The Bank for International Settlements published BIS Bulletin No 122 analysing how a small group of “AI giants” are positioning across the five-layer AI supply chain and becoming more macroeconomically significant. The Bulletin finds that the largest AI firms are concentrated in the United States, China, Chinese Taipei, Korea and the Netherlands, and that their scale increasingly coincides with breadth as leading firms expand into multiple layers of compute, infrastructure, data tools, models and applications. Among the top 20 AI firms worldwide by supply-chain presence and valuation, the top seven are publicly listed United States firms whose combined market value is more than twice that of the next 13 firms. United States and Chinese giants tend to span most or all layers, while other leading firms such as TSMC, ASML and SK Hynix are more specialised in critical computing inputs; private United States model developers such as OpenAI and Anthropic are described as being at the frontier of model development but with more limited external presence in compute and cloud infrastructure. By end-2025, the top 20 publicly listed AI firms accounted for around 30–40% of total market capitalisation in the United States, Chinese Taipei, Korea and the Netherlands, versus about 10% in China, and their shares of total capital expenditure and revenues were also rising in several jurisdictions (capex shares at end-2024: 26% in Korea, 21% in the United States, 4% in the Netherlands and 1% in China). Using analysis of United States 10-K business descriptions, the Bulletin estimates that major United States AI firms expanded from operating in about two layers on average in the early 2000s to three to four layers in 2020–24, and that these firms account for a large share of AI-focused deal activity, including nearly 70% of deals in AI models and 33% in AI applications in the latest periods. On implications, the Bulletin argues that vertical integration could internalise costs and support further innovation, while increased control over complementary inputs may restrict entry and dampen broader innovation over time. It highlights jurisdictional “sovereign AI” strategies as reflecting trade-offs between resilience and duplicating fixed costs, and points to policy levers such as fair access to key inputs (data and computing power), interoperable standards and multi-cloud strategies, stronger competition policy, and macro-financial oversight to monitor system-wide dependencies in compute, cloud infrastructure and models and the evolving financial structure of AI giants.
Bank for International Settlements 2026-02-23
Bank for International Settlements maps global AI giants and finds growing concentration and vertical integration across the AI supply chain
The Bank for International Settlements' BIS Bulletin No 122 examines the macroeconomic significance of "AI giants" in the AI supply chain, concentrated in the U.S., China, Chinese Taipei, Korea, and the Netherlands. It notes U.S. and Chinese firms dominate multiple supply chain layers, while others like TSMC and ASML focus on specific inputs. It discusses vertical integration, jurisdictional AI strategies, and suggests policy measures for fair access, competition, and oversight.