Doubleword raises £2M pre-seed to slash AI inference costs 100x yearly

London-based AI inference startup Doubleword has secured £2 million in pre-seed funding to advance its goal of making AI 100 times cheaper annually. Founded by Meryem Arik, the 11-person team focuses on lowering AI inference costs and establishing itself within the UK’s expanding AI ecosystem. The funding round was led by UCL Technology Fund with contributions from Creator Fund and others, signaling strong investor confidence in Doubleword’s mission and approach. ¹

Doubleword’s £2 million pre-seed round was spearheaded by UCL Technology Fund and included backing from Creator Fund and the UK government’s Future Fund Breakthrough. The startup is aligned with a wider UK initiative to strengthen AI infrastructure, especially sovereign compute capabilities. Founder Meryem Arik highlighted the strategic value of local talent, noting that Doubleword’s team features former researchers from prestigious institutions such as DeepMind and Oxford. ¹

Arik emphasized Doubleword’s ambitious target to reduce AI inference costs by 100 times every year. She explained that the high expense of running current AI models restricts access for smaller companies. Doubleword’s strategy combines hardware and software optimizations aimed at making AI more affordable. "We’re building the infrastructure to make AI inference 100x cheaper every year," Arik stated in the interview, underscoring the startup’s core mission. ¹

Doubleword’s focus on cost reduction reflects a broader industry trend toward democratizing AI access. Its pricing model, detailed on the company’s website, targets enterprises deploying AI at scale. For instance, Doubleword prices its Llama 3 70B model at $6,679 per month, a fraction of the $94 million charged by competitors like Together AI for similar deployments, demonstrating a significant pricing advantage. ¹

Arik also addressed the challenges of scaling a hardware-dependent AI startup within the UK. While the country boasts strong talent pools, securing sufficient hardware resources remains a key bottleneck. "We’re working with partners to ensure we have the necessary compute power," she said. The startup’s emphasis on sovereign compute is intended to reduce geopolitical risks tied to AI infrastructure dependencies. ¹

Doubleword’s team includes former researchers from DeepMind, Oxford, and other leading institutions. Arik credited the UK’s academic ecosystem for providing a continuous flow of talent. "The UK has world-class universities and research labs," she commented. The startup leverages this talent to build a globally competitive company, with its small size seen as an advantage for agility and innovation in a fast-moving market. ¹

The interview also explored the wider implications of Doubleword’s work for the UK’s AI strategy. Arik suggested that lowering AI costs could enable smaller companies and startups to adopt AI more broadly, thus fostering innovation. She stressed the importance of sovereign compute in maintaining the UK’s technological independence. "We need to build our own infrastructure," she said, "to avoid reliance on foreign providers." ¹

Looking forward, Arik outlined plans to expand Doubleword’s team and infrastructure using the £2 million funding. The startup intends to hire more engineers and secure additional compute resources. She also hinted at future collaborations with UK-based hardware providers to further reduce costs. "Our goal is to make AI inference as affordable as possible," Arik concluded, "so that every company can benefit from it." ¹