Lesson 10 — What If...?
How Do Chips Actually Work?
Learning Material
1 pagesLesson 10 — What If...?
Understanding the Complex: How Do Chips Actually Work?
This lesson asks three "what if" questions — not to predict the future, but to stress-test our understanding of the present. Each scenario illuminates something important about the structure of the semiconductor industry and the dependencies that have accumulated around it.
Thought experiments of this kind are used by risk managers, policy analysts, and corporate strategists to identify vulnerabilities before they become crises. The goal isn't certainty but clarity.
What if TSMC were disrupted — severely and suddenly?
Consider a scenario in which, for any reason — natural disaster, conflict, political crisis, or even a technical catastrophe — TSMC's leading-edge fabs in Taiwan were unable to produce advanced chips for 12 months.
The immediate consequences would be severe. Apple could not ship new iPhones, since its M-series and A-series chips are exclusively manufactured by TSMC. Nvidia could not ship AI accelerators — the H100, H200, and their successors are TSMC-produced. AMD's most competitive processors would halt. Qualcomm's modem chips for 5G smartphones would stop. Cloud providers would be unable to expand AI compute capacity. Defense contractors dependent on advanced semiconductors would face shortages.
The ripple effects would cascade outward. Industries dependent on chips for final products — automotive, medical devices, telecommunications equipment — would face production constraints within weeks to months. Consumer electronics would become scarce. The software companies and AI developers who depend on cloud compute would find their expansion plans stalled.
What could be done? Samsung's fabs in South Korea offer some alternative capacity, though Samsung's leading-edge process technology is somewhat behind TSMC's. Intel's fabs in the US could absorb some volume, though again with capability limitations. The industry would manage — imperfectly, expensively — but the disruption would take years to fully absorb.
The scenario illustrates why governments and companies now treat supply chain concentration as a strategic risk rather than a cost optimization opportunity. The question isn't whether disruption is likely, but whether its probability is low enough, given its potential severity.
What if RISC-V displaced x86 and ARM as the dominant architecture?
The current chip market is shaped by two dominant instruction set architectures. x86, controlled by Intel and AMD, runs virtually all personal computers and servers. ARM, whose designs are licensed by Arm Holdings and manufactured by dozens of chip companies, runs virtually all smartphones, tablets, and an increasing fraction of servers.
Both architectures are, in an important sense, historical artifacts — x86 dates to 1978, ARM to 1985. They persist not because they are technically superior to alternatives but because of ecosystem lock-in: the billions of lines of software written for them, the decades of compiler optimization, the libraries, toolchains, and developer expertise built around their instruction sets.
RISC-V, the open-source alternative, is beginning to accumulate its own ecosystem. China has invested heavily in RISC-V as part of its effort to build a chip industry that doesn't depend on US-controlled architectures. Several hundred RISC-V chips have been designed and are in production for embedded applications, IoT devices, and research systems.
But displacing x86 or ARM in their core markets — high-performance computing and mobile devices — would require overcoming ecosystem lock-in that has resisted decades of well-funded alternatives. Apple, IBM, and Intel have all launched their own RISC-based architectures at various points and found x86 compatibility requirements impossible to escape entirely.
A more likely scenario is that RISC-V takes significant share in embedded systems, IoT, and specialized accelerators — domains where the legacy ecosystem is less entrenched — while x86 and ARM retain their positions in the markets that matter most for revenue. The long-term outcome over decades is genuinely uncertain.
What if a European chip giant emerged within a decade?
The European Union's Chips Act commits €43 billion to building European semiconductor manufacturing capacity. TSMC is building a fab in Dresden, Germany (jointly funded with the EU and Germany). Intel is building fabs in Magdeburg, Germany, and Poland. Several smaller European chip companies — Infineon, NXP, STMicroelectronics — are expanding.
Could Europe, within a decade, develop a globally competitive position in leading-edge chip manufacturing?
The honest answer is: probably not at the leading edge, but possibly in specific niches where it already has strengths.
Europe's existing chip strengths lie in automotive semiconductors, industrial chips, and power electronics — areas where Infineon, NXP, and STMicroelectronics are genuinely world-class. These markets don't require 3nm or 2nm processes; they run on established nodes where manufacturing is less concentrated.
At the leading edge — the 2nm and below process nodes where AI chips and high-performance processors live — Europe has no current capability and would need 10-15 years and hundreds of billions of euros to build it from scratch. The TSMC Dresden fab will produce 12nm process chips, several generations behind the frontier, which is commercially useful for automotive and embedded applications but not competitive for AI accelerators or high-end processors.
The scenario also highlights a general principle: government subsidies can accelerate the building of facilities, but they cannot easily accelerate the accumulation of expertise. That takes time, training, and institutional knowledge that moves slowly.
What these scenarios share
All three thought experiments point to the same underlying feature of the semiconductor industry: its extraordinary resilience and its extraordinary brittleness coexist. The system has survived — through creative engineering, deep investment, and institutional cooperation — challenges that should have derailed it many times over. And yet it depends on a small number of critical nodes — TSMC's leading-edge fabs, ASML's EUV machines, a handful of materials suppliers — each of which represents a potential failure point.
The semiconductor industry is perhaps the clearest example of what strategists call a "complex adaptive system" — resilient under ordinary stresses, potentially fragile under the right combination of shocks, and very difficult to duplicate or replace quickly.
Next lesson: What Are You Taking Away? — synthesis, cross-links, and the central insight that ties eleven lessons together.
Reading time: approx. 9–10 minutes