Scott Bessent wants to sit down with He Lifeng and talk about artificial intelligence and rare earths as if they are line items on a bilateral trade spreadsheet. That is the consensus view echoing across every financial terminal in Manhattan and every think tank in D.C. It is also completely, catastrophically wrong.
The lazy narrative says America holds the high ground in silicon and software while Beijing holds the chokehold on the periodic table, and these two titans are about to horse-trade their respective monopolies.
I have watched corporate boards and government agencies blow tens of millions of dollars chasing this exact framing for the last five years. It is a comforting illusion for bureaucrats who love clean categories. Algorithms over here. Magnets over there. A neat little exchange of digital dominance for physical minerals.
Wake up. The physical and the digital are no longer separate domains. They are the exact same weapon system.
When Washington treats rare earths as raw commodities to be secured through mining subsidies and tariff waivers, it demonstrates a fundamental misunderstanding of what these elements actually represent in the twenty-first century. Neodymium, dysprosium, and yttrium are not just dirt found in the ground. They are the physical metabolism of computation. You cannot run hyperscale neural networks, guide autonomous guidance systems, or power electric propulsion without permanent magnets that maintain thermal stability under brutal operational stress.
China understands this at a cellular level. Beijing did not restrict gallium, germanium, antimony, and graphite exports because they wanted extra leverage for a trade deal. They did it because they realized years ago that controlling the refinery bottleneck of the material world renders America's software supremacy entirely theoretical.
What happens when you have the most advanced transformer model on the planet, but the motors driving your thermal cooling systems burn out because the specialized heavy rare earth supply chain is completely severed? You have a very expensive paperweight.
The prevailing media coverage wants you to believe that Bessent and He are negotiating a grand bargain where the United States trades compute access for mineral shipments. That assumes a symmetric exchange. It assumes America's AI advantage is permanent and China's processing monopoly is fragile. Both assumptions are laughably dated.
Let us look at the data the financial press conveniently ignores. America's rare earth processing capacity is still largely a rounding error compared to China's refining juggernaut. Even when a domestic mine pulls ore out of the ground in California or Texas, the concentrate frequently gets shipped straight back to Asia for separation and reduction because our environmental permitting and chemical processing infrastructure has atrophied over four decades of de-industrialization.
Conversely, China's indigenous AI hardware ecosystem has absorbed wave after wave of export controls and simply routed around them. By forced necessity, Chinese semiconductor firms and tech giants have optimized architectures around lower-yield lithography, accelerated domestic packaging innovation, and built alternative interconnect fabrics. They stopped trying to play catch-up to Nvidia's exact playbook and started rewriting the rules of hardware-software co-design.
So when Bessent brings up artificial intelligence in these talks, he is talking about yesterday's war. America is trying to protect an architectural moat built on software layers, while Beijing is weaponizing the physical substrate that makes those layers run.
To understand why this diplomatic dance is missing the target, we have to dismantle the core premise driving modern economic statecraft: the belief that technological leadership can be sustained independently of industrial depth.
For thirty years, Wall Street and Washington bought into the myth of the asset-light economy. We outsourced the smokestacks, the chemical refineries, and the dirty refining processes, convincing ourselves that we could maintain dominance by keeping the high-margin intellectual property right here at home. We became a nation of designers, financiers, and software architects, assuming the physical world would always supply itself cheaply and quietly in the background.
That model is dead. It died the moment global supply chains transformed from commercial efficiency engines into instruments of geopolitical coercion.
When you separate software design from physical manufacturing, you create a brittle illusion of power. American tech companies can code all day long in Silicon Valley, but if every single high-performance electric drive, precision servo-motor, and defense sensor relies on a refining supply chain controlled by a strategic competitor, your intellectual property is nothing more than digital ransom waiting to be collected.
This is why framing the Bessent-He talks as a simple negotiation over tech rules and mineral quotas is dangerous theater. It treats symptoms while ignoring the structural disease.
If Washington actually wants to shift the balance of power, it needs to stop asking how to buy more rare earths from China or how to stop them from buying advanced chips. The question we should be asking is why we spent forty years convincing ourselves that we didn't need to know how to build basic industrial chemistry in our own backyards.
The conventional wisdom dictates that the solution is simple: throw more subsidies at domestic mining companies, pass another clean energy bill, and wait for the private market to spin up alternative supply chains.
That approach fails because it misunderstands the timeline of industrial capacity. You cannot subsidize your way out of a thirty-year engineering deficit in thirty-six months. Building a domestic rare earth processing ecosystem is not a matter of writing a check; it requires a generational commitment to chemical engineering, hazardous waste management, and skilled trades that our education and economic systems spent decades devaluing.
There is an ugly downside to my own argument that I must acknowledge. A hyper-nationalist decoupling of the physical and digital supply chains is hideously expensive, fiercely inflationary, and guaranteed to slow down the pace of global technological innovation. Rebuilding domestic refining capacity means living with environmental remediation fights, high initial capital costs, and short-term consumer pain. There is no clean, frictionless way to decouple an integrated global economy.
Yet the alternative is far worse. The alternative is continuing to pretend that a trade envoy can talk his way past a structural monopoly with a clever compromise on export tariffs.
When Bessent and He sit down at the negotiating table, pay less attention to the joint communiques about artificial intelligence cooperation and mineral stability. Look instead at what both nations are doing away from the microphones. China is quietly cementing its control over the downstream manufacturing nodes where hardware meets material. Meanwhile, Washington is discovering that you cannot code your way out of a material shortage.
The real war is not being fought over software licenses or mining quotas. It is being fought over who controls the physical reality underlying the digital future.
Stop watching the software. Look at the dirt.
