Key Takeaways
- The first Trump administration ban on ASML EUV lithography machines shifted Chinese industrial policy from writing passive checks to forcing a full-stack supply chain rebuild.
- SMIC serves as the domestic substitute for TSMC, but equipment bans cap the process nodes it can reach.
- High-bandwidth memory is now the main operational bottleneck, with domestic producers like CXMT unable to supply enough volume.
- Huawei now sells neural processing units under a "BYOH" model, instructing AI customers to source their own high-bandwidth memory modules elsewhere.
- Massive state-directed capital investment solves national security targets, but it often erodes corporate profits and stock returns across Chinese hardware sectors.
The EUV Ban Forced an Impossible Domestic Rebuild
Before the White House blocked advanced lithography, Chinese industrial policy relied on broad financial support. Subsidies poured into fabs, but companies still bought their tools and inputs abroad.
The US ban on ASML EUV lithography equipment changed that calculation. Kyle Chan observed, “When that switch happened, when, for example, during the first Trump administration, you got the effective ban on EUV lithography machines from ASML to China, that was a huge wake-up call because that basically capped the ability of Chinese chipmakers to make chips of a certain process node.” (play from 28:44)
Once Washington blocked tool imports, Beijing realized that buying foreign gear was no longer an option. Chan explained, “I think that US export controls had a massively galvanizing effect on China's push for semiconductor self-reliance.” (play from 27:38)
No country has ever attempted to build an entire semiconductor stack inside its own borders. Modern chips depend on Japanese photoresists, Dutch lithography optics, American electronic design automation software, and Taiwanese foundries. Chan noted, “China's almost trying to replicate the entire global semiconductor supply chain in China, which I don't think any country has ever done before.” (play from 28:14)
SMIC took the lead as China's equivalent of TSMC. But running a foundry without access to global tools creates severe production limits. As Chan pointed out, “Huawei works a lot with SMIC. SMIC is sort of like the premier foundry in China for chipmaking. They are kind of like the analog of TSMC, and they are limited by their equipment, by the lithography machines that they have, and by their own supply chains.” (play from 25:09)
The Memory Bottleneck Creates BYOH Chips
The hard wall in advanced computing is rarely raw compute alone. Modern artificial intelligence models require fast data transfer between compute cores and memory dies. That makes high-bandwidth memory (HBM) just as critical as the processor.
While SMIC wrestles with lithography limits to print Huawei's neural processing units, memory suppliers like CXMT struggle to fabricate HBM at scale. Domestic packaging cannot keep up with Huawei's chip releases.
The result is a bizarre supply chain compromise. Chan shared, “At this point, there has even been reporting that Huawei is saying to customers like, 'We can make the NPUs, the neural processing units, but it's sort of BYOH, bring your own HBM. You're going to have to source that from elsewhere.'” (play from 26:28)
This dynamic creates a sharp divide between strategic self-reliance and commercial viability. State backing pushes foundries to expand capacity regardless of yield rates or capital efficiency. Chinese manufacturers excel at scale in electric vehicles and robotics, but state-driven expansion floods markets and crushes margins. Founders watching China's push must recognize that sovereign resilience does not translate into profitable enterprise economics.
What to Do With This
Map your compute infrastructure to its geographic chokepoints this week. If you rely on AI hardware providers, demand written disclosure of their memory suppliers and packaging locations to ensure your capacity does not depend on a fragile secondary market.