Critical Industries
The apex companies and chokepoints national security and frontier-technology production actually depend on.
The story most citizens hear about industrial power is the story of companies they can name — the consumer brands and the household-name technology firms. The story of where national security and frontier-technology production actually rest is different. It rests on companies most people have never heard of: ASML in the Netherlands making the lithography machines without which no advanced chip exists; Carl Zeiss SMT in Germany making the optics inside those machines; TSMC in Taiwan operating fabs that produce ninety percent of leading-edge logic. It rests on the chokepoints in critical minerals processing that China controls; on the gene-editing companies whose IP descends from a few academic labs; on a handful of defense primes the state buys almost everything from; on the cloud providers and the subsea-cable consortiums that carry every bit of information across every ocean.
This Atlas documents those chokepoints company by company. The point is not to alarm. The point is operating literacy: a citizen who can name the apex companies in eleven critical sectors, identify the choke at each, and find the canonical scholarship has more genuine grasp of industrial power than ninety-nine percent of the people who talk about industrial policy on cable news.
Semiconductor Lithography
Why it matters
No lithography, no leading-edge chips. No leading-edge chips, no modern AI compute, no advanced weapons guidance, no 5G/6G basebands, no modern automotive ECUs. This is the single most concentrated chokepoint in modern industrial civilization.
The chokepoint
EUV (extreme-ultraviolet) lithography is required for sub-7nm process nodes. The only producer of EUV machines is ASML (Netherlands). Each machine costs ~$300M (EUV) or ~$400M (High-NA EUV). Roughly 200 EUV machines exist worldwide; production is constrained to fewer than 60/year.
Apex companies
| Company | Country | Role |
|---|---|---|
ASML A ~€30B revenue company that effectively gates the entire frontier semiconductor industry. Holds export licenses to China that are tightly controlled by Dutch and US governments. | Netherlands | Sole producer of EUV lithography systems |
Tokyo Electron (TEL) | Japan | Deposition, etch, cleaning systems |
Applied Materials | United States | Largest semiconductor-equipment maker by revenue |
Lam Research | United States | Etch and deposition systems |
KLA Corporation | United States | Process control and inspection |
Carl Zeiss SMT | Germany | Optical systems inside ASML EUV machines — itself a sub-chokepoint |
Scholarship: Chris Miller, "Chip War" (2022) is the canonical popular treatment. SemiAnalysis (Dylan Patel) is the leading independent technical analyst.
Semiconductor Fabrication
Why it matters
A 3nm or 5nm chip is the product of a ~1,000-step process executed inside a fab costing $15–$25B to construct. Building one takes ~3–5 years. The world has ~3 companies that can produce leading-edge logic, and they are concentrated in two countries.
The chokepoint
TSMC alone produces ~90% of leading-edge (3nm, 5nm, 7nm) chips, and ~100% of the chips used by Apple, NVIDIA, AMD, and Qualcomm at the leading edge. TSMC operates principally in Taiwan, ~110 miles from mainland China. This is the single most-cited national-security concentration in US defense planning.
Apex companies
| Company | Country | Role |
|---|---|---|
TSMC (Taiwan Semiconductor Manufacturing Co.) TSMC Arizona (Fab 21) is the largest greenfield foreign direct investment in US history at ~$65B; reaches volume production 2025–2027. | Taiwan | Leading-edge logic foundry — 90%+ market share at the frontier |
Samsung Foundry | South Korea | Runner-up at leading-edge logic; leader in memory |
Intel Foundry | United States | Recovering competitive position with 18A node 2025 |
GlobalFoundries | United States / UAE | Trailing-edge logic; secured-fab US-defense focus |
SMIC | China | Largest Chinese foundry; making catch-up progress despite US sanctions |
SK hynix | South Korea | Memory and HBM (High Bandwidth Memory) for AI accelerators |
Micron Technology | United States | Memory and HBM |
Scholarship: TSMC’s annual report and SemiAnalysis’s fab-level coverage are the public sources of record.
AI Compute and Foundation Models
Why it matters
Frontier AI training requires tens of thousands of accelerators networked together. The accelerator supplier, the networking-fabric supplier, and the cloud landlord effectively gate which organizations can train frontier models — a small list.
The chokepoint
NVIDIA holds >85% of the discrete AI-accelerator market with H100/H200/B100/B200/GB200 product lines. Its CUDA software ecosystem is a deeper moat than the silicon itself. NVLink/NVSwitch interconnect is similarly concentrated. The three major cloud providers (AWS, Azure, GCP) control the at-scale rental market.
Apex companies
| Company | Country | Role |
|---|---|---|
NVIDIA ~$3T market cap as of 2024 — briefly the most valuable company on Earth. Sells primarily to hyperscalers (Microsoft, Meta, Google, Amazon, Oracle). | United States | AI accelerators (GPUs) and the CUDA ecosystem |
AMD | United States | MI300X and successors; meaningful but distant runner-up |
Broadcom | United States | Custom AI silicon for hyperscalers (Google TPU, Meta MTIA partnership) |
Microsoft (OpenAI partnership) | United States | Leading consumer AI deployment (ChatGPT, Copilot) |
Anthropic | United States | Frontier model lab backed by Amazon and Google |
Google DeepMind | United States / United Kingdom | Frontier research with deep custom-silicon integration (TPU) |
Meta AI (FAIR) | United States | Open-weight model leader (Llama family) |
xAI | United States | Frontier lab with rapidly scaled training cluster |
Mistral AI | France | Leading European frontier-model lab |
DeepSeek, Alibaba (Qwen), Baidu, Moonshot, Zhipu | China | Leading Chinese frontier labs operating under US chip-export restrictions |
Biotechnology, CRISPR, and Genetic Medicine
Why it matters
CRISPR-based genome editing is dual-use: it can cure sickle-cell disease (Casgevy, FDA-approved 2023) and it could in principle enable agricultural sovereignty, lineage tracing, and — at the dark end of dual-use — engineered pathogens. Biotechnology is now in roughly the position computing occupied in the 1970s. The next decade will decide who sets the rules.
The chokepoint
The IP foundations of CRISPR-Cas9 are split between the Broad Institute and UC Berkeley (the Doudna/Charpentier patents); the practical chokepoints are clinical trial throughput, capital, and GMP manufacturing of viral and lipid-nanoparticle delivery vehicles.
Apex companies
| Company | Country | Role |
|---|---|---|
Vertex Pharmaceuticals | United States | First FDA-approved CRISPR therapy (Casgevy, sickle-cell) — partnered with CRISPR Therapeutics |
CRISPR Therapeutics | Switzerland | Doudna-side IP licensee; therapeutic developer |
Editas Medicine | United States | Broad-side IP licensee; in vivo and ex vivo programs |
Intellia Therapeutics | United States | In vivo CRISPR delivery — transthyretin amyloidosis program (clinical proof of in-body editing) |
Beam Therapeutics | United States | Base editing (single-nucleotide changes without double-strand breaks) |
Prime Medicine | United States | Prime editing (David Liu lab) — most flexible editing platform to date |
Moderna, BioNTech, Pfizer | United States / Germany | mRNA platforms — same delivery infrastructure as future in vivo gene therapies |
Broad Institute, Berkeley IGI, Wellcome Sanger | United States / United Kingdom | Academic apex — where the foundational science continues to come from |
Scholarship: Walter Isaacson, "The Code Breaker" (2021) covers the Doudna story. Kevin Davies, "Editing Humanity" (2020) is the rigorous reference.
Nanotechnology and Advanced Materials
Why it matters
Nanotechnology is not a single industry; it is the substrate of modern semiconductors, drug delivery, energy storage, structural composites, sensors, and quantum devices. The leadership question is "who owns the techniques and the IP."
The chokepoint
No single chokepoint; rather, a constellation of specialist tool, material, and process companies, with cross-cutting research institutions (IBM Research, Argonne, MIT.nano, IMEC, RIKEN) sitting upstream of nearly everything.
Apex companies
| Company | Country | Role |
|---|---|---|
Applied Materials, Lam Research, Tokyo Electron, ASM International | US / Japan / Netherlands | Atomic-layer deposition, etch, and the entire toolchain that builds nanoscale features |
BASF, Merck KGaA | Germany | Specialty chemistry and lithography photoresists |
JSR Corporation, Shin-Etsu, Sumitomo Chemical, Tokyo Ohka Kogyo | Japan | Photoresists and ultra-pure silicon wafers — Japan controls roughly 50%+ of these markets |
Nanosys (Shoei merger), Quantum Materials | United States | Quantum-dot displays and lighting |
Carbon Inc., Markforged, Desktop Metal | United States | Advanced additive manufacturing |
IBM Research, Argonne, MIT.nano, IMEC (Belgium), RIKEN (Japan) | Various | Public-private research institutions that produce most of the precursor IP |
Defense Prime Contractors and Aerospace
Why it matters
The visible apex of military-industrial expenditure. Defense primes integrate thousands of subcontractors into the platforms governments actually procure — fighters, missiles, ships, satellites, ICBMs, command-and-control systems.
The chokepoint
Concentration is extreme. In the United States, post-Cold-War consolidation reduced the number of major prime contractors from ~50 to ~5. A small number of platforms (F-35, B-21, Columbia-class submarine, Sentinel ICBM, Aegis) drive a disproportionate share of procurement.
Apex companies
| Company | Country | Role |
|---|---|---|
Lockheed Martin Largest defense contractor by revenue (~$70B). F-35 program is the largest weapons program in history (~$1.7T lifecycle). | United States | F-35, Aegis, ICBMs (Sentinel), missiles, Skunk Works |
RTX (Raytheon Technologies) | United States | Missiles (Patriot, Tomahawk, Stinger, SM-3/6), radars, Pratt & Whitney engines, Collins Aerospace |
Northrop Grumman | United States | B-21 Raider, Sentinel ICBM, satellites, ground systems |
General Dynamics | United States | Submarines (Columbia, Virginia), Abrams tanks, IT services (GDIT) |
Boeing Defense, Space & Security | United States | F-15EX, KC-46, Apache, satellites |
BAE Systems | United Kingdom | Largest European defense prime; Eurofighter, naval, electronics |
Airbus Defence & Space | France / Germany / Spain / UK | A400M, FCAS partnership, satellites |
Thales | France | Defense electronics, radars, optronics, cybersecurity |
Leonardo | Italy | Helicopters, defense electronics, Eurofighter participant |
Rheinmetall | Germany | Land systems and ammunition — major beneficiary of post-2022 European rearmament |
AVIC, NORINCO, CASC, CASIC | China | Chinese state defense conglomerates |
Rostec | Russia | Russian state defense conglomerate |
Scholarship: SIPRI Top-100 arms producers (Stockholm International Peace Research Institute) is the canonical annual dataset.
Space and Launch
Why it matters
Whoever can put mass into orbit cheaply controls the satellite layer — communications, intelligence, navigation (GPS-equivalents), and increasingly orbital persistence. Reusable launch has collapsed the cost per kilogram by roughly 20× in a decade.
The chokepoint
SpaceX accounts for roughly 80% of mass-to-orbit worldwide. Its Starlink constellation now provides the dominant low-Earth-orbit broadband service. The Falcon 9 reuse cadence makes it structurally difficult for any competitor to close the cost gap without a comparable reusable platform.
Apex companies
| Company | Country | Role |
|---|---|---|
SpaceX | United States | Falcon 9 / Falcon Heavy / Starship; Starlink constellation |
United Launch Alliance (ULA) | United States | Vulcan Centaur — national-security launch alternative to SpaceX |
Blue Origin | United States | New Glenn (operational 2025), Kuiper constellation partner |
Rocket Lab | United States / New Zealand | Electron and Neutron launch; spacecraft components |
Arianespace / ESA | European Union | Ariane 6, Vega C |
CASC (China Aerospace Science and Technology Corp.) | China | Long March family; Tianzhou cargo; Tiangong space station |
CASIC | China | Solid-fueled launch and missile-derived systems |
Roscosmos | Russia | Soyuz, Angara |
ISRO | India | PSLV, GSLV, Chandrayaan-3 (2023 lunar success) |
Energy Storage and Batteries
Why it matters
Battery cost has fallen ~90% since 2010 and is the linchpin of electric mobility, grid storage, and increasingly defense systems (drones, autonomous platforms). Battery industrial policy is now a primary axis of US-EU-China competition.
The chokepoint
Chinese companies hold roughly 75% of global EV battery production. Cathode active material and battery-grade processed lithium, nickel, and cobalt are even more concentrated upstream.
Apex companies
| Company | Country | Role |
|---|---|---|
CATL (Contemporary Amperex Technology Co.) | China | ~38% of global EV battery market in 2024 |
BYD | China | Vertically integrated automaker and battery maker; LFP leadership |
LG Energy Solution | South Korea | Largest non-Chinese battery maker |
Samsung SDI | South Korea | Premium battery cells |
SK On | South Korea | Major US Ford joint-venture investments |
Panasonic Energy | Japan | Long-time Tesla partner; Nevada gigafactory |
Tesla | United States | Vertically integrated cell + pack production (4680 cells) |
Northvolt | Sweden | European-sovereign battery champion (operational and financial challenges 2024–2025) |
QuantumScape, Solid Power, SES AI | United States | Solid-state battery development |
Critical Minerals and Refining
Why it matters
Modern technology depends on roughly 50 elements that the US Geological Survey designates "critical" — including the rare earths, lithium, cobalt, nickel, gallium, germanium, neon, tungsten, helium, and graphite. The constraint is rarely the geology; it is the refining and processing capacity, which is heavily concentrated in China.
The chokepoint
China processes roughly 85% of global rare earth elements, 70% of cobalt, 60% of lithium, and 100% of natural graphite anode material. Gallium and germanium were placed under Chinese export controls in 2023 — the first overt weaponization of materials supply against the United States and allies.
Apex companies
| Company | Country | Role |
|---|---|---|
China Northern Rare Earth, Shenghe Resources, Lynas (partial Chinese ownership) | China / Australia | Rare earth mining and separation |
MP Materials | United States | Operates Mountain Pass mine in California — the largest non-Chinese rare-earth source |
Albemarle, SQM, Ganfeng Lithium, Tianqi Lithium | US / Chile / China | Lithium extraction and processing |
Glencore | Switzerland / UK | Cobalt and copper — major DRC operations |
China Molybdenum (CMOC) | China | Cobalt — operates two of the largest DRC mines |
Vale, BHP, Anglo American, Rio Tinto | Brazil / Australia / UK | Iron, nickel, copper at scale |
Air Liquide, Linde, Air Products, Iwatani | France / US / Japan | Industrial gases including neon (critical for chip fabs) |
Scholarship: USGS Mineral Commodity Summaries is the canonical annual reference. IEA "Critical Minerals Market Review" (annual) covers demand-side trends.
Cloud Infrastructure, Networking, and Subsea Cables
Why it matters
The cloud is where the computing happens; the networking equipment carries the bits; the subsea cables carry the bits across oceans. Whoever controls all three controls the infrastructure of digital sovereignty.
The chokepoint
Cloud is dominated by three companies. Networking equipment is split between US incumbents (Cisco, Juniper, Arista) and Chinese vendors (Huawei, ZTE) — with the latter blocked in most Western markets. Subsea cables are owned and maintained by a tiny number of consortiums; physical landing stations are obvious chokepoints.
Apex companies
| Company | Country | Role |
|---|---|---|
AWS (Amazon Web Services) | United States | ~31% global cloud share |
Microsoft Azure | United States | ~25% global cloud share |
Google Cloud | United States | ~11% global cloud share |
Alibaba Cloud | China | Dominant in China; ~4% global |
Oracle, IBM, Tencent Cloud | US / China | Specialty and second-tier cloud |
Cisco, Arista Networks, Juniper, Nokia, Ericsson | US / Finland / Sweden | Networking equipment |
Huawei, ZTE | China | Telecom equipment — restricted in US/UK/AU/JP/CA/NZ |
SubCom (formerly TE SubCom) | United States | Subsea cable systems |
Alcatel Submarine Networks (Nokia) | France | Subsea cable systems |
NEC Submarine Cable | Japan | Subsea cable systems |
HMN Tech (formerly Huawei Marine) | China | Subsea cable systems |
Quantum Computing and Post-Quantum Cryptography
Why it matters
A sufficiently large quantum computer would break the public-key cryptography (RSA, ECDSA) that protects roughly every authenticated transaction on the internet — and the encrypted communications adversaries have been intercepting and storing for years ("harvest now, decrypt later"). The race to deploy post-quantum cryptography (NIST PQC standards finalized 2024) is national-security urgent independent of any specific quantum-computing timeline.
The chokepoint
No clear chokepoint yet on the hardware side — multiple architectures (superconducting, trapped-ion, neutral atom, photonic) remain technically viable. On the software side, NIST’s PQC standardization (CRYSTALS-Kyber, CRYSTALS-Dilithium, FALCON, SPHINCS+) is the inflection point.
Apex companies
| Company | Country | Role |
|---|---|---|
IBM Quantum | United States | Superconducting; largest quantum cloud user base |
Google Quantum AI | United States | Superconducting; Sycamore quantum-advantage claim |
IonQ, Quantinuum | United States | Trapped-ion architecture |
PsiQuantum | United States / Australia | Photonic quantum computing, building utility-scale fab |
Rigetti, D-Wave | United States / Canada | Superconducting and quantum-annealing platforms |
Atom Computing, QuEra | United States | Neutral-atom quantum computing |
Origin Quantum, USTC (University of Science and Technology of China) | China | Chinese government-backed quantum program |
Cybersecurity and Identity
Why it matters
Every layer of the critical-infrastructure stack — power grid, financial settlement, healthcare records, weapons platforms, government services — depends on cybersecurity products and the identity systems that authenticate access to them. Both are dominated by a small number of US and Israeli companies.
The chokepoint
Endpoint detection and response (EDR) is dominated by ~5 companies. Hardware security modules (HSMs) are dominated by ~3. The DNS root and the certificate authority system are similarly concentrated.
Apex companies
| Company | Country | Role |
|---|---|---|
CrowdStrike | United States | Endpoint detection and response |
Palo Alto Networks | United States | Network and cloud security |
Fortinet | United States | Network security and firewalls |
Cisco Security, Microsoft Security | United States | Incumbent platforms |
Check Point Software | Israel | Network security |
Wiz (acquired by Google 2025) | Israel / United States | Cloud-security platform |
NSO Group, Paragon | Israel | Offensive cyber tools — controversial; sanctioned in several jurisdictions |
Thales, Entrust, Utimaco | France / US / Germany | Hardware security modules |
Okta, Auth0, Ping Identity | United States | Identity-as-a-service |
Recommended reading
- Chris Miller, Chip War (2022) — Semiconductor geopolitics
- SemiAnalysis (Dylan Patel) — Independent semiconductor analysis
- SIPRI Yearbook (annual) — Arms producers and military expenditure
- USGS Mineral Commodity Summaries (annual) — Critical minerals
- IEA Critical Minerals Market Review (annual) — Critical minerals demand
- AidData Global Chinese Development Finance Dataset — Chinese overseas lending
- NIST Post-Quantum Cryptography Standards (2024) — Post-quantum cryptography
- Kevin Davies, Editing Humanity (2020) — CRISPR history and ethics
What you just learned
Industrial power is not where you think it is. It is in a Dutch lithography company, a Taiwanese foundry, a Chinese refining complex, a handful of US defense primes, a Swiss gene-editing partnership, a few cloud landlords, and the subsea cables you have probably never seen. Knowing the apex companies and their chokepoints is the first prerequisite for thinking seriously about industrial policy.