In my last article on the UK–US tech deal I argued that ambition is cheap and reality is expensive. The handshakes were bold, the investment figures even bolder, but the fine print told a familiar story: Britain leaning on Silicon Valley for its future.
Nick Clegg put it more bluntly. The former Deputy Prime Minister turned Meta executive called the prosperity deal “sloppy seconds from Silicon Valley”. He warned that the UK was in danger of becoming “a vassal state technologically, holding on to Uncle Sam’s coat tails”.
That observation haunts much of the UK’s AI story.
We are quick to declare ambition. We are less convincing when it comes to building a lasting AI economy that works for everyone.
The question is whether the UK is creating a genuine foundation for AI leadership, or simply playing host to other people’s servers.
Startups and the Scaling Wall
Startups are the jewel in the crown. The UK boasts over 1,400 AI firms, more than any other European nation. Tech Nation estimates over 2,300 VC-backed companies valued at more than $200 billion. London, Cambridge and Edinburgh have become recognised AI hubs.
Table 1. Private AI investment against the size of the economy, 2024
| Private AI investment 2024 | GDP 2024 | Per $1,000 of GDP | |
|---|---|---|---|
| United States | $109.1bn | $29.2tn | $3.74 |
| China | $9.3bn | $18.7tn | $0.50 |
| United Kingdom | $4.5bn | $3.6tn | $1.24 |
Investment: Stanford HAI, AI Index Report 2025. GDP: World Bank, current US dollars, 2024. France and Germany are not shown: the available figures are euro-denominated European venture totals on a different definition of an AI company, so they cannot be set beside these three without inventing a comparison.
The figures are sobering. In absolute terms the US dwarfs everyone else, and per unit of GDP it is still three times ahead of us. Britain does beat China on that measure, by two and a half times, but this is cold comfort. British startups still struggle to scale without foreign capital.
We invent the engine, but someone else builds the car.
Without deeper pools of late-stage investment, Britain’s best ideas will continue to be acquired or relocated to California.
Finance: Proof It Can Work
If there is one sector where AI has taken root, it is finance.
Table 2. AI in UK financial services, 2024
| Share of firms or use cases | |
|---|---|
| Firms already using AI (58% in 2022) | 75% |
| Firms planning to inside three years (14% in 2022) | 10% |
| Use cases with some automated decision-making | 55% |
| Use cases where the decision is fully autonomous | 2% |
| Use cases built on foundation models | 17% |
| Firms with only a partial understanding of the AI they use | 46% |
Bank of England and FCA, Artificial intelligence in UK financial services 2024, published 21 November 2024. The first two rows carry their 2022 comparators from the same pair of regulators' earlier machine learning survey.
Three quarters of firms are running it. Just under half admit they only partly understand the thing they have bought. Adoption and control are not the same measurement, and only one of them is going up.
Banks now use AI for fraud detection, credit scoring and compliance. Insurers rely on it for claims. Asset managers employ it for faster analysis. Startups like Onfido, Eigen and Tractable have become global names.
The regulator has helped. The FCA’s sandbox created a way to test innovation without destabilising the system. London’s status as a global hub provided data, customers and talent.
Finance shows the UK can deliver real adoption, not just pilots. But finance is unusually fertile ground. The harder test is whether manufacturing, healthcare and SMEs can follow.
Healthcare: Promise, Pilots and Paralysis
The NHS holds more patient data than almost any other system in the world. Since 2019 the NHS AI Lab has funded projects ranging from breast cancer detection to cardiology and hospital demand forecasting.
Table 3. Three NHS AI programmes and what happened next
| Programme | What it measured | Where it got to |
|---|---|---|
| Brainomix e-Stroke, stroke scan reading | Door-in-door-out time fell from 140 minutes to 79. Patients reaching functional independence rose from 16% to 48%. | Five stroke networks in England |
| Kheiron Mia, breast screening | 12% more cancers found than routine practice across 10,889 patients, with no rise in unnecessary recalls and modelled workload savings up to 30% | One prospective evaluation, at NHS Grampian |
| Chest X-ray and CT reading for lung cancer | No outcome figure published | £21m spread across 64 NHS trusts |
Stroke: DHSC, 27 December 2022. Breast screening: NHS Grampian's prospective evaluation of Mia, funded by the AI in Health and Care Award, announced March 2024. Lung: DHSC, 30 October 2023.
Read the right-hand column rather than the middle one. Tripling the number of stroke patients who walk out of hospital independent is the sort of result that would justify a national rollout on its own. It got five networks.
The outcomes look strong on paper. But adoption beyond pilots is painfully slow. Hospitals use fragmented IT. Staff lack time for training. Regulators demand exhaustive evidence.
The pitfall is obvious. Pilots generate headlines, but patients rarely see the benefit.
Unless the NHS moves from experiments to national systems, AI will be remembered as another round of PowerPoint slides.
Industry: Still Stuck in the Slog
Manufacturing accounts for around 10% of UK GDP. The country ranks 12th globally. AI is hailed as the way to reverse decline.
Table 4. UK manufacturers and AI, 2024
| Share | |
|---|---|
| Intend to increase AI spending within the year | 75% |
| Already using AI and reporting better operational efficiency | 69% |
| Already using AI and reporting higher productivity | 61% |
| Deploying AI in production processes | 33% |
| Still at the conception stage of digital transformation | 30% |
| Consider themselves knowledgeable about AI | 16% |
| Describe their expertise as very knowledgeable | 7% |
Make UK, Future Factories Powered by AI. Survey of 151 manufacturers, July to August 2024. The efficiency and productivity rows are shares of those already using AI; the rest are shares of all respondents.
Three quarters are about to spend more on a thing one in six claims to understand. Nobody buys a lathe that way.
There are pockets of progress. Predictive maintenance works. Aerospace firms test digital twins. But most plants are stuck with old equipment, siloed data and under-skilled workforces.
Germany’s Industrie 4.0 push is more coordinated. China is embedding AI across state-backed factories. The US is pouring subsidies into advanced manufacturing. Britain risks falling further behind.
Energy: Feeding the Beast
AI consumes enormous power. Datacentres under construction in the UK will require gigawatts of energy and large volumes of water for cooling. Local concerns are growing.
The government links AI growth to new nuclear reactors, hoping clean energy can satisfy both households and hyperscalers. At the same time AI is being used by National Grid to forecast renewable output, and by Octopus Energy to manage household demand.
Table 5. What UK data centres draw from the grid
| 2023 | 2030, forecast | |
|---|---|---|
| Electricity consumed | 5.0 TWh | 26.2 TWh |
| Share of all UK electricity demand | 2% | 8.8% |
| Share of UK commercial electricity use | 7% | 30.4% |
| IT power capacity | 2.9 GW | about 9 GW |
Oxford Economics, The UK's data centre boom, 8 December 2025. The 2030 column is a forecast, not a measurement.
AI is both a consumer and an enabler of energy. The race between building clean energy and expanding AI infrastructure will determine whether the UK’s ambitions are sustainable or self defeating.
SMEs and Skills: The Real Test
The UK’s AI economy will stand or fall on the adoption by small and medium firms. They employ most of the workforce. Yet they remain on the sidelines.
Table 6. AI adoption by UK businesses
| Share using at least one AI technology | |
|---|---|
| Businesses with 10 or more staff, late 2023 | 12% |
| Businesses with 10 or more staff, June 2026 | 35% |
| Businesses with fewer than 10 staff, June 2026 | 28% |
| Businesses with 250 or more staff, June 2026 | 49% |
| Information and communication, June 2026 | 58% |
| Construction, June 2026 | 13% |
ONS, Artificial intelligence in UK businesses: 2023 to 2026. The ONS does not publish the 10 to 49 and 50 to 249 bands separately, so they are not shown.
Adoption tripled in under three years, which sounds like the argument is over. It is not. The gap between the smallest firms and the largest is still twenty-one points, and it is the same twenty-one points it always was, just further up the page.
Government promises are bold. The SME Digital Taskforce calls for a minister for SME tech, an “online CTO as a service”, and tax incentives. The ambition is to make UK SMEs the most AI-confident in the G7 by 2035.
But reality is harder.
Many SMEs cannot see the relevance. A regional accountancy firm or family-run logistics operator struggles to find a clear use case. Costs are another barrier. Even cheap tools require training, integration and process change. For thin-margin businesses this is not priority, it is distraction.
Skills are scarce. AI researchers cluster in universities and startups, not SMEs. Retraining the existing workforce is possible but requires decades, not months. Bootcamps and visas help, but they will not transform an entire economy overnight.
Without SME adoption, AI remains an elite sport for corporates and startups.
The pitfall is assuming SMEs will eventually catch up. Unlike websites or email, AI is not plug and play. It requires integration into workflows and data pipelines. Without deliberate design and policy, the gap could last for decades.
A lasting AI economy must include SMEs. That means simple, off-the-shelf tools that solve daily problems, clear regulations to build trust, and long-term investment in workforce skills. Otherwise AI will boost productivity for a few, while leaving the majority of businesses untouched.
Ambition or Achievement
Britain is very good at ambition. We are excellent at pilots, taskforces and glossy reports. We are even world class at attracting the next American hyperscaler to build a datacentre in our backyards.
What we are not yet good at is converting activity into achievement. AI could boost productivity, improve public services and create jobs. But until we solve the scaling problem for startups, pilots, and SMEs, we risk building an AI economy that looks impressive on slides but hollow in practice.
The pitfall is not ambition. It is execution. Unless Britain finds the capital, skills and courage to push through, the AI revolution will be another missed opportunity.
The question is not whether Britain has ambition. It is whether it has the will to turn that ambition into lasting change for everyone.
References
-
Stanford AI Index 2025
-
Financial Times, UK tech ecosystem report 2025
-
Bank of England/FCA AI adoption survey 2024
-
NHS AI Lab, programme documentation
-
TechUK, Industrial AI Sprint 2025
-
National Grid ESO, AI use cases 2024
-
Office for National Statistics, AI adoption in UK businesses 2023
-
UK SME Digital Adoption Taskforce report 2025
-
Guardian, UK–US prosperity deal