Tendrils Raises $30 Million to Rebuild the CPU From Scratch
Tendrils raised $30 million led by Benchmark to build a thousands-core parallel processor and Vine, an open-source language replacing 1970s CPU design assumptions.
By Amara Osei
3 min read

What's News
- Tendrils raised $30 million in a round led by Benchmark.
- Jack Altman joined Tendrils' board following the round.
- The startup is designing a processor with thousands of parallel cores.
- Tendrils is building Vine, an open-source parallel-first programming language.
- The company targets general-purpose computing, not just AI workloads.
Tendrils has raised $30 million to build a processor architecture with thousands of parallel computing cores — and to throw out five decades of CPU design convention in the process.
Benchmark led the round. Visionaries Club, the UK Sovereign AI Fund, FR8, Antimatter, Cloudberry VC, Conviction, Embassy Ventures and SV Angel also participated, alongside a group of angel investors. Jack Altman has joined the company's board as part of the deal.
The startup's core argument is blunt: modern CPUs still carry architectural decisions made in the 1970s. Cache-coherency protocols, reorder buffers, shared mutable memory and linear instruction sequences all create complexity, Tendrils says, and that complexity limits how efficiently processors can scale.
Why start over instead of iterating?
Conventional processors are constrained by backward compatibility. Sequential, single-core design assumptions are baked into nearly every general-purpose chip shipped today. Rather than patch existing architectures, Tendrils is designing a processor intended to run thousands of cores in parallel from the outset.
That decision forces a second one: hardware and software have to be developed together. You can't run a massively parallel chip well on languages and tools built for sequential machines. So Tendrils is also building Vine, an open-source programming language designed from the ground up for parallel-first architectures.
The company's bet is that tighter coordination between language design and processor architecture can eliminate legacy requirements entirely — and unlock more scalable general-purpose computing than incremental redesigns ever could.
How does the $30 million get spent?
The funding will support the research, engineering and development needed to turn the concept into functioning hardware and software. That's a long road: designing a novel CPU architecture and a new programming language simultaneously is one of the heavier lifts in deep tech.
Tendrils is also actively recruiting. The company says it's hiring hardware and software engineers interested in building the architecture from first principles.
Is this just another AI accelerator play?
No — and the distinction matters. AI accelerators have already proved the value of highly parallel computing for specialized workloads like model training and inference. Tendrils is targeting general-purpose applications instead.
The company's objective is an architecture that supports a wide range of software without relying on the sequential design assumptions embedded in conventional processors. In other words, it wants parallelism to be the default for everyday computing, not a specialized tool for machine learning.
That's a harder commercial problem than the accelerator market. General-purpose computing means compilers, operating systems, runtime environments and decades of software ecosystems — all of which assume sequential execution to some degree. Vine exists precisely because that gap can't be bridged with legacy languages.
What happens next?
The round gives Tendrils the runway to move from architectural theory to working silicon and a usable language stack. Benchmark's lead, alongside a syndicate that includes the UK Sovereign AI Fund, signals investor appetite for infrastructure-level bets on compute — not just applications on top of it. The test ahead is execution: whether thousands of parallel cores plus a purpose-built language can deliver general-purpose performance where five decades of sequential architecture still dominate.
By Amara Osei
Source: pulse2.com
More from Amara Osei
Show full bio
Senior reporter covering companies, deals and strategy in entrepreneurship, startups and venture funding at Business Bearings.
2,136 articles