The small club of AI giants | Number Theory

.

Artificial Intelligence (AI) will change the world economy as we know it. Who owns the technology though? What does it take to “produce” AI? Are there differences across firms leading the AI charge? A Bank of International Settlements (BIS) bulletin by economists Jon Frost, Kumar Rishabh and Vatsala Shreeti gives a useful summary of the answers. Here are three charts which describe this.

Representational image. (REUTERS)

The small club of AI giants

• 
  AI dominance entails a five-layered supply chain; only a couple of companies are present across all

  If one were to paraphrase the modern food supply chain analogy, AI is the chip-to-bot equivalent of farm-to-fork. At the first level of the AI supply chain is ‘compute’ hardware. It requires microprocessors and advanced chips. Then comes the data centre and cloud infrastructure, which is where the hardware described above is deployed. This infrastructure is then fed massive amounts of data—proprietary or public—and ‘trained’ on it which then leads to a model being generated which can claim to have AI. The final part of the AI supply chain is a consumer facing application which uses this AI to handle particular tasks. Each of these stages is so difficult to master that only a handful of companies have managed to break ground in the area. Even within this small club, only a couple or so are present across all of these verticals. Among the top 20 global AI firms, the top seven are all US-based listed companies, and together are worth more than twice the next 13 combined. By contrast, several pivotal firms stay specialised. ASML, SK Hynix and TSMC sit at chokepoints in chip equipment, memory and fabrication, but rarely cover the wider stack. To be sure, US private AI firms such as OpenAI and Anthropic sit at the cutting edge of model development, but unlike the tech incumbents they do not control chips, data centres or cloud platforms. They are scaling up indirectly, securing long-term compute partnerships and raising capital linked to infrastructure build-outs.
• 
  AI giants have a big economic footprint in their home countries now

  In several economies, these firms now make up a striking share of equity markets, which means stock indices are increasingly being driven by a narrow set of AI-linked names. BIS estimates that by end-2025, AI giants accounted for roughly 30% to 40% of total listed market value in the United States, Taiwan, South Korea and the Netherlands. China stands apart, with a much smaller share and a visible hit to tech-stock valuations after 2020 following regulatory reforms in the sector. AI firms’ heft is also showing up beyond markets. By end-2024, they accounted for about 21% of total capex in the US and around 26% in South Korea. In terms of GDP share this number was around 4.9% of US in 2025 and roughly 2.5% in the euro area and 3.5% in China in 2024. In the US and South Korea, AI giants already contribute around a tenth of listed non-financial revenues. The takeaway is that AI is no longer just a tech story, it is becoming a market and investment story for some of the world’s biggest economies.
• 
  AI giants are branching out through deals and investments

  With growing economic heft, AI giants are building a wider footprint across the AI supply chain through acquisitions, partnerships and investment. The BIS tracks this for US-listed firms using 10-K filings, and finds their roles have steadily broadened since 2000, from a presence in a couple of layers to three or more on average in recent years. Chinese tech firms show a similar drive to build across layers, though the paper notes that their expansion often takes a different route, shaped by domestic platforms and a policy environment that has also pulled down valuations in recent years. Even so, the underlying incentive is the same, to reduce dependence on external suppliers and strengthen control over key parts of the stack. Deal data suggests the pace has picked up. Since 2017, giants have tilted deal-making towards AI rather than non-AI transactions, with activity spreading across more layers, especially downstream applications where AI can be deployed and monetised at scale. In model development too, giants account for a dominant share of deals, reinforcing how frontier capability is increasingly shaped by a small group with the deepest pockets and best access to infrastructure.
• Conclusion

  AI may be the next general-purpose technology, but its supply chain is concentrating fast in a handful of firms and a few chokepoint suppliers. For everyone else including India, the strategic question now is less about building a full-stack champion overnight and more about choosing where to compete in the stack, while securing access to the layers they cannot easily replicate.