Who counts as the winner of the U.S.–China AI race: the country developing the most capable models, the one reliably supplying computing infrastructure, or the one translating adoption into economic benefits? This article examines competition principally during 2026–2030, with conditional scenarios extending to 2035. It separates research, semiconductors, electricity, finance, talent, enterprise adoption and international standards, distinguishing technological performance from outcomes accruing to states, businesses and users.

The analysis compares renewed leadership, expanded catch-up and practical deployment, asymmetric coexistence, and fragmentation combined with investment adjustment. It draws on government policy and export-control documents, public statistics, international organizations, university and research evaluations, and developer technical materials. Documented facts, source assessments and SG Group’s conditional judgments remain distinct. Model versions and evaluation dates, policy targets and realized outcomes, announced infrastructure and operational capacity, and experimentation and repeated use are examined separately, with attention to source boundaries and gaps in observation.

The subscriber article uses multiple clocks, capability comparisons, a trigger matrix, early-warning indicators organized by horizon, falsification conditions and a monitoring dashboard. It traces transmission from infrastructure and energy demand to business costs, trade, employment, fiscal accounts and financial conditions. For Japanese businesses, it examines information governance, supply dependencies, contracts, switching and recovery. The focus is macroeconomic analysis and business continuity, not recommendations concerning particular financial instruments.