Seventeen Years of Flat Power, and a Plan to Bring the Factories Home

Start with a number Helberg has only recently sat with: US electricity supply has flatlined since 2008. Seventeen years, no growth. In the same stretch the country has leaned on China for about 90% of the critical minerals it refines and on Taiwan for advanced semiconductors. The macroeconomic shape is textbook: 70 to 80% consumption-driven, more than two-thirds services, manufacturing around 10% of GDP.

That is the book Jacob Helberg is being asked to open. At the time of this No Priors taping he is the under secretary of state designate for economic growth, energy, and the environment, still waiting on Senate confirmation. Silicon Valley already knows the résumé: co-founder of the Hill and Valley Forum, which sits Washington policymakers next to tech companies, and author of *The Wires of War*, on technology and the global struggle for power. Sarah Guo and Elad Gil want a working list, not a vibe. He gives them one: brittle supply chains, rare earths, AI as a productivity engine, and nuclear as the main course.

His opening statement to the Senate, he says, put supply-chain security first. The US has one of the best innovation ecosystems in the world, but it is sitting on a system that can be disrupted. New partnerships abroad and reshoring at home are, in his telling, how you give builders the tools to ship products people love. Manufacturing’s share of GDP has not moved yet. He calls that a lagging indicator. The leading one is already twitching.

Helberg describes the current decade as a double exposure: Trump administration economics laid over very fast AI progress. He ticks through a first-day energy executive order, fast-tracked data-center permits, and a surge in oil, gas, nuclear, and what the president calls “clean, beautiful coal.” The mix, he says, is changing the makeup of the economy. Private capex is already over 2% of GDP and, in his projection, likely to double next year. Tax incentives matter. Permitting speed matters more. A company deciding whether a plant takes seven years or five is, he repeats, the variable that actually moves a board.

Gil asks where the money is going—defense, space, industrials, biotech? Helberg cites analysis that AI infrastructure alone added a full percentage point of GDP in the past year. The economy grew 3%, so roughly a third of growth was AI buildout. Energy infrastructure is the other obvious bucket: national energy demand is up for the first time since 2008. Raw industrials are in the mix too, including a $750 million Defense Department partnership with MP Materials to reboot domestic rare-earth magnet production.

The last leading indicator is global rearmament. World defense spending hit a record $2.7 trillion this year, 60% of it from a handful of large countries: the United States, China, Russia, India, and Germany. The trillion-dollar question, Helberg says, is what they buy. Inefficient capital allocation can leave a government with a paper-tiger military. Ukraine is already a window: AI and autonomy are changing battlefield outcomes. Where that $2.7 trillion lands will help define 21st-century hard power.

Why make supply chains the first fight? Because China is a systemic rival with a different theory of how the world should be run, and a lot of its overseas presence is an outflow of being the world’s factory floor. Africa, Latin America, the Belt and Road: Helberg reads them as byproducts of manufacturing dominance. China imports raw materials from Africa, makes the goods at home, and re-exports them. Solve the trade imbalance, he argues, and you loosen the peripheral influence problem in third markets. It is a national-security issue. It is also a corporate one. After Beijing’s rare-earth magnet export controls, America’s best companies have had to ask Beijing for licenses to build cutting-edge products. “The last thing we need,” he says, is that posture.

Gil notes that rare earths are badly named. There are large deposits in Canada, the United States, and India; the choke is refining, not geology. Helberg adds a timeline. China became a rare-earth superpower only about a decade ago. After Made in China 2025 in 2015, refining capacity skyrocketed, the market was flooded, prices sank, and Western refineries in Australia, Canada, and the US were squeezed. The US has had refineries before: a large one in Tennessee, others in Arizona and Georgia. The genie does not have to stay out.

The DoD–MP Materials deal is the blueprint he wants copied: an anchor buyer, an end customer, and a price floor. China’s playbook against Western refineries is, in his words, classic monopolistic behavior—artificially depress the global price, put the alternative out of business, then raise it again. Offtake agreements plus a floor break that loop. Transportation logistics is the other missing layer. Belt and Road already includes a global network to move African minerals to Chinese refineries and back out to the world. The US used to do analogous work with the Panama Canal. Helberg wants autonomous technology to leapfrog old infrastructure and redesign how goods move.

For decades the textbook said mature economies evolve into services and manufacturing leaves. American wages, in that story, are simply too high to bring factories home. Guo interrupts: Europe already violated the narrative. Germany kept an industrial base. Helberg agrees the US version became self-fulfilling. AI, he says, gives advanced economies a chance to break it. Raise productivity enough and the labor-cost advantage of developing countries erodes. Reindustrialization gets a fulcrum.

He borrows Peter Thiel’s *Zero to One* split. Horizontal competition is globalization; vertical competition is innovation. In the 2000s America grew sideways and not much upward. The last seven years, he argues, vertical growth has picked up. History is his exhibit. During the first industrial revolution Britain, tiny by population, had industrial output per capita more than 50 times China’s. China’s share of the world economy fell from about a third in 1800 to 7.5% by 1913. Today Nigeria is huge and Israel is smaller than New Jersey, with a bigger GDP and a stronger military, because of technology. People who say America cannot reindustrialize because China has more people, or because US workers cost more, are “totally missing the plot.”

He does not think AI is about to replace humans wholesale. He thinks it gives workers superpowers. The macro theory he has become fascinated with is Jevons paradox: a technology that massively raises efficiency tends to increase demand for the resource, because its relative cost falls. LED lighting made illumination cheaper, and the world turned on more lights. He expects the same pattern in compute and manufacturing. The arithmetic is a fork. A $10 trillion economy where agentic AI lets ten people do what a hundred did can either run with far fewer workers or produce ten times more. He bets on the second fork. Human wants are unlimited—look at what ordinary people consume versus their grandparents—and firms will raise output because rivals will. Workers will not vanish. Their range of work gets much wider.

Ten years ago the fashionable story was that the 21st century would belong to China, or to a rising East. Helberg’s counter-narrative is that the defining feature is the rise of superintelligence. First-mover economies that fold AI into production, he predicts, will reap a second great divergence, the first since the industrial revolution, and leapfrog laggards. A byproduct is the collapse of the cheap-labor advantage developing economies have lived on for 50 years. The plot twist the press has underplayed is Europe. Its share of global GDP went from 65% in the early 20th century to roughly a third in the 1980s and 1990s to 15% now.

Europeans, he notes, still blame the 1970s oil shock. He blames missed technological boats: late to the internet, late to consumer apps, and now the EU AI Act and a digital services tax. The AI Act, in his phrasing, is “single-handedly ensuring that Europe will not be a first mover in AI” because the rules are incredibly punitive. The other plot twist is the Middle East. GDP per capita in the UAE and Israel is now higher than in France, and higher than in South Korea. Western commentary still treats the region as war-torn. Helberg sees tech-forward leaders, and he treats Iran’s diminished regional influence after recent conflict as a silver lining that makes a less hostage-to-proxies Middle East possible.

The remaining features are a US–China race to control the world’s AI scaffolding, and the rearmament already discussed. Most countries will eventually import intelligence. They will not all need “super fancy Blackwell chips.” They will need ordinary compute. Who sells it—NVIDIA or Huawei—matters because China will bundle the stack: Ascend, DeepSeek, and a box of Chinese tools, competing hard for share. The United States, he says, needs a strategy for the Global South and third markets. On open source, Gil asks what Washington should do when Chinese labs push open weights with government wind at their backs. Helberg wants a plan to promote the American stack, open or otherwise. His DeepSeek read is harsh: incremental efficiency gains, a lie about compute (he says they have a billion-dollar cluster), and distillation of ChatGPT’s weights. “Open source” in quotation marks, if the weights came from a closed model. He still thinks open source will be central to China’s market-share push, which is why Meta’s effort matters, inside a holistic push to put the best models in the widest use.

Guo asks what the Middle East is in this fight: swing vote, capital, compute partner? Helberg sees a new kind of partner. Leaders there say they want a more pro-American direction. The US is energy-constrained; expanding supply takes time. Cheap, abundant Gulf energy lets American companies compete with China’s state-backed pace on raw power, compute, and speed—if Washington can write a security framework that keeps China off those clusters. He thinks that is eminently doable, and says the administration is looking hard at it.

Asked how the US actually closes an energy gap that some people describe as a doubling of production—more if factories come home—he does not offer a menu. “I think we need nuclear energy.” Two hundred years of history, in his telling, tie energy cost to growth. US electricity is half the European price, and that gap shows up in GDP. The obstacles are familiar: plants that cost tens of billions, build times that can run seven years, and regulatory uncertainty that can turn a five- or six-year project into twelve. Gil underlines the finance: delays do not add cost linearly. Interest and legal fees compound. The time value of money is the core variable. Compress the window and the entire project economics rewrite.

Capital can come from a policy flip. CFIUS has historically scrutinized foreign money in nuclear and other critical infrastructure. Helberg wants a lane for trusted partners so the huge investment commitments the president has already secured can flow into productivity-enhancing energy. France, for all its red tape, gets 75% of its electricity from nuclear. The US sits at about 17 to 18%, with almost no new capacity since the 1970s and a strong safety record. Guo’s observation is that data-center demand now matches nuclear scale: companies want to eat an entire large plant. Helberg calls the destination a new country—high investment of 2 to 4% of GDP in private capex, productivity compounding, a Goldman Sachs growth band of 3 to 5% instead of 1 to 3%, and GDP tilting from consumption and services back toward manufacturing. Nuclear, gas, and coal all count. Nuclear is the turbocharger.

Pressed for other domains that are essential to security, Helberg returns to a layered pyramid: energy, minerals, component manufacturing, semiconductor manufacturing, data centers, models, and apps. The country is in decent shape at most layers. The biggest exposure points remain components, chips, and minerals. Transportation logistics is the under-discussed extra. China already designed a full-stack move of goods. America, he says, used to make those bets—the Panama Canal again—and now has a president with appetite for an economic agenda that has been sitting idle. Autonomous logistics is how you skip the old roads.

For a tech audience, the practical question is how to deal with this administration. Helberg’s description is “builder friendly,” with a builder in the White House. The last six months, in his phrase, have been shock therapy: deregulation, lower taxes, a pitch for foreign capital, America as the best destination for money. He and David Sacks hosted an AI summit with the president, who declared that America started the AI race and would win it. Helberg heard an echo of Kennedy’s moon speech: optimism, bullishness, and boldness he thinks the White House actually has to show.

The working translation for operators is three modeling choices. Re-run factory, data-center, and energy projects on a five-year clock instead of seven; permitting compression changes IRR more than a slogan does. Study the MP Materials contract, because anchor-buyer-plus-price-floor is likely to spread across other critical minerals and heavy industry. And plan AI as a way to expand output, not as a layoff machine. Jevons is already the competitive logic: if you do not use the cheaper capacity to make more, someone else will.