The Trump administration recently approved Nvidia's H200 GPU for export to China — a chip one tier below the restricted H100 — while simultaneously watching Beijing instruct state-linked buyers to avoid U.S. devices "unless absolutely necessary," according to CSIS. That's the export control paradox in one sentence: the U.S. loosens restrictions on hardware China is already moving to replace.
The pattern here isn't new, but the data is getting harder to dismiss.
China's Equipment Vendors Are Growing Into the Gap
The clearest measurable signal right now is in semiconductor equipment market share. Chinese vendors — NAURA, AMEC, ACM Research, Hwatsing, Piotech, Kingsemi — collectively held 6.5% of the $41.4 billion global wafer fabrication equipment market in 2025, up from 5.6% in 2024 and 1.2% in 2021. Revenue growth for that group averaged 30.5% year-over-year in 2025, against 9.3% for the top non-Chinese incumbents.
To be clear about what 6.5% global share means: ASML, Applied Materials, Lam Research, and Tokyo Electron still dominate the market, particularly at leading-edge nodes. Chinese vendors aren't threatening EUV lithography or advanced etch at 3nm. But that's not the relevant question. The relevant question is whether sanctions are slowing China's domestic capability development — and the trajectory of those numbers suggests the answer is increasingly no, at least for mature nodes.
AMEC is improving plasma etch tools at more advanced nodes. Hwatsing is gaining traction in CMP, a critical planarization step. Piotech is moving into advanced packaging deposition. These aren't moonshot claims; they're incremental process steps where domestic substitution is technically achievable without EUV. The U.S. export control regime was designed around leading-edge chokepoints. China is building around the perimeter.
TSMC Arizona: One More Timeline to Track
Reuters reported in late April that TSMC plans to open an advanced packaging facility in Arizona by 2029, with construction already underway. The company had signaled intent to build advanced packaging capacity at its Arizona campus during its January earnings call but hadn't committed to a specific timeline. Now there's a date.
Advanced packaging — the process of combining multiple chiplets into a single high-performance package — is the current supply bottleneck for AI chips. Nvidia's most capable products depend on it. TSMC doing this work in Arizona rather than Taiwan matters for supply chain resilience arguments, and it's a real milestone worth tracking.
But 2029 is three years out. TSMC's Arizona fab history is instructive: Fab 1 broke ground in 2021, faced construction delays, workforce challenges, and equipment installation setbacks, and began limited N4 production years behind original projections. The packaging facility announcement is a directional commitment, not a delivery. Watch for construction progress photos, equipment procurement signals, and whether TSMC's quarterly calls start attaching capacity targets to that 2029 date.
The Supply Chain Scramble Has a $1 Trillion Backdrop
Bloomberg noted that the chip industry is on course to become a $1 trillion business in the coming years, with governments and companies actively rethinking where production is located. That framing is accurate but worth interrogating: "rethinking" and "remaking" are different things. Announcements, funding commitments, and groundbreakings are rethinking. Qualified fabs producing volume at competitive yield rates is remaking.
The gap between those two states is where most of the current policy action lives — and where most of the accountability failures will eventually surface.
What the Next 90 Days Should Clarify
Three things worth watching before the summer:
First, whether Beijing's reported push to boycott U.S. chips translates into measurable procurement shifts at major Chinese cloud providers. State encouragement and actual purchasing decisions are different signals.
Second, whether TSMC's Arizona packaging facility generates any equipment procurement filings or construction permits that would let outside observers verify the 2029 timeline is on track rather than aspirational.
Third, how the Commerce Department responds to the growing evidence that China's domestic equipment vendors are gaining share at mature nodes. The current export control framework was architected around leading-edge chokepoints. If the perimeter strategy is working, BIS hasn't shown its work publicly — and given the enforcement backlog I covered three weeks ago, the burden of proof is on the policy, not the skeptics.
The chip industry runs on optimistic timelines. Industrial policy runs on optimistic assumptions. When both are operating simultaneously, the gap between PowerPoint and production tends to be wide. The numbers coming out of China's equipment sector suggest that gap is narrowing — on their side of the ledger.