A subsidiary of CronBlocks  ·  Engineering Insights for Serious Practitioners

China Can Build Chips — But Why Can't It Catch TSMC?

China has demonstrated 7nm production without EUV, stockpiled ninety advanced lithography machines, and invested hundreds of billions to semiconductor self-sufficiency. Yet the gap with TSMC is not closing — it is, in fact, in some critical dimensions, widening. From the very outlook, it appears to be an engineering problem, not a political one.

August 2023 — stripdown of Huawei Mate 60 Pro sent shockwaves through the Western semiconductor and defence establishments. Inside the device was a Kirin 9000s chip — manufactured by SMIC at what appeared to be done by a 7nm processing node. At the time, SMIC was not supposed to be able to do that. Clearly, it appeared to be ahead of its time.

EUV lithography machines, widely considered as prerequisite for sub-10nm production, had been blocked from export to China since 2019. The US intelligence community had apparently missed their mark. Thus, policy circles scrambled. The headlines flashed, declaring that China had made a breakthrough.

Yes, they had.

But the nature of that breakthrough is almost universally misunderstood — and the serious misunderstanding matters for anyone trying to assess where China actually stands, what it can realistically achieve, and what its gap with TSMC actually represents in engineering terms.

Hence, this article attempts to provide that grounded assessment. Given that the semiconductor gap between China and Taiwan is not merely a story about politics and sanctions. It represents a story about expertise, of process engineering, accumulated institutional knowledge, and the sheer compounding nature of manufacturing excellence achieved over time.

some numbers
30×
TSMC's monthly wafer volume lead over SMIC at equivalent nodes (2024–25)
33%
SMIC's estimated yield at 5nm vs TSMC's 80%+ at equivalent node
50%
Higher cost per wafer at SMIC 5nm vs TSMC's EUV-based equivalent
34 steps
DUV passes needed for 7nm vs 9 passes with EUV — each adds defect risks
2nm
TSMC's production capability (in 2026) — China cannot manufacture that
10+ years
Taiwan's estimated lead over China at the cutting edge, per Taiwan's National Science Council

What China Actually Did — And What It Cost

The Kirin 9000s chip was real. SMIC's 7nm capability was real. But the method behind — and its implications for scale — is what most of the coverage entirely missed.

To manufacture chips at 7nm, TSMC uses EUV lithography: a process that uses light having 13.5nm wavelength, produced by firing laser pulses at 50,000 tin droplets per second, that in turn produces plasma, and that emits the radiation precise enough to etch features smaller than the size of a virus. A single EUV exposure can draw pattern, printing critical layer of a chip in one pass. In this regard, ASML holds complete monopoly over EUV tools. And China cannot legally obtain them.

SMIC's engineers resorted to solving this differently.

They used DUV immersion lithography — an older generation of the process that operates at 193nm. They combined it with a technique called self-aligned quadruple patterning (SAQP). Where EUV prints a critical layer in one pass, SAQP requires four exposures, with extreme precision and alignment at each step.

A modern 7nm logic chip requires more than 80 lithography steps in total.

With EUV, roughly nine of the most critical steps are handled cleanly in single passes. But, with DUV multi-patterning, those same steps require 34 separate exposures. Each additional exposure introduces alignment error. Each alignment error increases defect density. And defect density directly determines yield — the percentage of chips on a wafer that actually work.

"SMIC's 5nm wafers estimatedly cost 50% more than TSMC's equivalent in production, that too at yields reportedly as low as 33% — against TSMC's 80%+ in even comparison. And this, is not a minor efficiency gap. It represents structural manufacturing disadvantage that has potential to compound over every generation."

Resultantly, SMIC is producing chips at approximately 3,000–4,000 wafer starts per month at its 7nm-class process — sufficient for supplying Huawei's premium smartphones and AI accelerator business. But that's nowhere near being competitive at a commercial scale. On the other hand, TSMC produces roughly 100,000–120,000 wafer starts per month for Apple's A-series chips at 3nm. The volume gap is approximately 30 times.

For a specific customer, within a controlled supply-chain environment, SMIC's process works fine. But as a foundry business, competing in the open market for high-volume customers, the cost and yield disadvantage is strongly prohibitive.

TSMC — 2026 position
Leading node2nm (mass production)
LithographyEUV (single-pass)
Yield at 3nm80%+
Monthly volume (3nm)~110,000 wafer starts
Advanced node share>90% of sub-3nm globally
2025 capex$40–42B
SMIC — 2026 position
Leading node~5nm
LithographyDUV + multi-patterning
Yield at 5nm~33%
Monthly volume (7nm)~3,500 wafer starts
Advanced node shareNegligible, in open market
Cost premium against TSMC~50% per wafer

The EUV Barrier — And Potential Loophole Around It

By convention, the semiconductor containment strategy is straightforward: deny Chinese access to EUV, deny them sub-10nm process, and maintain the technology gap indefinitely. This understanding is correct in principle, but dangerously incomplete in practice.

In 2024 alone, Chinese entities acquired approximately 90 of ASML's most advanced DUV immersion machines — the NXT:1980Fi series — representing 70% of ASML's global DUVi sales that year, worth an estimated $5–7 billion. These machines lie in a regulatory grey zone. They are not EUV. Thus, not technically covered by most export control thresholds. But... they are capable ones. Through multi-patterning, they can produce near-frontier chips at scale.

An April 2026 report by the American Enterprise Institute identified this as a structural loophole within the US semiconductor policy: the controls are marked at specific technical thresholds that make the NXT:1980Fi drop just outside the line.

This indicates that China's DUV acquisition is not incidental. It is deliberate. And the Chinese fabs appear to be stockpiling machine capacity before further restrictions close that window.

In turn, the Chinese strategy looks like building the manufacturing infrastructure at scale, and run multi-patterning processes at volumes to make the yield disadvantage commercially manageable through sheer throughput only.

The servicing dependency — an underappreciated vulnerability
ASML's DUV machines require specialist maintenance approximately every six months, performed only by ASML and its authorised service partners. Over 200 DUVi tools are currently installed and operational in China. According to multiple assessments, without access to this ongoing maintenance, China's DUVi manufacturing-base can significantly degrade within a year — meaningfully disrupting SMIC's advanced manufacturing capability without a single new export restriction. This servicing dependency is arguably one of the most effective, yet underused lever in the Western containment toolkit.

Why This Is an Engineering Problem?

The framing of semiconductor competition in the form of sanctions and export controls misses the deeper reality: TSMC's advantage is not merely legal or geopolitical. It is also about the accumulated result of three decades of process engineering at an intensity and scale that is genuinely difficult to replicate.

Consider what ASML's EUV machines actually require to function.

The light source generates extreme ultraviolet radiation by firing lasers at 50,000 tin droplets per second. The optical system uses precise multilayer mirrors — manufactured by Zeiss (Germany) — with surface roughness measured in picometres. For idea: a single human hair contains millions of times more material than the tolerance margin in these mirrors. Hence, the entire machine operates as a system in which every component performs simultaneously at the cutting-edge of physical possibility. ASML took 30 years and approximately $8 billion in research and development to reach the current EUV capability. The knowledge embedded within that machine is not lying in any blueprint that China could just acquire.

The same principle applies to TSMC's process knowledge.

Semiconductor manufacturing is a discipline of yield improvement as the primary measure of engineering progress. Maturing a process from 40% yield to 80% at a given node requires thousands of small, accumulated insights about contamination sources, thermal management, material interactions, and equipment calibration. And this knowledge lives in the people, the documentation, the institutional memory, and the continuous experimentation cycles of a world-class manufacturing organisation. It cannot be just purchased and replicated. It cannot be forced into existence. And it does not transfer easily without building the founding substrate of understanding.

SMIC is capable and improving. Its engineers solved a genuinely hard problem in manufacturing 7nm without EUV. But the gap between "acquiring a node via an alternative route at low yield" and "manufacturing at commercial scale with competitive yield" is measured not in years but in engineering generations. And each generation requires the previous one to have fully understood the arena.

the trajectory
2019
US secured Dutch government agreement to block ASML EUV exports to China. Widely regarded as the defining early move in semiconductor containment.
2023
Huawei Mate 60 Pro teardown revealed Kirin 9000s built on SMIC's 7nm-class process via DUV + multi-patterning. Western policy caught off guard.
2024
China acquired 90 ASML DUVi machines in a single year — 70% of global sales. Yield on Ascend 910C AI accelerators climbed from 20% to 40%. TSMC began 3nm mass production.
2025
SMIC confirmed 5nm-class production via DUV multi-patterning. Estimated yield: 33%. Cost premium over TSMC's equivalent: ~50%. TSMC announced 2nm on track for mass production.
2026
Current state: TSMC in 2nm mass production. SMIC targeting 1.6M advanced dies for Huawei AI accelerators. AEI identifies DUVi loophole. MATCH Act introduced in US Senate to restrict DUV exports. Domestic Chinese DUV tools — still at 28nm capability commercially.
2030
Earliest credible estimate for Chinese domestic lithography tools to reach sub-10nm commercial viability. TSMC will be operating at nodes that do not yet have public names.

Where China Is Actually Winning?

A complete analysis requires acknowledging where China's semiconductor trajectory is genuinely formidable — and where Western complacency is at a greater risk.

In memory chips, the picture is substantially different from common logic. YMTC has captured approximately 18% of the global NAND market. CXMT is closing the gap in DRAM. Given the scenario, these are not trivial achievements. Memory chip production is less lithography-constrained than logic chips. Which means that the EUV ban is a smaller impediment. Hence, China's memory players are building real capability and capturing real market share at a rate that is faster than most Western industry analysis has acknowledged.

In chip design, Huawei's HiSilicon division has demonstrated that Chinese design capability — before being cut off from TSMC manufacturing and ARM licences — is genuinely world-class. The question of what Huawei's designers can produce when manufacturing constraints are removed is not trivial. Their design talent exists. Their manufacturing capability remains the limiting constraint.

In mature nodes — 28nm and above — China is no longer catching up. It has already arrived. Chinese foundry capacity of 28nm and legacy nodes is already sufficient to dominate the global market for automotive industry, industrial controllers, communications equipment, and a wide range of defence applications. This is strategically far more important than the most coverage suggests. Chips that control weapons, navigation equipment, aircraft subsystems, and military communications are not all manufactured at 3nm. They are mostly manufactured at nodes where China now has established and growing capability.

The defence electronics dimension — what debates miss
The public discourse on the semiconductor gap focuses almost entirely on AI accelerators and consumer electronics. The more strategically important and consequential question is regarding defence electronics. That's where chips produced at 14-28nm are overwhelmingly dominant. In this domain, China's manufacturing capability is not merely adequate. It is becoming increasingly self-sufficient. Avionics systems, missile guidance systems, and communications electronics of the next generation of Chinese military hardware can be built on nodes that China can manufacture domestically. This is the dimension that deserves far more critical attention than it currently receives.

What This Means for Nations, Institutions, and Engineers

The gap between China and TSMC is real, substantial, and in the domain of leading-edge logic, is likely to persist for a decade or more. Taiwan's National Science Council estimates the gap at over ten years given the TSMC's 2nm production capability. The ASML CEO cited ten to fifteen years for the same. These are not propaganda figures — they reflect the compounding nature of process engineering advantage.

But the strategic implications depend critically on which domain one is analysing. For AI training, TSMC's advantage is decisive and SMIC cannot compete at scale. For defence electronics, mature nodes, and memory, China's position is substantially stronger and improving faster than the headlines like "China can't catch TSMC" suggest.

For those outside the US-China-Taiwan triangle, the race presents a clarifying question: which supply chains are you actually dependent on, and what happens to your industrial and defence capabilities if any supply chain is disrupted? Those, building their electronics ecosystems on TSMC-fabbed components face a different risk profile from those who — by choice or necessity — rely on Chinese-manufactured chips at mature nodes. Neither position is inherently stable.

"The chips that power weapons systems, aircraft avionics, and military communications are not all manufactured at 3nm. They are built at processes where China has already arrived. This is the dimension of the semiconductor race that deserves far more attention than it currently receives."

The Conclusion

China cannot catch TSMC at the cutting edge in the foreseeable future. It's a defensible statement. But it is also, at the same time, strategically misleading.

China does not need to catch TSMC to achieve its prime objectives. It needs to manufacture enough advanced chips to power its AI ecosystem, enough mature-nodes to equip its military, and enough domestic capability to reduce its vulnerability to supply chain disruptions. And on those objectives — not on competing with TSMC for Apple's business — China is making faster progress than the dominant narrative suggests.

Yes, the gap is real.

Yet the gap is being actively worked around — funded at a scale that simply dwarfs any comparable national industrial programme in history, and staffed by engineers who have demonstrated that they can solve problems that the rest of the world had assumed unsolvable.

Underestimating that is a strategic mistake. So is overestimating.

Comments

Popular posts from this blog

Technology: The New Colonialism

Empires no longer arrive with armies. Yet, they do come. In modern days, with software, semiconductors, cloud platforms, and export controls. Implying, colonialism didn't disappear — it evolved into a much more sophisticated, invisible, and arguably more durable form of possession and control. — 12 min read Alexander the Great built one of history's largest empires in barely a decade. The Mongols forged the most expansive contiguous land empire the world had ever seen. The British Empire ruled so much of the planet that the sun supposedly never set on it. Separated by centuries, all such powers shared something in common: control something critical that others depend on, and the power shall follow. For most of history, that formula to control translated into professional armies, naval fleets, seaports, trade routes, and vast occupied territories under direct control. In-short, physical domination of the physical world persi...

The Most Misunderstood Keywords in Embedded C/C++

Six keywords appearing in almost every embedded project, cited commonly in code reviews, and understood correctly by almost a negligible number of new developers. And this isn't a report from a random academic survey — it is a field report from development teams of systems where getting these wrong costs weeks and months. — 18 min read Particular kind of bug keeps haunting embedded systems: it is the kind where the code is correct, logic looks sound, unit tests pass, but the system still fails in hardware. You revisit the algorithm. Verify the peripherals. Add printf s for debugging — which, in turn, changes the timing enough for the fault to disappear. Then you remove it. And the problem returns. In a significant proportion of such cases, the root cause can be traced to misunderstood keywords. It is not about a missing keyword — it is about a keyword that is present, used with confidence, but doing something entirely different from what the developer beli...