Why ASML and TSMC Are the Chokepoints in Global Chipmaking
Sign up for ARPU: Stay ahead of the curve on tech business trends.
The explosion in artificial intelligence capabilities, particularly since the advent of large language models, is driving unprecedented demand for advanced computing power. But this power relies on physical chips — trillions of transistors packed onto silicon wafers.
The manufacturing of these semiconductors is incredibly complex, and concentrated among a few key global players, primarily Taiwan-based TSMC and Netherland-based ASML. Their unique positions in the global supply chain mean they are increasingly becoming chokepoints, controlling the flow of the most advanced chips needed for the AI era.
What makes manufacturing advanced chips so difficult and highly concentrated?
Fabricating advanced semiconductors is one of the most complex and technologically challenging processes in the world. It involves intricate steps like photolithography, etching, doping, and metallization, repeated layer upon layer to build circuits with features measuring just a few nanometers. This requires state-of-the-art equipment that costs hundreds of millions of dollars per machine and facilities, known as fabs, that demand billions in capital investment — a single leading-edge fab can cost north of $30 billion.
Mastering these processes to achieve high manufacturing success rates, known as yields, takes decades of experience, extensive R&D, and an army of highly skilled engineers and scientists. Even slight errors or contamination can ruin an entire wafer containing hundreds or thousands of chips. This immense difficulty and cost limit the number of companies capable of operating at the leading edge, naturally concentrating manufacturing power in the hands of a few, creating the conditions for chokepoints.
Why is TSMC so dominant and essential for the AI supply chain?
Taiwan Semiconductor Manufacturing Company (TSMC) is the world’s largest and most advanced dedicated independent semiconductor foundry. This “pure-play” model means they focus solely on manufacturing chips designed by other companies, including virtually every high-end semiconductor designer like Apple, Nvidia, and AMD.
TSMC holds an immense 61% market share across the entire foundry industry globally, rising to around 67% in advanced nodes (below 7nm), and this dominance is growing. Their technological lead allows them to mass-produce the most cutting-edge chips (like 3nm and ramping up 2nm) at scale and with high yields, which competitors struggle to match.
This makes TSMC the essential go-to partner for advanced chips, with reports indicating their cutting-edge capacity is fully booked and customers are fighting for capacity at their fabs. This intense demand on TSMC’s limited cutting-edge capacity, combined with their unparalleled manufacturing capability, firmly establishes TSMC as a primary chokepoint for the supply of the most sought-after AI chips, such as Nvidia’s high-end GPUs.
What is ASML’s unique role and why is it a choke point?
ASML holds a virtual monopoly on the advanced lithography machines required to pattern the tiny features on cutting-edge chips. Their Extreme Ultraviolet (EUV) lithography machines, developed over 17 years with billions in R&D, were first shipped widely in 2017 and remain the only tools capable of printing circuitry at the densities needed for the latest nodes (like 3nm and 2nm).
These machines are the size of a double-decker bus, and cost over $100 million each, with the next-generation High-NA EUV tools costing upwards of $380 million. As the sole provider — in the world — of this essential technology, ASML is an undeniable chokepoint in the supply chain. Foundries like TSMC, Intel, and Samsung rely entirely on ASML’s machines to produce their most advanced chips. Any constraint on ASML’s ability to innovate, produce, or ship these machines — whether due to manufacturing challenges, supply chain issues, or geopolitical restrictions — directly limits the ability of the entire industry to ramp up the production of advanced chips, regardless of demand.
Who is challenging TSMC’s dominance, and how are they faring?
TSMC’s closest competitors are Samsung Foundry and Intel Foundry. Samsung has also developed advanced nodes, including 3nm using Gate-All-Around (GAA) transistor technology, which is ahead of TSMC’s adoption of GAA at 2nm. However, Samsung has struggled with lower yield rates compared to TSMC and attracting major external customers for its leading-edge processes. Recent reports indicate Samsung’s trial production yield for its 2nm process is around 30–50%, still significantly behind TSMC’s reported yields (60+%) for comparable nodes. Samsung reportedly cut its foundry capital expenditure in half in 2024 to $3.5 billion, focusing on upgrading existing facilities rather than major expansion, potentially impacting its ability to compete for future high-volume orders.
Intel is also investing heavily to re-establish its position as a leading-edge foundry, with an ambitious roadmap targeting 18A (comparable to 2nm) and 1.4nm nodes. Intel is increasing its capex to $12 billion–$14 billion in 2025 and has secured some external customers like Microsoft for its future nodes.
However, based on TSMC’s claims and industry analysis, Intel appears to be years behind TSMC in terms of process maturity, yield, and cost efficiency for advanced manufacturing, even reportedly placing billions of dollars in 3nm orders with TSMC itself for 2024 and 2025. While these efforts aim to increase global foundry options, they face significant hurdles to alleviate the current demand pressure on TSMC’s cutting edge and do not diminish ASML’s foundational chokepoint status.
How do geopolitics and the US-China rivalry impact manufacturing?
The strategic importance of advanced chip manufacturing has made it a central focus of the US-China technology rivalry. The US government has implemented stringent export controls, notably restricting China’s access to advanced chips (like certain Nvidia GPUs) and, crucially, advanced manufacturing equipment from ASML and others. These restrictions, often justified by national security concerns, prevent ASML from selling its most advanced EUV and some DUV machines to customers in China, impacting ASML’s potential revenue but also solidifying the US’s ability to control a key chokepoint.
In response, China is accelerating efforts to build domestic capabilities and achieve self-sufficiency in semiconductor manufacturing. Companies like SMIC (China’s largest foundry) and SiCarrier (reportedly developing lithography tools) are receiving state support. While China has made progress, even producing 7nm chips, they face significant technological hurdles, particularly in replicating ASML’s advanced lithography and building a complete, cutting-edge domestic supply chain, which many experts believe will take many years.
This geopolitical dynamic effectively creates two diverging supply chains, but it underscores that the US is leveraging the existing chokepoints (ASML’s equipment, TSMC’s advanced processes for US designs) to maintain a technological lead, while China attempts to bypass or replicate these chokepoints domestically. The concentration of critical R&D and manufacturing in politically sensitive locations like Taiwan also adds significant geopolitical risk to the global supply chain.
Why is global capacity expansion happening, and what are the risks?
The concentration of advanced manufacturing in Taiwan and the surging demand for AI chips have spurred governments and companies to invest heavily in building new manufacturing capacity globally, particularly in the US, Japan, and Europe. The US CHIPS Act, for example, provides billions in grants and loans to incentivize companies like TSMC and Intel to build new fabs in the United States.
However, bringing significant new capacity online is a multi-year process, requiring years for construction, equipment installation (heavily dependent on ASML’s constrained supply), and yield ramp-up.
Reports of ballooning lead times for equipment needed to build out data centers underline the broader supply chain pressures originating from manufacturing constraints, which are amplified by the ASML chokepoint. There’s also a risk of potential oversupply in certain nodes if the growth in AI demand moderates or if geopolitical dynamics shift rapidly.
Furthermore, while new fabs can increase manufacturing output, the core R&D and expertise for the next generation of processes often remain concentrated in existing hubs like Hsinchu (TSMC’s primary R&D center), Hillsboro (Intel), and Pyeongtaek (Samsung), meaning new facilities might still depend on developments elsewhere for future process nodes. This highlights that while geographical diversification is occurring, the fundamental manufacturing chokepoints centered around TSMC’s process leadership and ASML’s equipment monopoly are not easily bypassed in the near to medium term.
Reference Shelf:
Erik van Klinken
Finbarr Bermingham