Elon Musk’s Terafab initiative has moved beyond speculation into tangible execution, and its true objective isn’t land acquisition, tax breaks, or capital raised—it’s access to the world’s scarcest semiconductor manufacturing asset: ASML’s extreme ultraviolet (EUV) lithography machines. This ambition not only exposes the fundamental bottleneck in AI chip scaling but thrusts ASML into the geopolitical triangle between the United States, Taiwan, China, and mainland China.
ASML stands alone as the sole commercial supplier of EUV systems. A single Twinscan NXE:3600D costs over $200 million, with delivery lead times exceeding 18 months, all tightly governed by Dutch export controls. In 2025, ASML shipped just 72 EUV tools globally—over 80% went to TSMC, the foundry based in Taiwan, China that manufactures more than 90% of NVIDIA’s H100 and B100 AI accelerators on advanced nodes. For Musk to build sub-7nm logic capacity via Terafab, bypassing ASML is impossible.
Terafab’s strategic intent extends far beyond securing redundant compute for xAI or Tesla. Musk is attempting to replicate SpaceX’s vertical integration playbook—controlling foundational infrastructure from end applications like Starlink and autonomous driving. Yet semiconductor manufacturing is orders of magnitude more complex than rocket assembly. Even with an EUV tool, Terafab would lack TSMC’s decades of process recipes, yield management expertise, and EDA ecosystem integration. Crucially, EUV machines are only one node in a tightly controlled supply chain: photomasks, photoresists, and metrology tools are equally constrained under U.S.-Japan-Netherlands export regimes.
This places ASML at the epicenter of geopolitical friction. Following the U.S. Commerce Department’s tightened semiconductor equipment restrictions in 2024, ASML halted shipments of its NXT:2050i and newer deep ultraviolet (DUV) systems to mainland China. But if Terafab locates in the U.S., it could theoretically skirt some restrictions. The real barrier, however, is allocation: ASML’s EUV capacity through 2027 is already locked in by TSMC, Samsung, and Intel. Even with premium pricing, Musk faces near-zero near-term availability. I judge that Terafab will likely adopt a hybrid approach—starting with mature nodes (e.g., 28nm) for automotive AI chips while lobbying to be designated a CHIPS Act priority to secure future EUV access.
Underlying this scramble is a structural fragility in the AI supply chain. NVIDIA dominates AI accelerators but relies entirely on TSMC for manufacturing. TSMC, in turn, concentrates its EUV capacity in just two campuses in Hsinchu and Tainan. Any disruption—geopolitical tension across the Taiwan Strait, seismic risk, or sudden policy shifts in Washington or The Hague—could halt global AI training clusters. Terafab is fundamentally a hedge against this single point of failure. Yet building a greenfield EUV-capable fab requires over $20 billion and five-plus years—capital intensity even Musk’s combined SpaceX-Tesla cash flow may struggle to sustain alone.
More concerning, Terafab could accelerate the fragmentation of global semiconductor manufacturing. Nations are racing to establish domestic advanced-node capacity: the U.S. backs Intel and Micron, the EU bets on STMicroelectronics and Infineon, while mainland China pushes SMIC and YMTC. But EUV scarcity ensures only a handful of players will reach the 3nm era. ASML’s production ceiling is quietly becoming the dividing line between “AI-sovereign” states and “compute-dependent” ones.
ASML itself is caught in a bind. As a Dutch company, it must reconcile U.S. export controls, European industrial autonomy demands, and global customer commitments. Its 2025 market cap briefly surpassed LVMH to become Europe’s most valuable firm—but that valuation carries immense geopolitical risk. If Terafab secures EUV tools, it could erode TSMC’s near-monopoly on advanced logic, reshaping AI hardware power dynamics. If it fails, the TSMC–NVIDIA alliance becomes even more entrenched.
Musk’s bet is, in essence, a stress test of the global semiconductor order. As the AI race shifts from algorithms to physical fabrication, control over lithography equals control over the next decade of compute. The critical question now is whether ASML can remain a neutral technological hub in a fractured world—or whether it will be forced to pick sides, becoming a new frontline in the tech cold war.