⚡ America needs an Energy and Science Act
The major long-run constraint on the AI Revolution will be power, not chips
The core case for the first part of the subsidy-laden 2022 CHIPS and Science Act: addressing critical vulnerabilities in the supply chain for these indispensable technological marvels. By reducing reliance on Taiwan — which makes nearly all of the most advanced processors — and bringing advanced chip production here to American soil, the law is meant to enhance national security and economic resilience through industrial policy.
Of course, cutting checks and making forecasts is the easy part here. In other words, the coast is far from clear, especially if America decides that attracting workers to the US is no longer a big priority. As I recently wrote:
The construction of the TSMC plant in Arizona has brought to light a significant culture clash between the work practices and management styles of TSMC’s home culture in Taiwan and the expectations of the American workforce. The Washington Post reports that “the companies have struggled to hire enough construction workers, especially welders and pipe fitters.” There’s also no guarantee that government checks will offset a series of poor past strategic decisions by Intel — including reluctance to invest in the most advanced manufacturing technologies — and enable it to again be a global leader in making cutting-edge chips. At least TSMC has shown it can produce cutting-edge chips, albeit in some other place. My point: It’s still a long way from breaking dirt to churning out chips, even for companies already successful at doing so.
It’s important for policymakers to think hard about economic vulnerabilities, resilience, supply chains, and bottlenecks. On the other hand, government intervention in markets means taxpayers might end up funding excess capacity if too much supply swamps too little demand from customers. But reducing our vulnerability is worth that risk, according to my AEI colleague Chris Miller, author of Chip War. All of which got me to thinking: Is there any bigger vulnerability right now than generating an adequate and growing supply of clean energy to power the emerging AI Revolution?
The Conservative Futurist: How To Create the Sci-Fi World We Were Promised
“If America is to fully recover from the COVID-19 pandemic, take full advantage of emerging tech from generative AI to CRISPR to reusable rockets, and launch itself into a shining tomorrow, it must again become a fully risk-taking, future-oriented society. It’s time for America to embrace the future confidently, act boldly, and take that giant leap forward.” - James Pethokoukis
I’m hardly the only one thinking about this. Our entire tech sector is concerned! Including Bill Gates!
It’s an issue explored at length in a recent interview that Goldman Sachs conducted with Brian Janous. He’s the co-founder of Cloverleaf Infrastructure, which helps manage and optimize the electricity needed for data centers and AI applications, and formerly vice president of energy at Microsoft. Janous explains that AI-related demand is causing a surge in power consumption, particularly for data centers, which is expected to continue growing significantly in the coming years.
When generative AI first exploded onto the scene, people debated what would constrain its potential — a shortage of chips or a shortage of power. That debate has now been settled, with everyone agreeing that over the medium-to-longer term the major constraint will be power. … AEP, one of the largest US electric utility companies, has reportedly received 80-90 gigawatts (GW) of load requests. Only 15 GW of that is likely real because many of the AI projects that companies are currently envisioning will never actually see the light of day. But 15 GW is still massive given that AEP currently owns/operates around 23 GW of generating capacity in the US. And even if overall grid capacity grows by only 2% annually—which seems like a reasonable forecast—utilities would still need to add well in excess of 100 GW of peak capacity to a system that currently handles around 800 GW at peak. The increase in power demand will also likely be hyperlocalized, with Northern Virginia, for example, potentially requiring a doubling of grid capacity over the next decade given the concentration of data centers in the area. So, grid capacity will need to expand substantially across the US, and likely even more in certain regions.
There are two compounding issues here that exacerbate the strain on our creaky old power grid, Janous goes on to explain. First, there’s a fundamental mismatch between the pace of AI development — which can see major improvements in months — while power infrastructure takes years to build.
Second, there are other factors increasing electricity demand, including transportation and building electrification, manufacturing onshoring driven by the Inflation Reduction Act and CHIPS Act, and potential hydrogen economy development. And right now, sorry to say, America simply isn’t able to expand grid capability and we have to settle for squeezing more from what we already have. Janous:
Expanding the grid is no easy or quick task. The electric utility industry is highly regulated, and utility companies must go through a lengthy permitting and approval process before starting to construct new capacity. The total capacity of power projects waiting to connect to the grid grew nearly 30% last year, with wait times currently ranging from 40-70 months, and lead times for critical electrical components such as transformers and switchgears have substantially increased. Until those issues can be resolved, and the grid can catch up, a significant power crunch will likely force utilities and states to pick and choose who receives power. The US has unfortunately lost the ability to build large infrastructure projects — this is a task better suited for 1930s America, not 2030s America. So, that leaves me a bit pessimistic. That said, utilities and policymakers are starting to take seriously the need to invest in America’s transmission infrastructure, which isn’t designed for today’s energy generation mix. The transmission infrastructure was built from the coasts into the country, but today, massive wind resources are located in the center of the country and solar in the southwest. So, the transmission system ideally needs to run from the inside out.
So maybe what America needs is an Energy and Science Act. The second part would involve funding federal R&D at levels far above those in the CHIPS and Science Act — and then actually spending it.
The first part is about regulatory reform, without which the existing tidal wave of clean energy subsidies will prove wasteful beyond the imagination of even the most skeptical GOP opponents. As James Coleman, a nonresident AEI senior fellow at AEI, and a law professor at Southern Methodist University, explains in his recent paper “Permitting the Energy Transition,” whether this tremendous expenditure by American households will reduce carbon emissions “will depend on how rapidly we can build the new clean energy infrastructure it is designed to fund. Permitting reform will ensure that Americans get something for this unprecedented economic commitment and that the United States can do its part to reduce global carbon emissions.”
Among the key regulatory reform elements suggested by Coleman that would be included in my proposed Energy and Science Act:
Preventing undue judicial delay of federally-approved infrastructure. Place a time limit on how long courts can hold up construction of a project using injunctions. After four or more years of environmental review, courts would no longer be able to enjoin project construction while the government fixes environmental reviews.
Cutting red tape for projects with no significant negative environmental impacts. Allow agencies to approve projects with no significant environmental impacts without extensive documentation or environmental assessments.
Preventing states from blocking federally approved infrastructure. Give federal approving agencies authority to overrule state denials of permits for federally approved projects. The federal agency could overrule if the state decision was unreasonable, conflicted with federal reasoning, or if the project is in the national interest despite state-identified impacts.
The Energy and Science Act of 2025 would really encapsulate my Up Wing, conservative futurist diagnosis of What’s Gone Wrong over the past half century, especially What We Did To Ourselves: too little spending on R&D, too much anti-growth regulation. The next president and next congress need to address both.
Micro Reads
▶ Business/ Economics
Nearly 40 percent of Texas firms use AI with little employment impact so far - Dallas Fed
Tech Investor Sean Parker Leads Rescue of Struggling AI Startup - WSJ
Nvidia’s Stock Tumbles. What Does it Say About A.I.? - NYT
Emotions and Subjective Crash Beliefs - NBER
The A.I. Boom Has an Unlikely Early Winner: Wonky Consultants - NYT
Will Debt Sink the American Empire? - WSJ
Is Software Eating the World? - NBER
▶ Policy/Politics
Vestager slices into Apple again - Politico
Music Labels Take On AI Startups With New Lawsuits - WSJ
Challenging China’s Electric Vehicle Export Surge: US and Europe Diverge - AEI
▶ AI/Digital
Political deepfakes top list of malicious AI use, DeepMind finds - FT
▶ Biotech/Health
Supershoes are reshaping distance running - MIT Tech Review
▶ Clean Energy/Climate
The giant Exxon project that could create the world’s last petrostate - FT
▶ Robotics/AVs
Self-healing 'living skin' can make robots more humanlike — and it looks just as creepy as you'd expect - Live Science
▶ Space/Transportation
Pooping on the Moon Is a Messy Business - Wired
China Becomes First Country to Retrieve Rocks From Moon’s Far Side With Chang’e-6 - NYT
DARPA Teases Newest 'X-Plane' Images as Mysterious Hybrid Electric Propulsion Craft Gets Official Designation - The Debrief
NASA’s commercial spacesuit program just hit a major snag - Ars Technica
▶ Up Wing/Down Wing
Europe May Be Declining, but America Isn’t - WSJ
Why waiting is the hardest part: Science shows us how uncertainty fuels anxiety - Study Finds
▶ Substacks/Newsletters
Will We Ever Get Fusion Power? - Construction Physics
Thoughts on Leopold Aschenbrenner's “situational awareness” - Understanding AI
The triumph of electromagnetism over thermodynamics - Noahpinon
Harvesting Sunlight - by J.K. Lund - Risk & Progress - Risk & Progress
The Next Generation of Robotic Surgeons are Arriving - AI Supremacy
Advanced Nuclear Learns to Share the Dance Floor With Renewables - The Breakthrough Journal
The Environmental Trinity - Jesse H. Ausubel
Breaking software bottlenecks - AI Prospects
Latent Expertise: Everyone is in R&D - One Useful Thing