.
Battery storage is frequently touted as the answer. Yet current grid-scale systems provide only limited support, not multi-day or seasonal coverage. Scaling to that level would require investments measured not in billions but trillions, with extraordinary land-use and mineral-supply implications.
Leading technology firms used to compete at Davos to announce their net-zero pledges. Now, confronted with electric grid constraints that could adversely impact their investment returns, their tone has changed markedly. Mark Zuckerberg identifies electricity supply as the primary bottleneck to AI growth. Larry Fink, long an ESG proponent, now concedes AI data centers cannot be powered by wind and solar alone.
These are not climate skeptics. However, they are facing operational and engineering realities. Several U.S. jurisdictions have already delayed or paused issuance of new data-center permits due to inadequate transmission capacity or insufficient baseload capacity.
Microsoft is investing $80 billion in fiscal 2025 to build and expand data centers for AI workloads. This investment covers infrastructure, software, and partnerships with AI companies like OpenAI. Microsoft has even agreed to a $16-billion plan to restart a nuclear reactor at the Three Mile Island nuclear power plant in Pennsylvania. Under the 20-year power purchase agreement with Constellation, Microsoft will purchase approximately 835 megawatts of energy to match the power its data centers consume in the PJM region.
Meanwhile, China, already responsible for approximately 25% of global data-center demand, continues to expand its coal- and nuclear-generating capacity rapidly. The competitive implications are self-evident.
Gates’ shift is not an ideological retreat. It reflects recognition that human welfare, not symbolic emissions accounting, is the appropriate metric for environmental progress. AI will generate historic gains in productivity, medical research, defense capability, and economic opportunity. But it can only if supported by abundant, affordable, dispatchable electricity.
Climate goals can coexist with innovation and prosperity. They cannot supersede them. A nation that restricts energy supply will limit its own future. A nation that expands reliable energy could lead the 21st-century economy.
Bill Gates’ climate policy “U-Turn” should be understood as a reflection of these conflicting policy trade-offs between promoting “net-zero” and AI.
Direct Line News is a Maryland-based free-access publication committed to the Republican reform tradition.
About The Author
