The UK lost WWI in AI – we can’t afford to lose WWII

In 2016, Britain let go of arguably its most valuable asset for the AI era: Arm. Its energy-efficient chips already power over 90% of smartphones - the gateway through which billions will use AI - and are now climbing the stack into servers, edge devices, and custom accelerators. Selling Arm to SoftBank meant handing over a platform that could have underpinned everything from pocket assistants to hyperscale clusters. At the time, some hailed the deal as proof of confidence in post-Brexit Britain. In reality, it was a retreat. As Washington funnelled billions from the CHIPS Act into domestic fabs and Beijing bankrolled mega-plants, Britain ceded the high ground and walked away from a position of strategic strength in one of the century’s most consequential industries.

This wasn’t just a missed opportunity - it was an early warning sign. Throughout the 2010s, the UK watched from the sidelines as AI slowly, then quickly, took off. The talent was here. Researchers like Demis Hassabis, Zoubin Ghahramani, and Neil Lawrence were making foundational contributions. DeepMind was a moonshot bet, focused on AGI before anyone else took it seriously. But when it came to funding, infrastructure, and long-term strategy, we simply weren’t in the fight. While the US and China built trillion-dollar AI ecosystems, the UK allowed its best ideas, talent, and companies to be absorbed into Big Tech. No one here was writing the big cheques. The government stood by as we sold off our most valuable assets.

World War I in AI is over. That chapter is closed. But World War II is just beginning.

The Second AI War: 2025–2040

The next phase of AI will look different. The deep learning era - powered by transformer-based generative models like LLMs - is already dominated by a small group of large labs: OpenAI, DeepMind, and Anthropic. That wave has crested. The next wave won’t be about who has the biggest model, but who has the smartest strategy. AI is shifting toward integrated, multi-agent intelligence. Victory will hinge on robust sovereign infrastructure: compute, energy, and tight software–hardware integration - domains where the UK still has a fighting chance.

Compute

Access to compute will define national power in AI. Building and running state-of-the-art models requires vast amounts of processing power. Countries without domestic semiconductor capabilities or high-performance AI clusters will be relegated to second-tier roles. Right now, the UK lacks a sovereign chip industry. We have only a few AI-specific supercomputers and remain reliant on companies like NVIDIA and TSMC. That dependency will come at a cost.

Energy

AI is rapidly becoming an energy game. The next generation of models will require gigawatt-scale power just to train and operate. Without reliable, affordable energy, AI development will stall. The UK needs to rethink its energy strategy through the lens of AI - investing in stable supply, expanding nuclear, doubling down on renewables, and integrating energy planning with digital infrastructure. Otherwise, soaring costs and shortages will choke innovation.

Software–hardware integration

For two decades, the gospel was “software eats the world.” AI’s next chapter depends just as much on hardware. Frontier models need not only custom silicon but also high-bandwidth interconnects, advanced packaging, cooling, and the grid capacity to keep hyperscale clusters running. Success will rest on teams that co-design algorithms with the entire hardware stack. The UK has genuine assets here: world-leading materials science research, a long tradition of precision engineering in aerospace and life sciences manufacturing, and globally respected micro-architecture talent built around Arm. But without a concerted national drive to convert that expertise into large-scale design and fabrication, Britain risks being locked out of the foundations of tomorrow’s economy.

Financing a wartime economy

Right now, the funding mechanisms for AI in the UK are not built for this kind of war. If we do not act, we’ll keep seeing the same story unfold - our best talent and companies sold to overseas buyers with more capital and greater risk appetite. The US has long used defence contracts and federal R&D funding to drive AI investment. China has poured billions into a coordinated national effort. The UK has yet to define a strategy for how it plans to fund its AI future. We need new financial architecture. Government should act as a catalyst - de-risking long-horizon bets, unlocking pension fund capital, and signalling long-term ambition. Without that, we will keep punching below our weight.

From battle plans to battle stations

Matt Clifford’s AI Opportunities Action Plan, released earlier this year, is an important step forward. It lays out a clear framework for keeping AI breakthroughs in the UK and creating the conditions for long-term leadership. But a plan is only useful if it leads to action. That’s why we launched the AI Forum - a community of operators, researchers, and policymakers from across civil society and the public and private sectors, focused on turning the plan into progress. We are not just admiring the problem. We are asking what it will take to win.

Over the next two years, AI Forums will be held quarterly across the UK, each focused on one of the Action Plan’s key pillars: infrastructure, data, talent, regulation, procurement, and building national champions. The next forum will be held in Cambridge later this month and will focus on data and regulation. If you want to take part, reach out to the team: hello@ai-forum.fyi

The raw ingredients are already here. Talent. World-class universities. Over 500 startups generating more than $100 million a year in revenue. What we are missing is strategic alignment - and the ambition to lead.

We lost the first AI war. This time, we need to fight to win.