The Chip War Nobody Can Win: Why Weaponized Interdependence Is Cutting Everyone

I've been following this chip fight for years now, and I keep waiting for someone to just say the obvious part out loud.

For the last four years the US strategy has been pretty clean on paper: choke off China's access to advanced chips. You control the chokepoints — the Dutch machines that make the chips, the American software that designs them, the Taiwanese fabs that actually build them — and you control the future. Academics gave it a fancy name, "weaponized interdependence." Using the supply chain as a noose. And for a while, it looked like it was working.

Then RAMageddon hit.

That's what the industry is calling the memory chip shortage in 2026. AI data centers are hoovering up every stick of DRAM and HBM on the planet, because that's where the money is. And suddenly it's not just a supply chain hiccup. It's the moment the whole strategy started eating itself.

Because when you point the entire world's memory capacity at the most expensive AI chips, you don't just starve China. You starve everyone. Nine US industry groups wrote to the White House warning that it's going to raise prices for American consumers. They have a name for this too — the boomerang effect. Sanctions have costs, and the costs land on you too.

But here's the part that should actually keep people in Washington up at night. The restrictions that were supposed to kill China's chip industry are giving it a walled garden to grow in.

Huawei's HiSilicon is designing AI chips on SMIC's 7nm, working around the EUV ban that was supposed to be impossible to work around. SMIC had record revenue in 2025 and is now the third biggest foundry in the world. YMTC — Yangtze Memory — got put on the Entity List in late 2022, and now it has 13% of the global NAND market. 

Exclusion as accelerant. That's the pattern. You cut them off from the free market, you force them to build their own. And they are building it.

And it's not just chips anymore. That's the other thing everyone misses. The real fight moved.

For years the moat was Nvidia's CUDA — the software platform that locks the entire AI world into American hardware. If you wanted to do AI, you had to do it Nvidia's way. But Huawei has CANN, and Chinese labs like DeepSeek are building stacks that don't need the latest Nvidia card at all. The Koreans said it best: at a certain stage, architecture beats physics.

China hit the physical wall at sub-3nm and just went sideways. Link older chips together cleverly. Optimize the software to hell. Rearchitect the whole system. The AI race isn't about who can etch the smallest transistor anymore. It's about software plus hardware plus supply chain plus pure geopolitics all tangled up together.

And China didn't just sit there and take it. It hit back where it hurts.

Gallium, germanium, antimony. China controls 94% of gallium production, 83% of germanium. These aren't obscure rocks for geology nerds. They're in every chip, every fiber optic cable, every solar panel, every missile guidance system. Beijing banned exports to the US. Antimony shipments fell 97% in October after the earlier curbs. 

The message was pretty clear: you can choke our chips, we can choke your raw materials. And unlike chips, you can't just stockpile your way out of this. These minerals are baked into everything.

Brookings put out a piece in February arguing maybe we should loosen the controls, because superintelligence isn't around the corner and cutting US companies out of the biggest chip market in the world is self-defeating. There's logic to that. If AI has plateaued, then the strategic benefit of denying China a few extra H100s is lower than the economic cost of losing China as a customer.

But that argument assumes China will stay dependent forever. It won't.

Every restriction teaches Beijing the exact same lesson: build your own. And every year, the gap gets a little smaller.

What bothers me most is how this gets framed as binary. Either you're tough on China or you're handing them the future.

Reality is messier. Did export controls slow China down at the top end? Yes. At 3nm, 2nm, absolutely. Did they also force China to learn how to work around us, build domestic capacity, and go all-in on self-sufficiency? Also yes. The net effect in five years might be a China that is *less* dependent on American tech than if we had done nothing at all. That's not a win. That's an own goal dressed up as toughness.

Meanwhile, everyone else is caught in the middle.

The Netherlands and Japan got pressured into restricting exports to China. Their champions — ASML, Tokyo Electron — are losing revenue and market share for a fight they didn't start. Europe and Asia are being told to pick a side in a war they can't win. The global semiconductor supply chain, which was already fragile as hell, is now actively splitting into two parallel systems. One American-led, one Chinese-led.

And the countries stuck in the middle — South Korea, Taiwan, Singapore — the ones who actually make this stuff work — they're the ones who will pay the most for a decoupling they never wanted.

I don't have a clean answer. I don't think anyone does.

But weaponized interdependence is a double-edged sword, and the blade is getting dull on both sides. The US can slow China down. It can't stop it. China can retaliate with minerals and market share. It can't make the US back down.

What we're left with is a slow, grinding, expensive stalemate that reshapes the entire global economy and leaves everyone a little poorer, a little less secure, and a little more dependent on whoever happens to control the next chokepoint.

The chip war isn't a war anyone wins.

It's a war everyone loses, just at different speeds.