Is it time for an AI sovereign wealth fund in the UK?
By Peter Springett ยท 16 September 2026
I had a bit of fun with Claude pulling together a proposal for an AI wealth fund in the UK. It's written as a two-page Treasury briefing note to attract the attention of a Whitehall wonk. Let me know what you think.
An AI Sovereign Wealth Fund for the UK
Converting AI-generated economic surplus into shared public capability
THE PROBLEM
AI will generate enormous productivity gains over the next decade. The central question is no longer whether those gains arrive, it is who captures them.
The countries that build mechanisms to capture and reinvest AI productivity will compound national competitiveness. Those that do not will become net consumers of foreign-owned intelligence infrastructure.
Left uncaptured, the surplus will concentrate in a small number of private actors. That is not a neutral outcome. It means deeper inequality, weakening democratic accountability, and a society structurally dependent on systems it does not own.
The moral case is straightforward: AI productivity rests on decades of public investment: research funding, education, infrastructure, legal systems, publicly generated data. Without a capture mechanism, that public foundation becomes private rent.
THE PROPOSITION
A UK AI Sovereign Wealth Fund, structured as two distinct institutions with a shared mandate.
The Wealth Portfolio: global, diversified, professionally managed. Holds equities, bonds, infrastructure funds and private markets exposure, with capped AI-sector concentration. Governed at arm's length from ministers. Its purpose is intergenerational wealth preservation, not industrial policy.
The Capability Bank: a sister institution that invests in the public infrastructure of an AI economy: compute capacity, national data trusts, AI safety institutes, technical education, care infrastructure, community institutions, clean energy. Closer to a development bank, with a democratic mandate and independent charter.
The Wealth Portfolio captures upside. The Capability Bank builds the society able to use it.
FUNDING MODEL
Corporation tax on AI profits is too easily avoided. A more robust hybrid:
Compute levy: tax on GPU/TPU usage above a threshold. Targets the input to automation rather than declared profit. Harder to offshore, legally defensible, already gaining traction in policy circles.
Initial collection could occur through UK datacentre operators, hyperscale cloud providers, or regulated frontier-model API providers, using existing energy and digital-services reporting infrastructure.
Equity issuance: frontier-scale AI firms required to issue warrants into the fund as a condition of operating at scale in the UK market. Minority stakes only, no governance entanglement.
Licensing fees for frontier-model deployment, and a headcount-adjusted productivity levy for large firms growing revenue without growing payroll.
The compute levy should be the first mechanism established. It generates early revenue and builds institutional credibility before more contested mechanisms are introduced.
GOVERNANCE
Norway's sovereign wealth fund is the relevant precedent: Parliament sets the framework, the Ministry holds accountability, and operational management is fully delegated under clear fiduciary rules. That firewall (elected purpose, independent execution) prevents the fund becoming a political instrument.
Statutory ring-fencing requiring supermajority parliamentary approval for changes to drawdown rules, investment mandate or institutional purpose.
Cross-party appointment to the oversight board. Transparent annual reporting to Parliament. The Capability Bank governed separately, with its own independent charter and mission lock.
THE TRANSITION PROBLEM
Displacement will arrive before the fund is large enough to help. Years one to five require a separate AI Transition Authority, funded by near-term levies, not fund returns, to manage income smoothing, retraining and regional support. This is the bridge to the fund.
THE ASK
We recommend that HM Treasury commission an independent feasibility study into the fiscal, legal and governance architecture of a UK AI Sovereign Wealth Fund, reporting within twelve months.
The question is no longer whether AI wealth will concentrate. The policy question is whether the UK chooses to remain a tax base for foreign-owned intelligence, or becomes a co-owner of the infrastructure shaping its economic future.