Tendrils, a company building a general-purpose processor with thousands of cores operating in parallel, has raised £23 million in a seed round led by Benchmark. Visionaries Club, the UK Sovereign AI Fund, FR8, Antimatter, Cloudberry VC, Conviction, Embassy Ventures and SV Angels also took part, along with angel investors. The round is intended to rebuild general-purpose computing.
Tendrils is answering the question of what it would look like to redo the CPU. Contemporary CPUs are built under the constraint of backwards compatibility with sequential, single-core chips, and are bottlenecked by cache coherency protocols and reorder buffers. Clock speeds stalled in the mid-2000s, and per-core performance gains have slowed to roughly 5 to 10 per cent a year. In some agentic workloads, tool execution on the CPU accounts for up to 90 per cent of end-to-end latency. Rather than starting from a sequential design, the processor runs on interaction nets, a graph model of computation in which each step is a small, local rewrite, independent of the others. Nothing is shared, so nothing needs to be kept in sync, and work that can run in parallel does so, with a simpler core that can scale to thousands.
Alongside the hardware, Tendrils is working on Vine, an open-source programming language designed from the beginning to target parallel-first architectures, with the two developed in tandem. Vine, built by co-founder T6, reads like Rust and compiles straight to the graph. An ordinary nested loop can expose hundreds of operations ready to run at every step, without threads or CUDA. In summer 2025, Visionaries Club led a pre-seed round for Tendrils, alongside FR8 and angel investors, after first supporting the team with a small grant. Jack Altman has joined the board.
Nils, T6 and Hugo are the co-founders. Nils came across interaction nets in 2024 while studying computer science at ETH Zurich, and was finishing his master's thesis when Visionaries Club first met him in Munich. T6, who built Vine, has also contributed to Parity. Hugo previously worked at lowRISC, an independent company out of Cambridge working on open-source chip architectures.
If it works, this isn't a faster chip. It's a reset of general-purpose computing, and the substrate the next wave of AI will run on.
