Paper 05: Build Big Things

Plank: Freeze reactor designs, rebuild the shipyards, and get to a full domestic chip stack in twenty years, funded and structured to survive a change of party in the White House.

1. What we want

We want the federal government to pick three nuclear reactor designs, freeze them, and order fifty of each, so the country finally rides the same cost curve every other country that builds reactors in volume has ridden. We want the shipyards rebuilt, because a Navy that cannot build ships at scale cannot project power, and the country that now outbuilds us in commercial shipping tonnage will keep outbuilding us in warships too. We want the full chip supply chain back on American soil within twenty years, including the extreme ultraviolet lithography machines at the very top of it, because a country that cannot make its own advanced chips does not control its own industrial future. And we want every one of these programs funded and structured the way the Interstate Highway System was, not the way Apollo was: dedicated revenue that keeps flowing without a new vote each year, and construction work spread into enough congressional districts that no single election can kill the program.

2. The evidence

Start with cost. Georgia Power's Vogtle Units 3 and 4, the two new AP1000 reactors that came online in 2023 and 2024, ended up costing roughly $35 billion for 2.2 gigawatts of capacity, against an original budget of about $14 billion [1][2]. That works out to roughly $16 billion per gigawatt. Unit 3 was supposed to enter service in April 2016; it actually started commercial operation in July 2023, seven years late. Unit 4 was supposed to follow in 2017; it arrived in April 2024 [2].

Compare Korea. Korea Hydro and Nuclear Power built four APR-1400 reactors at Barakah in the United Arab Emirates for roughly $4,000 to $4,500 per kilowatt, or $4 billion to $4.5 billion per gigawatt [3], a quarter of Vogtle's cost, using a design KHNP had already built repeatedly at home. France's fleet tells the same story from the build side. Between the early 1970s and the early 1990s, EDF built the fleet of 56 reactors still running today, using a small number of standardized designs: three size classes (roughly 900, 1,300 and 1,450 megawatts) with six design variants across them, rather than a new plant engineered from scratch each time [4][5]. That is six related designs built repeatedly across three size classes, and it is still radically more standardized than anything the United States has done. Today France gets roughly two-thirds of its electricity from nuclear power; the United States gets about 18 percent, from 94 reactors [4][6].

At the V.C. Summer plant in South Carolina, two more AP1000s were under construction as a sister project to Vogtle, using the same design. This provided the clearest evidence of the cost problem during Vogtle's own build, when the second set of reactors should have been getting cheaper. Crews and suppliers were expected to bring those costs down as they climbed the learning curve. Instead, Westinghouse, the contractor building both projects, filed for Chapter 11 bankruptcy in March 2017 after roughly $9 billion in losses on the two jobs. Four months later, in July 2017, the V.C. Summer owners walked away and canceled their project outright [7]. Vogtle carried on alone, which meant it absorbed the entire first-of-a-kind engineering and tooling cost of the AP1000 program by itself. The learning curve that was supposed to bring the price down for the second plant never got a second plant to apply to.

On shipyards, an unclassified U.S. Navy intelligence briefing slide that circulated in 2023 put Chinese shipbuilding capacity at roughly 232 times that of the United States, based on gross tonnage capacity: about 23.2 million tons in China against under 100,000 tons here. A Navy spokesperson confirmed the slide was genuine, while cautioning it was a rough estimate rather than a rigorous study [8]. Rough or not, the order of magnitude is not in dispute: the United States is not close.

The Dutch company ASML makes the extreme ultraviolet lithography machines required to print the most advanced chips. It is the only company in the world that makes them; no competitor, American or otherwise, currently builds one [9]. The 2022 CHIPS Act appropriated $52.7 billion for domestic semiconductor manufacturing, research and supply chain work. That funding was split roughly $39 billion for manufacturing, $13.2 billion for research, and the rest for supply chain coordination. The law also provided a 25 percent investment tax credit, and the appropriation runs through fiscal year 2027 [10]. The money is helping build fabrication plants, but it does nothing to fund the machines those fabs depend on. It runs as a multi-year appropriation through 2027 rather than a dedicated revenue stream, which matters for the next section.

Finally, the funding-design comparison behind the whole plank. The Highway Trust Fund was created by the Highway Revenue Act of 1956, financed by a dedicated federal gas tax, and it paid for the Interstate Highway System without Congress needing to reauthorize the money every year. The system was declared complete in 1992, when the last segment, a stretch of Interstate 70 through Glenwood Canyon, Colorado, opened to traffic, 36 years after the program started and 17 years past its original 1975 target [11]. NASA's budget, by contrast, peaked in 1966, three years before Apollo 11 landed on the Moon, and was already falling by the time the landing happened, because it depended on an annual vote and a sense of national threat that faded once the Soviets were beaten [12]. Once the landing happened, the remaining missions were cut fast: Apollo 20 was canceled in January 1970, and two more missions were canceled that September as part of a budget reduction [12]. Apollo cost about $25.8 billion between 1960 and 1973, or roughly $257 billion in 2020 dollars [12], and none of it kept flowing once the goal was met, because none of it was built to.

3. Why it happened

The United States has never built the same reactor design enough times to bring costs down through the learning curve, which explains why nuclear power is expensive here. American engineers can build cheap reactors, but the industry and its regulators have never given them the chance to build one design fifty times. The V.C. Summer cancellation in July 2017 is the clearest single case of this failure, because it ended the one project that could have proven the point. It was a second AP1000 order using the same design as Vogtle, and it collapsed under first-of-a-kind engineering and construction costs. The AP1000 program had front-loaded those costs onto its first two sites, with the expectation that later orders would benefit from the learning curve. Cancelling it guaranteed that no later AP1000 order ever got the benefit of the learning curve those costs were supposed to buy.

Shipbuilding has a cleaner, more precisely dated cause. Since 1936, the federal government had run a construction differential subsidy that covered up to half the cost gap between building a commercial ship in a U.S. yard and building the same ship abroad. In 1981, the Reagan administration terminated funding for that subsidy as part of a broader push to get government out of private markets. The commercial shipbuilding market collapsed within a few years: annual commercial vessel construction in U.S. yards fell from an average of 72 ships a year in the 1970s to 21 a year in the 1980s, oceangoing commercial orders dried up almost entirely by decade's end, and nearly half the country's private shipyards closed [13]. American shipyards became, in effect, a one-customer industry serving the Navy alone, which is a much smaller and much less steady order book than a yard needs to keep skilled crews and suppliers in business. The subsidy cut was not the only thing hitting shipbuilders that decade: a global shipbuilding depression and escalating foreign subsidies, especially from Japan and South Korea, were squeezing yards worldwide at the same time, and American yards would have lost some orders regardless. But no other major shipbuilding nation cut its own support to zero in the middle of that downturn the way the United States did, which is why the U.S. collapse was total where competitors' was not.

Chips have a similarly specific moment, and it runs through a piece of technology America built and then let go. Starting in 1997, the EUV LLC consortium, funded mainly by Intel, joined by Motorola, AMD, IBM and Micron, and working inside U.S. national laboratories at Sandia, Livermore and Berkeley, did the foundational research on extreme ultraviolet lithography [14]. Two of the three toolmakers inside that consortium were American: Silicon Valley Group of San Jose and a spinoff of Ultratech. In 2001, ASML moved to buy Silicon Valley Group. Congress asked for a 45-day national security review of the deal, conducted by the Committee on Foreign Investment in the United States under the Exon-Florio provision and scheduled to conclude on April 23, 2001, over objections from a group warning about the loss of defense-relevant technology [15]. The deal closed that year. That review was the moment the government had to decide whether to keep the successor technology to modern chipmaking in American hands, and the deal went through anyway. Cymer, a San Diego company that developed the laser light source EUV machines depend on, was later acquired by ASML as well, in 2013. None of this was inevitable or accidental. A specific government review cleared this transaction on a specific date.

Underneath all three is the same funding-design failure the Interstate avoided and Apollo did not. Vogtle, the shipyards, and the CHIPS Act all run, or ran, on money that has to be re-voted, re-appropriated, or re-justified on a rolling basis, which means every one of them is exposed to a change in the political weather in a way the Interstate, running on dedicated gas-tax revenue that did not require a new vote, never was.

4. What to change

Federal: freeze three nuclear reactor designs at different sizes and order fifty of each, so there is one supply chain and one trained crew per design instead of a dozen competing one-off projects. Rebuild the shipyards, because naval capacity depends on having yards capable of building at volume over sustained years, the way a wartime industrial base has to. Commit to a twenty-year program to reconstitute the full domestic chip stack, including lithography. Fund all three programs the way the Interstate was funded rather than the way Apollo and the CHIPS Act were: dedicated revenue rather than an annual appropriation, with the program continuing by default and requiring an affirmative vote to stop it, rather than an affirmative vote to keep it going.

State: reform siting and transmission permitting so that a reactor or a shipyard expansion approved on the merits does not still take a decade to get a line run to it. Fund trades training programs tied to specific, named projects, so welders, pipefitters and electricians are trained for work that is already under contract, ahead of a hiring date they can see on the schedule.

Local: towns that host a reactor or a shipyard get the surplus power the plant produces at a discount, so the benefit of hosting the project is visible and immediate to the people living next to it, on their own power bill, and worth more than a promise on a fact sheet.

5. What a bill would do

The federal bill would direct the Nuclear Regulatory Commission to certify three specific reactor designs at different capacity classes and would appropriate funds for the government, acting as anchor customer alongside utilities and large power buyers, to order fifty units of each. The bill has to name the designs and cap the count at three because otherwise it will attract amendments adding a fourth and then a fifth, eventually recreating the design-sprawl problem the country already has with small modular reactors. A companion title would authorize shipyard capital investment through long-duration contracts with cancellation penalties rather than annual appropriations, so a yard that wins a contract can hire and train crews against a schedule that will not disappear in the next budget cycle. A third title would establish a twenty-year chip stack program with its own dedicated revenue stream, structured as a trust fund the way the Highway Trust Fund was. The program would continue automatically unless Congress affirmatively votes to end it, and this default-on clause is essential because it makes the program survive a change of administration.

Every title requires geographic distribution of contracts and construction work across a minimum number of states and congressional districts, written into the bill rather than left to agency discretion, because the geographic spread is what creates a constituency willing to defend the program in Congress later. Companion state-level model legislation would streamline siting and transmission permitting timelines and set up trades training grants conditioned on a signed contract for a named project rather than a general workforce fund. A final clause requires host communities to receive a defined discount on power from any reactor built under the program.

6. What would prove us wrong

If the third and fourth reactors built under a frozen design do not come in meaningfully cheaper than the first two, the way Korea's and France's did, the standardization theory is wrong and the money should stop. If shipyard capital investment does not translate into a measurable increase in vessel throughput, tracked in ships delivered per year, within a decade of funding, the shipyard title has failed. If twenty years from now the United States still has no domestic capability at or near the lithography frontier and remains entirely dependent on foreign toolmakers, the chip stack program did not work and should be said so plainly. And if the program structure itself does not hold: if a change of party control in Congress and the White House cuts these programs the way NASA's budget was cut after 1969, then the design lesson from the Interstate did not actually transfer, and that failure belongs to how we built the program, while the underlying case for building these things at all still stands.

Sources

1. World Nuclear Report, "Utilities Abandon V.C. Summer AP1000 Reactor Construction Following Westinghouse Bankruptcy," 2017. https://www.worldnuclearreport.org/Utilities-Abandon-V-C-Summer-AP1000-Reactor-Construction-Following-Westinghouse.html

2. Wikipedia, "Vogtle Electric Generating Plant," accessed 2026-09-22. https://en.wikipedia.org/wiki/Vogtle_Electric_Generating_Plant

3. World Nuclear Association, "Nuclear Power in the United Arab Emirates," accessed 2026-10-04. https://world-nuclear.org/information-library/country-profiles/countries-t-z/united-arab-emirates

4. World Nuclear Association, "Nuclear Power in France," accessed 2026-09-22. https://world-nuclear.org/information-library/country-profiles/countries-a-f/france

5. Relance2030, "The French Nuclear Fleet — 56 Reactors, 70% of Electricity, and the EPR Program," accessed 2026-09-22. https://relance2030.com/encyclopedia/french-nuclear-fleet/

6. U.S. Energy Information Administration, "U.S. Nuclear Industry," 2025-2026 data. https://www.eia.gov/energyexplained/nuclear/us-nuclear-industry.php

7. Wikipedia, "Nukegate Scandal," accessed 2026-09-22 (V.C. Summer cancellation and Westinghouse bankruptcy timeline). https://en.wikipedia.org/wiki/Nukegate_scandal

8. The War Zone, "Alarming Navy Intel Slide Warns Of China's 200 Times Greater Shipbuilding Capacity," 2023. https://www.twz.com/alarming-navy-intel-slide-warns-of-chinas-200-times-greater-shipbuilding-capacity

9. The Review (Strange VC), "ASML's 30-Year Monopoly: The Moonshot Bet No One Can Replicate," 2026. https://thereview.strangevc.com/p/asmls-30-year-monopoly-the-moonshot

10. U.S. Congress, Senate Commerce Committee, "The CHIPS Act of 2022 Section-by-Section Summary," 2022. https://www.commerce.senate.gov/wp-content/uploads/media/doc/CHIPS%20Act%20summary.pdf

11. American Society of Civil Engineers, "History Lesson: The Last Piece of the Interstate Highway System May Have Been the Toughest," Civil Engineering Source, 2021. https://www.asce.org/publications-and-news/civil-engineering-source/civil-engineering-magazine/issues/magazine-issue/article/2021/03/the-last-piece-of-the-interstate-highway-system-may-have-been-the-toughest

12. The Planetary Society, "How Much Did the Apollo Program Cost?," accessed 2026-09-22 (budget peak, mission cancellations, total cost). https://www.planetary.org/space-policy/cost-of-apollo

13. Congressional Research Service, "U.S. Maritime Administration (MARAD) Shipping and Shipbuilding Support Programs," R46654 (Construction Differential Subsidy termination in 1981 and the decline from 72 to 21 commercial vessels built per year). https://www.congress.gov/crs-product/R46654 ; American Maritime Voices, "Michael Lind: A Libertarian Radical Sank America's Maritime Industry. It Needs to Be Restored," accessed 2026-09-22 (shipyard closures and one-customer-industry effect). https://americanmaritimevoices.org/michael-lind-a-libertarian-radical-sank-americas-maritime-industry-it-needs-to-be-restored/

14. Construction Physics (Brian Potter), "How ASML Got EUV," accessed 2026-09-22 (EUV LLC consortium founding and members, Silicon Valley Group acquisition). https://www.construction-physics.com/p/how-asml-got-euv

15. EE Times, "ASML Refutes Claim That SVG Purchase Will Threaten EUV Technology," 2001 (45-day national security review of the ASML-SVG deal). https://www.eetimes.com/asml-refutes-claim-that-svg-purchase-will-threaten-euv-technology/