This London Startup Just Raised $100M to Make AI Smarter and Cheaper

Jan de Vries · 2026-08-20

The AI race has a new kind of winner emerging. It's not the company building the biggest model or the fastest chip. It's the one that figures out how to make them all work together seamlessly. Callosum, a London-based startup, just closed a massive $100 million Seed round to do exactly that. The company is building the software layer that unifies different AI models and chips, a concept they call "heterogeneous intelligence." The goal? To make AI faster, more efficient, and dramatically more cost-effective. This isn't just another funding announcement. It's a signal that the AI industry is shifting from a focus on training giant models to the messy, practical work of running them efficiently in the real world. ### The Big Bet: Orchestration Over Domination Founded in 2025 by Jascha Achterberg and Danyal Akarca, Callosum's core belief is simple: the future of AI isn't about owning the best single piece of infrastructure. It's about orchestrating many different models across many types of silicon. Think of it like a master chef. You can have the best ingredients (chips) and the best recipes (models), but without a chef who knows how to combine them perfectly for each dish, you're wasting resources. Callosum wants to be that chef for the AI world. The round was led by Atomico, with major participation from Plural, DCVC, and the UK Sovereign AI Fund. Notably, this marks the first-ever investment from that fund, a $660 million UK government initiative launched in April 2026 to back strategically important British AI startups. ### Why This Matters for Your Bottom Line Niklas Zennström, CEO of Atomico, put it bluntly: "As AI moves from training to inference, the question is no longer just who builds the best models. It's how we deliver intelligence efficiently, across an increasingly diverse landscape of models and chips." That's the key shift. Training is expensive, but running AI (inference) is where the real, ongoing costs hit. If you can cut those costs while boosting performance, you unlock serious competitive advantages. Callosum is already showing results. Their "Tailored Inference" APIs are live in production, delivering performance and cost improvements in fields like cybersecurity and finance. ### Three Moves That Could Change the Game With this fresh capital, Callosum is pushing forward on three fronts: - **Global Compute Partnerships:** A "flagship partnership" with Cerebras Systems to deliver low-latency, multi-agent AI at scale, plus deals with chipmaker Rebellions and other infrastructure providers worldwide. - **Programmable Heterogeneity:** The ability to break AI workloads into smaller tasks and route each one to the model and chip that runs it most efficiently, balancing cost, energy, and speed. - **Tailored Inference in Production:** A family of APIs already proving its worth in real-world applications. Andrew Feldman, CEO of Cerebras Systems, explained the value: "By integrating Cerebras into Callosum's platform, we're making ultra-low-latency inference available exactly where it creates the greatest impact." ### The Bottom Line The $10.25 million pre-Seed round in February 2026 was just a warm-up. This $100 million raise is a validation. Callosum's team, drawn from Cambridge, Oxford, MIT, and Intel, isn't just building another AI tool. They're building the plumbing that could make the entire AI ecosystem work better. For businesses, this means one thing: the era of choosing between speed, cost, and capability in AI might be ending. With smart orchestration, you can have all three. That's not just good news for Callosum; it's good news for anyone who wants to use AI without breaking the bank.