Global Hubs
Global hubs are the cities that serve as command nodes for the space of flows: New York, London, Tokyo, Shanghai, San Francisco, and a small number of others whose connectivity to the global network gives them disproportionate influence over what happens in every place connected to them. They are not the largest cities by population, nor necessarily the wealthiest, but the most connected — the nodes through which global flows pass at highest density, producing the specific kind of power that comes from being indispensable to the network's operation. In the AI economy, global hubs have been joined by a specific subset — the cities that host the foundation model companies and the data centers that run them — whose concentration reproduces and intensifies the hub structure Castells documented in earlier network transitions.
In the AI Story
The hub structure of the network society reflects what network scientists call preferential attachment: new connections prefer nodes that already have many connections, producing a power-law distribution in which a few hubs accumulate disproportionate centrality while the vast majority of nodes remain sparsely connected. The phenomenon is not specific to technology networks; it appears in citation networks, social networks, transportation networks, and economic networks. What makes the AI hub structure distinctive is the speed at which it consolidated and the specific form of power hub-hosts exercise over the networks they anchor.
The geography of AI hubs is strikingly concentrated. A handful of cities host the companies building foundation models. A slightly larger set hosts the data centers and research institutions that support them. The economic and political implications of this concentration are not yet fully understood — Castells's framework suggests they will be substantial and durable, because the accumulated advantages of the hub positions are self-reinforcing through the mechanisms that produced them in the first place.
The response Castells's framework recommends is not to oppose hub formation — hub structures appear to be structural features of network systems — but to govern them. The task is to ensure that hub positions do not translate into unaccountable power over the broader network, that peripheral nodes retain meaningful capability to participate and withdraw, and that the benefits of hub productivity flow through the network rather than accumulating only at the hub itself. This is the structural analog of antitrust regulation in the earlier industrial economy, adapted to network conditions.
Origin
The concept is implicit in Castells's earliest work on the informational city and developed explicitly across his analyses of global production networks, media systems, and digital governance.
Key Ideas
Hubs form through preferential attachment. New connections prefer well-connected nodes, producing power-law distribution in which hubs accumulate disproportionate centrality.
AI hubs are geographically concentrated. A small number of cities host the companies, infrastructure, and institutions that anchor the AI network.
Hub positions are self-reinforcing. Accumulated advantages produce new advantages through the same mechanisms that generated the initial position.
Governance is the response. The task is not to prevent hub formation but to ensure that hub power remains accountable to the network it anchors.
Appears in the Orange Pill Cycle
Further reading
- Manuel Castells, The Rise of the Network Society (Blackwell, 1996)
- Albert-László Barabási, Linked (Perseus, 2002)
- Saskia Sassen, The Global City (Princeton University Press, 1991)