The Situation

The United States will reportedly be telling other countries they must pick sides in the artificial intelligence race with China. This is related to a U.S. December 2025 initiative known as the Pax Silica, an initiative aiming to secure global supply chains for AI and other advanced technologies by countering Chinese market dominance in areas like critical minerals and semiconductors.

The news follows a speech made last month by Chinese President Xi Jinping, who extolled the virtues of an open weight AI ecosystem at the World Artificial Intelligence Conference in Shanghai. The irony is delicious. An authoritarian Communist country is supporting an open weight ecosystem against a democratic market-driven economy’s attempts to control access to a transformative technology.

China’s support of open weight models is clothed in magnanimity (China insists it will provide resources to support the development of AI, a “digital public good” in the Global South countries, for the small price of accepting Chinese standards and governance norms that come with it). But in truth, it belies where China is behind. China is resorting to open weights because it lags the U.S. in terms of computing power and AI capex. China is also exposed to U.S. export controls on advanced AI chips and other critical technology inputs.

And China’s approach is working. Open-source Chinese AI models are demonstrating capabilities close to OpenAI and Anthropic at a lower potential cost to the consumer and with more efficient use of compute – because they don’t have a choice. By making the system open source, China is enlisting the help of any smart person anywhere in the world to improve this AI ecosystem.

In this, the Chinese government is demonstrating a better grasp of the potential economic benefits of the AI revolution (and yes, I do think it’s a true tech revolution) than the U.S. government. The most recent tech revolution was in information and communication technology (ICT). It is difficult to overstate just how this revolution changed our lives. A modern home is equipped with devices and technologies that seem mundane today, but would have been mind-blowing to someone from the 1950s.

The ICT revolution also created a huge amount of wealth – the titans of Silicon Valley, the Mag-7, the tech-bros, the investors who gobbled up shares of Apple, Microsoft, and others in the 1990s and are sitting on piles of ever-increasing wealth as a result.

Artificial intelligence is a different kind of technology. In a provocative essay published in Colossus last year, retired venture investor Jerry Neumann compared AI to containerization. The winners of the containerization story were not tech start-ups and their founders, employees, and investors – but instead individual consumers who could suddenly buy more goods at lower prices because of the way containerization greased wheels of globalization. If containerization was about the physical economy, AI then is about the containerization of the digital economy – and its gains will be enjoyed by individuals in the form of much lower prices for things like legal and accounting services.

Dror Poleg, an economic historian, sharpens this view: He describes AI as a “medium.” No, not as a way to commune with the dead – but as an alternative to thinking of AI as a “tool.” Our cell phones, the laptop on which I’m typing – these are tools. Tools make tasks easier, but they don’t fundamentally change the world. A medium by comparison is the substrate everything runs on. One of Dror’s metaphors for AI is the phonograph – which did not destroy the music industry, but did change it. Rather than asking whether AI will replace our jobs, Dror suggests asking, “how will AI change the income distribution of the game I’m playing?”

It is not surprising that the U.S. is jealously clutching its AI pearls. As the war on Iran and the ongoing closure of the Strait of Hormuz has demonstrated, the U.S. economy no longer runs on oil – it runs on AI.

The statistics on AI’s importance to the U.S. economy are mind-boggling. Data centers account for over three percent of all U.S. construction. Private U.S. data center construction spending has surpassed public spending on all transportation – airports, marine terminals, mass transit, you name it. In June, the U.S. Commerce Department reported that the annual rate of outlays on data centers was up almost 50 percent year-on-year, to $68.3 billion. Outlays on all other private construction – houses, hospitals, schools, underwater basket weaving academies – fell by $101.6 billion over the same time-period. If the AI train stops, the economy tanks.

This is where China’s deeper advantage lies. AI is not going to become a medium or a harbinger of digital containerization, or anything else for that matter, unless the physical infrastructure  necessary for its build out happens along-side it. China may be lagging when it comes to compute, but when it comes to building out infrastructure, there is no country in world history that can match China’s capabilities.

Right now, the focus in the U.S. is on the data center build out. But the bottlenecks in bringing AI to the edge are acute. The U.S. has cheap energy, but its power infrastructure is antiquated. It’s not built for the level of demand that is coming. “Boring” long-lead electrical gear like transformers, switchgear, and grid-tie batteries are seeing significant (30 months+ ) and increasing lead times. High bandwidth memory is basically sold out through 2026, and much of that market is dominated by South Korea, which the U.S. just slapped in the face (more on that below). McKinsey projects optical transceiver production will fall short by 40-60 percent of demand through 2027.

Even in something as boring as optical fiber cable, for which data centers have gone from accounting for 5 percent of global demand to ~30 percent by 2027, the U.S. is short. Fiber optic cable is a rare product in that the U.S. still controls production at significant levels (either domestically or in Mexico). China’s dominance on a critical mineral (germanium) and preforms/glass are creating fears of shortages.

China’s open weight system is softening its weaknesses. Pax Silica does not soften the U.S. weakness – it shines a light on it.

Another weaknesses, which both the U.S. and China share, is an aging population. China’s position is materially worse here than the U.S., but it is worth noting that U.S. population growth slowed to 0.5 percent in 2025. U.S. net migration has plunged to near historic lows, which the current U.S. administration celebrates even as it creates labor shortages and wage inflation. The U.S. ace-in-hole for cheap labor – Mexico – is mired in USMCA melodrama.

China has responded to its demographic problem by attempting to spur consumption from its lower middle class (not going so well) and by installing one out of every two industrial robots in the world (succeeding wildly). Chinese manufacturing workers are losing their jobs to robots and autonomous machines – precisely because China can build out the infrastructure necessary to handle the massive amounts of data and power necessary to take advantage.

To blunt the effects of slowing population growth, the U.S. must either a) have more babies, b) attract more migrants, or c) build lots of robots. The data-center build out and the AI race should put the U.S. in the cat-bird seat for c), but the U.S. is not doing that. It is making high-falutin plans to replace critical mineral dependence on China with domestic or other sources over the next 5-7 years, thinking that will be enough.

The U.S. is ahead in the AI “race,” but losing ground and without an effective strategy for maintaining its technological primacy. A Communist rival is beating the U.S. at its own game. Meanwhile, the focus on the admittedly gaudy capex going into data-center build out in the U.S. is obscuring the extent to which the rest of the U.S. economy is on shaky ground (to put it nicely).

It also reveals a deep misunderstanding: the “winners” of the AI revolution will not be the toll-collectors for compute that will continue to get cheaper, but to the downstream users who combine cheap data, energy, and information to compete in ways they couldn’t before. According to a recent Gallup poll, 71 percent of Americans oppose the construction of a data center in their area. That’s a combination of ignorance and mistrust in institutions. In five years, those same 71 percent will be furious if the price of compute is not falling and if AI’s glittering gains are being kept from them because the U.S. has a geopolitical “AI race” to lose.

Map/Chart of the Week

Year Over Year Change in US Population

Blind Spot

On Sunday, U.S. President Donald Trump posted this on social media: “Based on my very good relationship with Kim Jong Un, of North Korea, I am not happy with the fact that the United States has, long ago, agreed to participate in Joint Military Exercises with South Korea. These exercises are not only costly, with much of these costs paid for by the United States of America (as usual!), but send a signal that is totally inappropriate and hostile to a Country that, as long as Donald J. Trump has been President, has been unthreatening and respectful. Therefore, and based on the fact that it is too late to cancel, I have instructed the Sectary of War, Pete Hegseth, to substantially reduce the Joint Military Exercises! While somewhat unrelated (?), I recently asked the President of South Korea if they would like to join us in the Denuclearization of the Islamic Republic of Iran, and they said, “No thanks!” Thank you for your attention to this matter.”

As recently as last month, the U.S. government was pushing for South Korea to help expand and modernize US naval forces. We need the help. The U.S. accounts for ~0.1 percent (that’s not a typo) of commercial shipbuilding capacity. South Korea, by comparison, is number 2 in the world, at 29.1 percent, behind only China’s 53.3 percent. South Korean companies SK Hynix and Samsung produce roughly two-thirds of the world’s memory chips, critical to artificial intelligence infrastructure. South Korea is a U.S. defense treaty ally and has attempted to satisfy the U.S. governments concerns about protectionism repeatedly. South Korea is a major energy importer and like another U.S. defense treaty ally in the neighborhood, Japan, has been severely impacted by the closure of the Strait of Hormuz.

Is South Korea a critical U.S. ally? One can make a case that it is not. The Korean War started in part because then Secretary of State Dean Acheson did not include Korea in the U.S. definition of its security umbrella in East Asia, which the Soviet Union took for tacit permission to encourage North Korean forces to overrun Seoul. They almost did, until the U.S. changed its mind and defended South Korea, which went from being a war-torn backwater to one of the most technologically advanced countries in the world. The division of North and South Korea is a Cold War anachronism – not unlike U.S. vitriol for Iran. And the U.S.’s goals are perfectly maintained by defense treaties with Japan, the Philippines, and Australia, as well as Taiwan’s de facto autonomy.

Still, that is a stretch. In a world where U.S. interests are increasingly in the Pacific, and where China is the biggest geopolitical rival to the U.S., treating South Korea in this manner seems terrifically short-sighted. Moreover, there is no sane case for treating North Korea like a partner, let alone an ally. The U.S. is forfeiting decades of good will with Seoul so that President Trump can support what Christhopher Hitchens so brilliantly called the world’s first and only necrocracy, the last truly totalitarian country left in the world, which has obtained nuclear weapons and also has enough artillery pointed at Seoul to melt the city without having to fire one of its warheads? President Trump’s stated goal is to prevent Iran from getting a nuclear weapon, but he’s ok with North Korea having them?

This isn’t the end of the world, but it is the beginning of the end of the U.S.-South Korea relationship. The latter cannot afford to bet its future on such a mercurial ally. In Casa Blanca, Captain Renault chides Rick Blaine after he rejects a beautiful woman named Yvonne, “How extravagant you are, throwing away women like that!” How extravagant the U.S. is, throwing away allies like South Korea. Multipolarity is in its early stages, but things like this hasten its coming.

Reader Question

Finally… 

What I’m watching: The Silo Season 3

What I’m reading:  Ren Xiao: A Chinese Theory of International Relations, on China’s Own Terms – this is an excellent essay that helps frame Chinese foreign policy analysis, which is extremely different than the West’s.

What I’m listening to: The Undiplomatic Podcast, Ep 314