Rachel Reeves talked up UK sovereign AI, but the numbers show why that's a hard sell

[Glenn]

UK Chancellor Rachel Reeves used her Mansion House speech to call AI the defining technology of our generation, saying it will be crucial to national security and the country's economic future. She pointed to a new Sovereign AI unit, an advanced market commitment to quantum, an AI hardware plan and a new AI Economics Institute as evidence the government is being active and strategic rather than just watching from the sidelines

One genuine win she highlighted, Mastercard has chosen the UK as the first country in Europe to roll out its AI agent based payment system, Agent Pay for Machines, which lets software make payment transactions across the Mastercard network without a human clicking approve

The harder problem sits underneath all of it, UK datacentre running costs are dramatically higher than competing countries. A think tank report found powering a 100 megawatt datacentre in the UK costs around 226.5 million pounds a year, compared to 156.3 million in France and just 67 million in Sweden, with UK electricity costs running almost four times higher than the United States

Forrester analyst Dario Maisto, who co-authored a global sovereignty forecast, argues it isn't realistic for a mid sized country like the UK to build genuinely sovereign cloud and AI capabilities from scratch, given how many years hyperscalers have had to refine their offerings. He points out there's effectively no real alternative to Microsoft Excel's macro functionality except Google Workspace, and even that lacks equivalent advanced features, meaning true sovereignty in practice usually comes down to deciding which countries to depend on and on what terms rather than building everything domestically

The government's own British Industrial Competitiveness Scheme, aimed at cutting electricity costs for manufacturing, explicitly does not cover datacentres, despite them being exactly the kind of high energy user the scheme was designed to help. Industry body TechUK has called this a cherry picking approach that leaves strategically important digital infrastructure feeling abandoned while other sectors get relief

Reeves Speaks Up About Uk Sovereign Aiwww.computerweekly.com

[StoneCold]

The Excel macros example is such a perfect illustration of why sovereignty rhetoric is so much easier to say than actually deliver in practice

[Nina26]

Datacentres being explicitly excluded from the very scheme designed to cut energy costs for high usage industries is a strange gap in the policy

[CosmicRay67]

226 million versus 67 million in Sweden for the exact same 100 megawatt datacentre shows just how uncompetitive the UK's energy costs really are for this specific industry

[Amber Tiger]

Maisto's point about sovereignty really being about choosing who to depend on rather than building everything yourself feels like the more honest and useful framing than the political speech language

[RomanReigns02]

The Mastercard Agent Pay announcement is a nice concrete win to point to, but it doesn't really address the underlying structural cost problem at all

[Andy99]

Well she is gone now. But and we have an AI minister

[Dom_8]

Curious how much of this AI hardware plan and AI Economics Institute actually translates into real capacity versus just being another layer of strategy documents

[QuantumDay]

A new government having a fresh look at AI could be a good thing. The important part is moving beyond speeches and actually building the foundations: skills, infrastructure, public sector adoption and sensible regulation. A lot of countries can talk about AI, fewer can make it work.

[MattHardy]

The sovereign AI idea is interesting, but it is a huge challenge. The UK does not need to build every single piece of technology from scratch to benefit from AI. Sometimes being a smart customer and creating strong partnerships is more realistic.

[Dark Jaguar]

Hopefully the new priorities include practical uses rather than just chasing headlines. Things like reducing NHS admin, improving public services and helping small businesses adopt useful tools would probably have a bigger impact than another flashy announcement.

[AsteroidCandle]

The tricky bit is that AI moves so quickly. Governments are naturally slower because they have to consult, budget and create policy. By the time a committee has finished a report, the technology has already had three updates.

[Nomad22]

A focus on education would be a sensible starting point. More people understanding how to use AI properly could have a bigger effect than simply owning more computing power.

The next generation of workers will need AI skills almost regardless of their industry.

[Router]

The UK does not necessarily need to beat the US and China at everything AI-related. There is still a valuable role in specialising in areas where we already have strengths, like research, science and certain industries.

[Golden Dan]

A new minister gives an opportunity to reset priorities, but the results will depend on execution. AI investment needs patience because the benefits may appear gradually rather than as an overnight transformation.

[QuantumToken]

The public sector could be the biggest test. If AI can help reduce paperwork and make services easier to access, people will notice. If it just creates another layer of forms and meetings about AI, then we have learned nothing.

[Kieron83]

The phrase "sovereign AI" sounds impressive, but the details matter. Does it mean domestic data centres, British models, secure government systems or something else? Without a clear goal it becomes a nice slogan rather than a plan.

[Summer75]

There is also the energy question. More AI infrastructure means more demand for electricity and data centre capacity. A serious strategy has to include the boring practical stuff as well as the exciting technology side.