The exit option: portable workloads and a sovereign British cloud

British government and much of the country's critical infrastructure now run on three American cloud providers. The arrangement works well in normal conditions. The case for it rests on those conditions holding. The risk sits elsewhere. The CLOUD Act obliges American providers to comply with United States legal process wherever the data physically lives, and no contract signed in London binds a future executive order signed in Washington.^[1] Contractual safeguards fail in exactly the circumstances where sovereignty matters, because those are the circumstances in which the counterparty's own government is the one acting.

The current implicit policy is to accept this and manage it with contracts, data classification, and a thin layer of sovereign infrastructure for the most sensitive workloads. Whether that position holds depends on a judgement about abnormal times, and that judgement is being made today by default, renewed silently with every framework contract, rather than being taken once, in the open, by people who answer for it.

The proposal

The proposal has two moves, and the first gives the second its force.

Make government workloads portable by design. Public cloud deployments should be built on open-source orchestration so that workloads can move between providers, and every major government cloud deployment should hold one node on sovereign infrastructure. For ordinary application workloads, an exit programme measured in months rather than years then becomes credible. The exit option earns its value unexercised. A supplier prices and behaves differently when the customer can leave, and a foreign government weighing an action that would trigger a British exit is also weighing the loss of a major revenue stream for its own companies.

Underwrite a sovereign cloud, with the state as founding client. The United Kingdom has no hyperscaler-class provider to move to; no British provider operates at anything approaching hyperscaler scale, and the most ambitious previous attempt went into administration in 2022.^[2] The capability has to be built, through a joint venture drawing together the domestic operators that exist, a European partner with real scale, and a state investment vehicle, anchored by a founding-client commitment from government and the regulated sectors. France and Germany have providers of genuine size with explicit sovereignty positioning, OVHcloud among them, and a British-European venture would combine engineering depth and market scale that neither side can reach alone.

Portability has limits. Open orchestration covers most ordinary government workloads well, and managed databases only partially, at the price of running them yourself. It barely reaches frontier artificial intelligence workloads, because proprietary accelerator stacks tie them to their providers. The leverage is therefore strong where most of government computing sits, and weak at the frontier until sovereign compute catches up. That is a longer build, and the venture should start on it early.

The champion that must not sleep

European industrial policy has attempted versions of this several times, and the recurring failure is the slumbering champion, a protected entity that never becomes good because it never has to. The discipline that prevents it has four parts, and each one must be in the founding documents rather than the press release.

• The founding-client commitment carries a sunset. Government anchors the venture for five years and re-tenders at year five, so the venture must be competitive against the hyperscalers on the open market by then or lose its anchor revenue.
• The board is independent, with no ministerial appointments, and its mandate is explicitly time-limited.
• The chief executive is a founder-quality operator with equity at stake, recruited from the market, because seconding a senior official is precisely how the slumbering pattern starts.
• The venture gets no protection from competition. Government keeps using the American providers for the bulk of workloads through the transition, and the venture wins share on merit.

The design underwrites a competitor through demand commitment and time-bounded capital, then forces it to stand on its own. Airbus is the precedent, in both halves: the founding-client discipline that built it, and the eventual exit of state ownership that proved it could stand alone.

The research base makes it durable

The venture should be paired with a small number of named university centres, each owning one of the hard problems: open orchestration at scale, sovereign accelerator design, power-efficient data centres, confidential computing, safety and alignment. The centres run on commissioning discipline of the kind the Advanced Research and Invention Agency (ARIA) was built for, with fixed-term programme directors judged on deployable capability, because the national institute model has already shown how a distributed research body becomes slow and paper-bound without it. Universities need this as much as the venture does; more than four in ten are in deficit,^[3] and applied income flowing from real problem statements, with intellectual property licensed back to the venture, is better money than another round of grants.

What it costs

A hyperscaler-class region with artificial intelligence capability is a build measured in the low billions; a ten-year programme of several regions with serious compute sits in the tens of billions. The funding is multi-vehicle by nature, spanning the Treasury, the national investment institutions, departmental commitments and the savings that a credible exit option extracts from renegotiated hyperscaler contracts. Those savings would fund part of the build. Power is the harder constraint. The grid cannot connect new hundred-megawatt loads on demand in the south east, so the venture sites its capacity where headroom exists, in the north and the devolved nations, and the regional economies get the jobs the constraint creates.

Objections

Europe already tried this and failed. GAIA-X failed on governance and the absence of a paying anchor customer, and the previous British attempt failed for want of exactly the founding-client commitment this proposal centres. Airbus shows the premise can work, and every sovereign cloud attempt that failed lacked its disciplines. The four rules above are the answer to that record.

It is unaffordable in this fiscal climate. The headline number ignores the other side of the ledger. Renegotiated hyperscaler pricing under a credible exit option offsets part of the cost, adoption is deliberately partial rather than total, and the alternative is not zero spend but continued full-price dependence, renewed indefinitely.

Contracts and classification are enough. They are enough for every scenario in which nothing serious goes wrong. How much sovereignty matters when conditions are abnormal is a question of strategic risk appetite, and ministers who answer for that appetite should settle it deliberately, rather than let it accumulate out of procurement defaults.

Government should not pick winners. That is correct, and this design does not. It creates a competitor and exposes it to the market with a five-year clock running. The state's contribution is demand certainty and patient capital, both of which it already gives the incumbents, unpriced, with every framework renewal.