In the 20th century, the nation that controlled oil controlled the world. In the 21st century, that resource is semiconductors. Every smartphone, every server, every missile guidance system, every AI model, every modern vehicle — all of it runs on chips. And the global supply chain that produces them is the most fragile strategic bottleneck in human history.
The Scale of Dependence
The global semiconductor market reached $620 billion in 2025, but its downstream economic impact is measured in trillions. Chips are embedded in $8 trillion worth of global electronics. A single advanced chip — the kind that powers AI training, autonomous vehicles, and missile systems — can cost $10,000 to $30,000 to manufacture. The machines that make them cost $200 million each. And only one company on Earth can build those machines.
This is not hyperbole. It is the defining strategic reality of our time. The semiconductor supply chain is the most complex, concentrated, and geopolitically significant industrial ecosystem ever constructed.
TSMC: The Most Important Company You May Not Know
Taiwan Semiconductor Manufacturing Company (TSMC) is, by any strategic measure, the most important company in the world. Founded in 1987 by Morris Chang, TSMC invented the foundry model — manufacturing chips designed by other companies. Today, it produces approximately 90% of the world's most advanced semiconductors (sub-7nm).
TSMC by the Numbers (2026)
| Metric | Value |
|---|---|
| Market cap | ~$900 billion |
| Global foundry market share | 58% |
| Advanced node share (sub-7nm) | ~90% |
| 2025 revenue | $95 billion |
| 2025 capex | $36 billion |
| Employees | ~76,000 |
| Key customers | Apple, NVIDIA, AMD, Qualcomm, Broadcom |
Apple's A-series and M-series chips, NVIDIA's H100 and B200 GPUs, AMD's EPYC server processors, Qualcomm's Snapdragon mobile chips — all manufactured by TSMC. Without TSMC, Apple cannot ship iPhones, NVIDIA cannot supply AI data centers, and AMD cannot compete with Intel. The entire AI revolution runs through a handful of fabrication facilities on an island 100 miles off the coast of China.
This concentration is not an accident. It is the result of decades of relentless investment, engineering excellence, and a business model that created a natural monopoly. Building a cutting-edge semiconductor fabrication facility (fab) costs $20-30 billion and takes 3-5 years. The engineering talent required to operate at TSMC's level takes a generation to develop. You cannot simply throw money at this problem — a lesson China is learning the hard way.
The ASML Monopoly: The $200 Million Machine
If TSMC is the world's most important company, ASML is the most important company most people have never heard of. The Dutch firm holds a complete global monopoly on extreme ultraviolet (EUV) lithography machines — the tools required to manufacture chips at 7nm and below.
Each EUV machine weighs 180 tons, contains 100,000 components sourced from 5,000 suppliers across 60 countries, costs over $200 million, and requires three Boeing 747s to ship. ASML sells roughly 50 of these machines per year. Every single one goes to TSMC, Samsung, or Intel. Zero go to China.
The physics behind EUV are staggering. The machine fires a laser at droplets of molten tin 50,000 times per second, generating plasma that emits extreme ultraviolet light with a wavelength of just 13.5 nanometers. This light is focused through a series of mirrors polished to sub-atomic smoothness to etch circuit patterns onto silicon wafers. The precision required is equivalent to hitting a golf ball on the moon from Earth.
This is why semiconductor export controls work. You can restrict access to chips, and countries can stockpile. You can restrict access to chip designs, and engineers can reverse-engineer. But you cannot replicate ASML's EUV technology without the entire Western industrial ecosystem that supports it. China has spent an estimated $150 billion trying. They are still 5-8 years away.
The CHIPS Act: America's $280 Billion Bet
The CHIPS and Science Act, signed in August 2022, represents the largest US industrial policy investment since the Interstate Highway System. The act provides $52.7 billion in direct subsidies for semiconductor manufacturing, $24 billion in tax credits, and over $200 billion in authorized research spending.
Major US Fab Investments (2023-2028)
| Company | Location | Investment | Process Node |
|---|---|---|---|
| TSMC | Phoenix, AZ | $65 billion (3 fabs) | 4nm / 3nm / 2nm |
| Intel | Ohio, AZ, OR | $100 billion+ | Intel 18A (1.8nm eq) |
| Samsung | Taylor, TX | $17 billion | 4nm |
| Micron | Syracuse, NY | $100 billion (20yr) | DRAM |
| GlobalFoundries | Malta, NY | $11 billion | 12nm-14nm |
The strategic logic is sound: the US designs the world's most advanced chips but manufactures only 12% of global supply, down from 37% in 1990. That gap represents an existential vulnerability. If Taiwan were blockaded, invaded, or even just subjected to prolonged military pressure that disrupted TSMC operations, the global economy would face a crisis that would make COVID supply chain disruptions look minor.
But the CHIPS Act faces real challenges. Construction costs in the US are 30-50% higher than in Taiwan. Skilled semiconductor engineers are in short supply — the industry needs 100,000 new workers by 2030. And TSMC's Arizona facility has already encountered delays, with the company citing difficulties adapting its engineering culture to US labor practices.
China's Catch-Up: $150 Billion and Counting
China has poured more money into semiconductor development than any country in history. The Big Fund (National Integrated Circuit Industry Investment Fund) has deployed three rounds of capital: Big Fund I ($21 billion, 2014), Big Fund II ($29 billion, 2019), and Big Fund III ($47 billion, 2024). Combined with provincial and private investment, total Chinese semiconductor investment exceeds $150 billion since 2014.
The results are mixed. China has made genuine progress in mature-node manufacturing (28nm and above), LED chips, power semiconductors, and chip packaging. Chinese companies now produce roughly 15% of global chip output by volume. But in advanced logic chips — the ones that power AI, smartphones, and modern military systems — China remains fundamentally dependent on foreign technology.
SMIC's achievement in producing a 7nm-class chip for Huawei's Mate 60 Pro was impressive engineering but commercially problematic. Without EUV, SMIC uses multi-patterning with older DUV equipment — a process that requires 5-10x more lithography steps, dramatically reducing throughput and yields. Industry estimates suggest SMIC's effective yield at 7nm is 20-30%, versus TSMC's 80%+ at the same node.
The Technology Gap in Context
TSMC is currently ramping 2nm production. Samsung is close behind. Intel is targeting 1.8nm equivalent with Intel 18A. China's SMIC is at 7nm with poor yields. The gap is not shrinking — it is widening. Advanced nodes are not just smaller transistors; they represent entirely new physics, materials science, and engineering methodologies that compound existing advantages.
The Supply Chain Map: A Web of Dependencies
The semiconductor supply chain is the most globalized industrial network in existence. A single advanced chip might cross 25,000 miles and touch 6 countries before reaching a consumer device:
- Design: United States (NVIDIA, AMD, Apple, Qualcomm) and UK (ARM)
- EDA Software: United States (Synopsys, Cadence, Siemens EDA)
- Equipment: Netherlands (ASML), Japan (Tokyo Electron, Screen Holdings), US (Applied Materials, Lam Research, KLA)
- Materials: Japan (silicon wafers — Shin-Etsu, SUMCO), Germany (specialty chemicals — BASF, Merck KGaA)
- Fabrication: Taiwan (TSMC), South Korea (Samsung), US (Intel, GlobalFoundries)
- Packaging: Taiwan, China, Malaysia, Vietnam
- Testing: Southeast Asia (ASE Group, Amkor)
Every node in this chain represents a potential point of failure — or a point of strategic leverage. The US and its allies control the design, software, and equipment layers. China controls significant portions of packaging and testing, plus critical raw material inputs (rare earth minerals, gallium, germanium). Neither side can function without the other in the short term, which is precisely what makes this a cold war rather than a hot one.
Market Implications: How to Invest in the Chip War
Semiconductor Investment Framework
- AI compute demand: NVIDIA (NVDA), AMD (AMD), Broadcom (AVGO) — insatiable demand for training and inference chips through at least 2028.
- Equipment monopolies: ASML (ASML), Applied Materials (AMAT), Lam Research (LRCX), KLA (KLAC) — picks and shovels plays with pricing power.
- Foundry buildout: TSMC (TSM) — dominant position strengthening, not weakening, despite diversification efforts.
- Memory supercycle: Micron (MU), SK Hynix — HBM demand driven by AI creating structural supply tightness.
- EDA duopoly: Synopsys (SNPS), Cadence (CDNS) — every chip designed anywhere uses their software. 90%+ market share.
The Bottom Line
Semiconductors are not just another industry. They are the substrate of modern civilization — the resource upon which all other strategic capabilities depend. The nation (or alliance) that controls chip production controls the 21st century, just as oil controlled the 20th.
The current strategic landscape favors the US and its allies. America controls design and software. The Netherlands controls lithography. Japan controls materials and equipment. Taiwan controls fabrication. Together, this alliance holds checkmate over China's semiconductor ambitions — for now.
But strategic advantages erode. China's $150 billion investment will eventually yield results. Mature-node flooding is already reshaping market economics. And the Taiwan vulnerability remains a ticking clock. The semiconductor war is not a conflict that will be won or lost in a single battle. It is a generational competition that will shape the balance of power for decades.
The Collective tracks semiconductor supply chains, export controls, and investment flows in real-time. Visit our Situation Room for the latest developments in the chip war.
