The Tony Blair Institute published a policy paper arguing that "AI sovereignty" should be redefined from full technological self-sufficiency to strategic agency within an interdependent system, structured around trade-offs across the AI stack (compute, data, models, applications, governance). The paper documents that the US hosts approximately 75% of global AI compute capacity, China 15%, and the remaining 10% is distributed mostly in Europe, with only 32 countries hosting AI-specific data centres. The structural implication is that most governments will remain dependent on a small number of hyperscaler and chip-supply actors, making vendor lock-in risk and partner-selection decisions core sovereign choices rather than procurement details. For financial institutions, this framing reinforces that hyperscaler and frontier-model revenue from government and quasi-government enterprise channels is structurally durable but subject to geopolitical reorientation risk as states negotiate bilateral AI access arrangements.
540 other tracked developments touching the same claims.