Britain is making a big bet on building its own AI technology. Instead of relying on American giants, it wants to create a homegrown AI model. A startup called Cosine is leading the charge. Cosine has gathered major British companies like BT, HSBC, Lloyds, NatWest, and BAE Systems to back a new AI project called Lumen Sovereign.

This AI model will be trained entirely on British soil. It will run on the Isambard-AI supercomputer in Bristol, one of Europe’s most powerful machines. The UK government is backing this effort with hundreds of millions of pounds. At London Tech Week, Prime Minister Keir Starmer announced a £400 million plan to buy AI chips and expand the country’s computing power. He wants Britain to be an “AI maker, not an AI taker.”

One big reason behind this push is control. For banks, defense companies, and critical infrastructure operators, sending sensitive data to foreign servers is risky or illegal. They need AI systems they can fully trust and control. Cosine’s model will avoid using foreign infrastructure at any point. It can even run in secure, offline environments. This eliminates many security and privacy worries.

Joining Forces for a British AI Model

The coalition behind Lumen Sovereign includes some of the UK’s biggest names. Alongside the banks and telecoms, firms like Babcock, Thales UK, Leonardo UK, PwC, and Telefónica Tech have signed up. These companies will help decide how the AI model works, what rules it follows, and how it is used. The goal is to launch the model by the end of 2026.

Cosine’s CEO, Alistair Pullen, says many businesses are waking up to the risks of depending on foreign AI providers. He points out that vendor lock-in leads to security problems, rising costs, and lack of flexibility. Cosine’s model will be built from scratch on proprietary data covering over 30 regulated workflows. It supports many programming languages, including old ones like COBOL and Fortran, which still run parts of the UK’s financial and defense systems.

These companies want AI that fits their unique needs. For example, Babcock operates in complex defense environments that require strict security. Financial firms want AI to help with anti-money laundering checks and legal document reviews. Healthcare providers need AI to manage patient data safely. Britain’s AI adoption has stalled in these areas because no one offered a fully sovereign solution — until now.

Backing British Chips and Computing Power

Training large AI models takes enormous compute power. The UK government has invested £500 million in the Sovereign AI programme to support homegrown AI labs and infrastructure. Isambard-AI runs on 5,400 Nvidia GH200 Grace Hopper Superchips, powered by zero-carbon electricity. Nvidia itself is investing £2 billion in British AI startups and expanding AI infrastructure in the UK.

Britain is also becoming the first buyer of domestically made AI chips. The government plans to spend up to £100 million buying AI inference chips from startups like Fractile. These chips are designed to run AI tasks faster and cheaper than current options. Fractile claims their chips can handle inference up to 25 times faster and at a tenth of the cost compared to existing tech.

This matters because inference—the process of running AI models to answer questions or complete tasks—happens billions of times a day. Cutting costs here could reshape AI adoption across all industries. The UK’s approach is different from the US and EU. It focuses on backing chip design and AI systems where the country already has talent, rather than trying to build full manufacturing chains from scratch.

Graphcore, another British AI chip firm, recently received $450 million from its Japanese parent company SoftBank. This shows investors still believe in UK chip engineering despite past challenges. The government’s support gives startups a chance to prove their technology in real data centers. That’s vital for attracting more private investment.

Challenges and the Road Ahead

Building a sovereign AI model is no small task. The UK still relies on American chips like Nvidia’s for now. The supercomputer’s power is impressive but small compared to giant US data centers. The real test will be if British companies accept the higher costs that come with sovereignty. Will they pay a premium to keep data safe and under control?

The coalition behind Lumen Sovereign treats this as a procurement decision, not just a slogan. They want AI that fits strict legal and security rules. The UK government is making sure its AI champions “start here, scale here, and stay here.” If Cosine and other startups can turn this compute time into real products and profits, Britain could build a strong AI sector independent of Big Tech.

This push comes at a moment when the UK’s AI startup scene is booming. British AI firms have raised more than $11 billion this year alone. The wider UK tech market is now worth $1.6 trillion. It’s a sign that homegrown innovation can thrive with government support and the right partnerships.

In short, Britain is betting big on its AI future. By building its own models, chips, and infrastructure, it hopes to cut reliance on foreign tech. If it succeeds, it could set a new example for other countries worried about digital sovereignty.