Editorial: We Need Faster Sites, Not Just Larger Data Centers

Goldman Sachs' two analyses from June and July, seemingly on different topics—one focusing on South Korea's humanoid supply chain and the other on shifts in AI investment landscapes—reveal a fundamental change in AI competition rules. During this transition, a significant opportunity gap has emerged in South Korea. The shift from training-centric AI to inference-driven applications is reshaping the industry, with bottlenecks moving from semiconductors to processor communication and from data centers' copper wiring to fiber optics. While top 5% firms now consume three times more tokens than average companies, the true bottleneck lies in data. South Korea's potential in humanoid robotics hinges on domestic deployment, not just production. Despite government investments targeting 2029 production goals, the scale remains insufficient to drive the data flywheel critical for leadership in this new paradigm.