The matrix of regional self-sufficiency

For the discussion of digital sovereignty in Europe to make sense, it must rest on a realistic comparison: what can each major geopolitical bloc build alone, if it had no access to its partners' capabilities starting tomorrow?

The answer is sobering. Nobody on the planet is fully self-sufficient. Not the United States, with its dominant cloud and AI but its dependence on Asian fabrication and European lithography. Not China, with its rapid build-out and rare-earth dominance but its lockout from leading-edge equipment. Not Taiwan, Korea, and Japan combined, whose fabrication leadership rests on American design tools and European photolithography. Not the EU, with its ASML monopoly and automotive-chip strength but its absence from advanced logic, memory, hyperscale cloud, and frontier AI.

The matrix below makes this concrete.

Regional self-sufficiency across the technology stack (2026)

Rating: ✅ full self-sufficiency · 🟡 partial / with compromises · ❌ dependent on external supplier

Stack layer	USA	China	Taiwan + Korea + Japan	India	EU
EUV lithography	❌ (depends on ASML)	❌ (export-banned)	❌ (depends on ASML)	❌	✅ (ASML, NL)
Advanced logic ≤3 nm	🟡 (TSMC AZ from 2025)	❌	✅ (TSMC, Samsung)	❌	❌
Logic 7–14 nm	🟡	🟡 (SMIC, low yield)	✅	❌	❌
Mature processes ≥28 nm	🟡	✅ (~25–31 % world capacity)	✅	🟡 (from 2026)	🟡 (from 2027 ESMC)
HBM / advanced memory	🟡 (Micron)	🟡 (CXMT, HBM in development)	✅ (SK hynix, Samsung)	❌	❌
EDA software	✅ (Synopsys, Cadence)	🟡 (Empyrean, partial)	❌	❌	🟡 (Siemens EDA)
Wafer-fab equipment (WFE)	✅ (Applied, Lam, KLA)	🟡 (NAURA, AMEC; ~10–11 % global)	🟡 (Tokyo Electron)	❌	🟡 (ASML, ASM)
Hyperscale cloud	✅ (AWS, Azure, GCP)	✅ (Alibaba, Tencent, Huawei)	🟡	❌	❌ (~15 % home market)
Frontier AI models	✅ (OpenAI, Anthropic, Google)	✅ (Qwen, DeepSeek, Doubao)	🟡	🟡	🟡 (Mistral)
Rare earths and processing	❌ (depends on CN)	✅ (>90 % processing)	❌	🟡	❌
Energy for AI data centres	✅	✅	🟡	🟡	🟡

Sources for the table figures. China's mature-node share is TrendForce / SemiAnalysis 2024–2025 capacity projections. The China WFE share is the Center for Security and Emerging Technology's December 2025 review (Chinese suppliers ~11 % of global wafer-fab equipment, NAURA and AMEC roughly 5 % each in the segments where they compete). The rare-earth processing figure rests on USGS, IEA 2025 and CSIS critical-minerals analysis.

What the matrix shows

No region is self-sufficient across the entire stack — not the U.S., not China, not the Northeast Asian coalition. Full technological autarky is in 2026 physically and economically infeasible.

ASML remains the world's only EUV-lithography supplier. That means every advanced chip produced anywhere in the world passes through European equipment. This position is often weighted in European sovereignty discussions below its actual strategic significance.

The U.S. is dominant in design, AI, and cloud, but dependent on Asian manufacturing and European lithography. Without ASML no advanced chip would be fabricated in Arizona, Texas, or Ohio. Without TSMC, U.S. firms would have nowhere to fabricate.

China is building capacity rapidly but is cut off from EUV and therefore from anything below 7 nm. SMIC manages 7 nm using DUV multi-patterning, but at yields estimated at less than a third of TSMC's level and at 40–50 % higher cost per die.

The EU holds structural advantages in lithography and automotive chips but lacks advanced logic, memory, hyperscale cloud, and frontier AI models. The asymmetry between strengths (high-leverage, narrow) and weaknesses (broad, structural) is exactly what the operations and partnerships sections of this site try to address.

The strategic implication

When no region has everything, sovereignty stops being a question of capability accumulation and becomes a question of partnership architecture. Whom you depend on, on what terms, with what right of exit — these become the strategic variables, not how much you build at home.

Two lines follow from this.

The first concerns partnerships. After Intel cancelled Magdeburg and STMicro–GlobalFoundries put their planned Crolles megafab on hold in 2025 , the U.S. ceased to be a realistic source of new fabrication capital for Europe in 2026–2028. The natural partnership axis shifts to Korea (HBM, advanced packaging), Japan (manufacturing equipment, Rapidus collaboration via imec), and selectively Taiwan (consolidate ESMC , expect no more). The partnerships section develops this in detail.

The second concerns contracts. If sovereignty is partnership architecture, the practical building blocks are contractual. The operations section provides a checklist for what to put in IT and cloud agreements — the European customer's actual instruments of leverage in the daily reality of supplier relationships.

Anyone claiming that one region holds all the cards either does not understand the stack or is running a political campaign.