Mapping How Alliances, Acquisitions, and Antitrust are Shaping the Frontier AI Industry

I’m excited to share my working paper: “On Labs and Fabs: Mapping How Alliances, Acquisitions, and Antitrust are Shaping the Frontier AI Industry.” It will be on Arxiv in a few days.

This paper began with a question: Why are there so many strategic partnerships in the frontier AI industry? How might this impact regulatory proposals?

Driven by this, I spent several months studying the AI supply chain. I mapped 25 leading companies, from litography to AI labs, listing their hundreds of interrelationships and dozens of antitrust litigations and M&As they were involved in. The full mapping is available here.

Here’s what I found out:

Horizontal integration throughout the AI supply chain is mainly driven by market consolidation through natural growth, not acquisitions.
There is a noticeable trend of backward vertical integration in both the lithography and chip manufacturing industries.
Downstream in the chain, we predominantly see a significant number of strategic partnerships between AI labs and cloud companies.
Big tech companies frequently engage in conglomerate integration by acquiring startups focused on narrow AI applications or setting up strategic partnerships with foundational AI labs.
Governmental actions significantly shape the AI supply chain through subsidies, sanctions, and industrial policy. However, there has been limited antitrust action on the supply chain, with the most noteworthy cases being Applied Materials and Tokyo Electron, and ARM and Nvidia.

I tentatively conclude that the factors leading to this substantial vertical integration and quasi-integration include:

Companies seeking to ensure compute access for large training runs.
Big tech companies balancing specialization with broad capabilities.
High transaction costs in R&D, especially for companies upstream that develop or deeply engage with EUV technology for chip manufacturing.
A market in its initial stages of development, which has not yet developed large markets for major, impersonal transactions due to the lack of established ways of working.
A desire for companies to be secretive.

Most importantly, these factors share many similarities with the rise of the internet and digital economy but also have significant differences. We need lots more empirical research! I hope this mapping will be helpful towards that goal.

Some research questions I’m most curious about:

How does the prevailing market structure influence the trajectory of AI industry advancements?
How does the current market structure within the AI supply chain affect the implementation and effectiveness of current regulatory proposals?
Will structural remedies be necessary to establish effective regulatory frameworks in the AI industry?

This is still a work in progress, so comments are very welcome! I’d like to thank the invaluable mentorship of Charlotte Siegmann and Andrew J. Koh along the way, the support of the Swiss Existential Risk (kudos to Tobias Häberli) and the Cambridge-Boston Alignment Initiative, as well as the interviewed experts and those who commented on earlier versions of this working paper.