Lesson 8 — What's Contested? What Don't We Know?
How Do Chips Actually Work?
Learning Material
1 pagesLesson 8 — What's Contested? What Don't We Know?
Understanding the Complex: How Do Chips Actually Work?
The semiconductor industry generates strong opinions. Governments treat it as a matter of national security. Companies stake multi-billion-dollar bets on technology roadmaps that may or may not materialize. Analysts disagree profoundly about which scenarios are likely and which are wishful thinking. In this lesson, we look at four genuinely contested questions — presenting the strongest arguments on each side, without attempting to resolve them.
Controversy 1: Can Europe and the United States catch up with TSMC?
In 2021 and 2022, the United States and European Union announced major initiatives to rebuild domestic chip manufacturing capacity. The US CHIPS Act directed $52.7 billion in subsidies. TSMC, Intel, Samsung, and others announced new fabs in Arizona, Ohio, Germany, and elsewhere. The stated goal was to reduce dependence on Asian manufacturing, particularly Taiwan.
Those who believe the effort will succeed point to several factors: the scale of subsidies, the alignment of corporate and government interests, and the strategic imperative — if advanced chip production remains in Taiwan and Taiwan's security situation deteriorates, the consequences for global supply chains would be catastrophic. They also note that the US and Europe already have significant assets: the EDA software companies, semiconductor equipment makers, materials suppliers, and research institutions that underpin the global industry.
Skeptics point to a harder reality: leading-edge chip manufacturing requires decades of accumulated expertise that cannot be quickly purchased or legislated into existence. TSMC's Arizona fab has faced significant delays, cost overruns, and challenges hiring skilled workers. Europe's ambitious 20-percent market share target by 2030 would require constructing multiple leading-edge fabs in a time frame that experts consider optimistic. And while subsidies help with capital costs, they don't solve the expertise problem. History suggests that moving manufacturing to new locations is far slower and more expensive than proponents typically assume.
Both camps make legitimate points. The outcome will depend on execution, geopolitical developments, and factors that are genuinely uncertain.
Controversy 2: Is Moore's Law dead?
In the classic formulation, Moore's Law predicts that the number of transistors on a chip doubles every two years, at roughly constant cost. This prediction sustained itself for five decades. The question of whether it continues to hold — and what happens if it doesn't — divides the industry.
Pessimists note the physical constraints. Transistors at 2-3nm node are only a few atoms across in their critical dimensions. Quantum tunneling — electrons passing through barriers they classically shouldn't — becomes a significant problem below certain scales. The cost per transistor, which fell dramatically for decades, has leveled off at leading-edge nodes. A 2nm wafer costs more to produce than a 5nm wafer; whether the added transistors are worth the cost depends on the specific application.
Optimists argue that Moore's Law is less a statement about physical limits than about engineering creativity. When planar transistors hit limits, FinFETs extended the road. When FinFETs run out of road, Gate-All-Around transistors take over. When 2D scaling exhausts, 3D integration — stacking chips vertically, connecting them with dense arrays of microscopic pillars — offers a new dimension. Companies like Nvidia already ship products with multiple silicon dies stacked and interconnected (chiplets), achieving effective scaling without requiring a single die to push physical limits.
The honest answer may be that Moore's Law, as originally formulated, is slowing, but that the underlying trend of improving compute-per-dollar is continuing through architectural innovation rather than pure geometric scaling.
Controversy 3: Taiwan — what if?
This is the most sensitive question in the chip industry, involving geopolitical judgments that reasonable people hold very differently.
Taiwan produces roughly 90 percent of the world's leading-edge chips. The island's political status — claimed by the People's Republic of China as a province, governed by the Republic of China with its own democratic institutions and military — has been a source of tension for seven decades. The question of whether this arrangement is stable, and what would happen if it weren't, generates intense debate.
Those who argue that the status quo is manageable note that the economic interdependence cuts in multiple directions. China is a major customer of chips made in Taiwan; disrupting that supply would hurt China's own economy. The global community has strong incentives to keep Taiwan open and its factories running. TSMC itself has described its production as so technically complex and so geographically distributed in terms of supply chain that destroying it would serve no aggressor's interests.
Those who argue that the risks are real and growing point to changing military balances, hardening political positions on both sides of the strait, and the historical record of conflicts that most observers thought were economically irrational occurring anyway. They argue that the world's failure to maintain supply chain diversity is a genuine vulnerability.
This is not a question that has a clear correct answer. It involves values about sovereignty, security, deterrence, and the management of great-power competition that people across the political spectrum weigh differently. What is clear is that the concentration of chip production in one politically contested island is, at minimum, a risk that governments and companies are now taking very seriously.
Controversy 4: Could RISC-V displace x86 and ARM?
For most of computing history, chip architectures — the instruction sets that CPUs speak — were proprietary. Intel's x86 architecture dominated personal computers; ARM's architecture dominated mobile devices. Both extract significant value from licensing or from selling chips that implement their architectures.
RISC-V (pronounced "risk five") is an open-source instruction set architecture released by UC Berkeley in 2010. Unlike x86 or ARM, it is freely available to anyone, without license fees. Any company, research institution, or government can design and manufacture RISC-V chips.
Proponents argue that RISC-V represents a transformative shift: just as Linux made enterprise-grade software available without license fees, RISC-V could make chip design accessible to a much wider range of actors. China has been particularly enthusiastic about RISC-V, partly because it offers a path to chips that don't depend on US-controlled ARM or x86 architectures.
Skeptics note that architectural openness doesn't solve the manufacturing problem — RISC-V chips still require fabs to make them — and that the software ecosystem built around x86 and ARM represents decades of accumulated optimization that RISC-V will struggle to replicate. In the markets that matter most — high-performance servers, consumer devices — x86 and ARM have enormous incumbency advantages.
Whether RISC-V becomes a genuine alternative to the dominant architectures or remains a niche is a question that the next decade will begin to answer.
Next lesson: What Comes Next? — from 2nm to neuromorphic chips, the technological and geopolitical trajectory of the semiconductor industry.
Reading time: approx. 10–11 minutes