South Korea Launches $500 B ‘San Francisco AI Declaration’ with Nvidia–SK Megaproject

State-orchestrated tech pushes are not new: Japan’s 1982 Fifth-Generation Computer Project and Taiwan’s early-1990s Hsinchu Science Park similarly fused government capital with foreign know-how to capture strategic computing sectors—success for Taiwan, disappointment for Japan. Lee’s declaration tries to exploit South Korea’s memory-chip dominance just as U.S.–China tech decoupling (Huawei ban 2019, CHIPS Act 2022) splinters semiconductor geography. By binding Nvidia’s GPU roadmap to Korean HBM factories and 2-GW domestic compute farms, Seoul hopes to shift the locus of AI infrastructure toward U.S. allies, diverting supply chains that currently flow through Taiwan and, increasingly precariously, mainland China. If the plan sticks, it could define where the next century’s “digital oil”—AI compute capacity—is minted; if electricity, talent or geopolitical frictions falter, the effort may echo Japan’s stalled bid. Either way, the move signals that, eight decades after the first transistor, control over fabrication rather than algorithms is again the currency of geopolitical power.