Can the UK Become a Global AI Leader? | Meryem Arik, Doubleword

As part of Balderton Capital’s “What’s Next” series James Wise interviews Meryem Arik, Co-founder and CEO of Doubleword (a London-based AI inference infrastructure startup, formerly TitanML).

Entitled “Can the UK Become a Global AI Leader?” the conversation explores AI infrastructure, sovereign AI, open-source dynamics, and the UK’s/Europe’s potential to lead in AI despite US dominance.

Doubleword specializes in self-hosted, secure AI inference—enabling enterprises and governments to run their own models (or open-source ones) privately in their data centers or clouds, making AI faster, cheaper, and more controllable than relying on third-party APIs like OpenAI.

Key points from the discussion:

• Doubleword’s focus and use cases: Arik explains that Doubleword optimizes inference (the usage phase of AI, not training) for privacy-sensitive sectors like finance, pharma, and healthcare. Customers want sovereignty to avoid data exfiltration, rate limits, or dependency risks. There’s a growing shift toward enterprises seeking control even beyond strict privacy needs.
• Sovereign AI redefined: Arik stresses that true sovereignty is about inference location and control, not just building national models (“Brit GPT”). Nations gain power by inferring models onshore (e.g., recent UK moves like OpenAI inferencing locally). James Wise connects this to avoiding past UK pitfalls in outsourcing critical tech like energy.
• Open source’s future and competitiveness: Arik is optimistic—open source has historically won despite lower funding, with contributions from Meta, Google, and others creating ecosystem efficiencies (e.g., optimizing for Llama benefits everyone). She warns of China’s lead (e.g., DeepSeek, Qwen) and urges Western labs to open-source near-frontier models to compete. Distillation (training smaller models on larger ones’ outputs) is widespread but limited and raises IP concerns.
• Innovation, research vs. commercialization: Doubleword balances cutting-edge research (e.g., batched inference optimizations) with reliable products. Arik highlights the UK’s research strengths (attracting mission-driven talent) but commercialization weaknesses—UK enterprises adopt AI ~12 months behind the US, with more risk aversion. She sees London as an ideal R&D hub but targets North America for sales.
• Challenges and opportunities for the UK/Europe: The UK excels in deep tech talent and research but lags in commercial scaling, risk-taking culture, and buyer maturity compared to the US. Arik praises aligned investors and suggests reviving a “do things without permission” mindset. Europe could lead in deep tech by focusing on open source, sovereignty, and practical deployment. Both express optimism, with Wise noting government support (e.g., sovereign compute investments) and Arik seeing vast untapped potential in rebuilding workflows across industries.
• Broader outlook: Remaining frontiers include robotics, high-value industry transformations, and adapting to rapid change. The talk ends on an upbeat note about AI’s transformative scale and the need for obsession with real problems.

Overall, Arik is bullish on Europe’s (and specifically the UK’s) ability to lead in pragmatic, sovereign, and open-source-driven AI infrastructure, provided it bridges research excellence with bolder commercialization and ecosystem collaboration. James Wise frames it as a call to action for the UK to scale its AI breakthroughs nationally.