Tendrils raises $30M led by Benchmark to build a many-core CPU
Tendrils announced on its own blog that it has completed a $30M round led by Benchmark. The stated aim is to rebuild general-purpose computing around a processor with thousands of cores working in parallel.
The company's argument is that today's CPUs are held back by history. In its words, contemporary CPUs are made under the constraint of backwards compatibility with sequential, single-core chips, and accumulated layers of historical design lock engineers into decisions dating back to the 1970s. Tendrils says the result is that all existing CPUs are bottlenecked by complex cache coherency protocols and elaborate reorder buffers, and that CPU scaling is sabotaged from the start. It believes the future of general-purpose computing lies beyond legacy architectures for legacy software.
The company frames its work as a single question: what if we redo the CPU? Instead of starting from a sequential design, it says it is building its processor to have thousands of cores operating in parallel. That, it says, requires moving beyond shared mutable memory and linear instruction sequences. To make it possible, Tendrils is working in tandem on Vine, an open-source language designed from the beginning to target a new wave of parallel-first architectures.
On the money: Tendrils says it has raised $30 million. Benchmark leads the round, with significant participation from Visionaries Club, the UK Sovereign AI Fund, FR8, Antimatter, Cloudberry VC, Conviction, Embassy Ventures and SV Angels. The post also thanks its angel investors and welcomes Jack Altman as its newest board member.
The post closes with a recruiting line. Tendrils insists that the chips of the future will result from close collaboration between hardware and software engineers, and invites readers who agree to join the team. The post is signed "The Tendrils Team".
Key facts
- Tendrils has completed a $30M round led by Benchmark, announced on the company's own blog.
- Other named investors with significant participation: Visionaries Club, the UK Sovereign AI Fund, FR8, Antimatter, Cloudberry VC, Conviction, Embassy Ventures and SV Angels.
- Tendrils says it is building a general-purpose processor with thousands of cores operating in parallel, rather than a sequential design.
- It is developing the open-source language Vine alongside the chip, designed to target parallel-first architectures.
- Jack Altman joins as the newest board member, according to the post.
Why it matters
The post is a bet against the shape of the mainstream CPU. Tendrils argues that backwards compatibility with sequential, single-core chips, with design decisions dating back to the 1970s, leaves all existing CPUs bottlenecked by cache coherency protocols and reorder buffers. Its answer is a clean-sheet processor with thousands of parallel cores, which it says means dropping shared mutable memory and linear instruction sequences. A $30M round led by Benchmark gives that idea funding, but the argument itself is the company's own.
Who it affects
The post speaks to hardware and software engineers, since Tendrils says future chips will come from close collaboration between the two and invites readers to join. Developers who write for parallel machines are the natural audience for Vine, the open-source language built for parallel-first architectures. The investors named in the round are Benchmark (lead), Visionaries Club, the UK Sovereign AI Fund, FR8, Antimatter, Cloudberry VC, Conviction, Embassy Ventures and SV Angels.
How to use it
There is nothing to buy or run yet. The only action the post offers is to consider joining Tendrils if its view of hardware and software collaboration appeals. Vine is described as open source, but the post gives no release status or link for it.
How solid is it
The funding claim comes from Tendrils itself, in a signed company blog post, and names Benchmark as lead investor. The technical claims, such as all existing CPUs being bottlenecked and scaling being sabotaged from the start, are the company's own framing. The post gives no valuation, no round stage, no individual investment amounts, and no prototype, tape-out, benchmark results or performance figures.
Risks and caveats
The processor is described as something Tendrils is building, not a finished chip, and the post offers no timeline. It does not describe the fabrication process, instruction set, memory model or core design, so the thousands-of-cores goal cannot be judged from this text. The post does not say what Jack Altman's affiliation is, or whether he joins through Benchmark or another investor. It also names no founders, executives or headcount.
“what if we redo the CPU?”
— Tendrils, company blog post