This article covers Tendrils, a hardware startup, which has raised £22.4m in a growth funding round led by Benchmark. The funding will support development of a parallel-first processor architecture and an open-source language, Vine, which Tendrils says are intended to target large-scale parallel and high-throughput workloads and to influence tooling and ecosystem support.
Tendrils, a hardware startup, has raised £22.4m in a growth funding round led by Benchmark to pursue a controversial rethink of the CPU. The company says it will use the capital to build a parallel-first processor architecture and an accompanying open-source language, Vine, aiming to move beyond decades of backward-compatible CPU design that the team argues constrains performance at scale.
Modern general-purpose processors retain design compromises made to preserve backwards compatibility with single-core, sequential code. Tendrils positions its work as a direct response to those limits: instead of incrementally extending existing CPU designs, the company proposes an architecture built around many thousands of cores and a programming model that avoids shared mutable memory. If realised, that approach could change how performance and concurrency are exposed to software, with potential implications for high-throughput and parallel workloads such as large-scale data processing and some classes of machine learning.
The stakes are high. Redesigning the processor-software boundary can unlock efficiency gains, but it also raises practical questions about software migration, tooling and ecosystem support. The funding round signals investor appetite for hardware-first experiments even as such projects remain capital intensive and technically risky.
Tendrils describes two tightly coupled elements to its approach. On the hardware side, the company is building a processor that deliberately abandons the single-core, sequential model in favour of massive parallelism. That choice, Tendrils says, removes the need for complex cache coherency protocols and large reorder buffers that, in their view, limit scaling in current CPUs.
On the software side, Tendrils is developing Vine, an open-source language designed from the outset for parallel-first architectures. The language aims to express concurrency without relying on shared mutable state and linear instruction sequences, which Tendrils argues are mismatched to a multi-thousand-core design. The company has framed its work as hardware and software co-design, emphasising that the chips of the future require collaboration between both disciplines.
Key open questions remain: how legacy applications will be supported or ported, what performance characteristics will look like across diverse workloads, and how the tooling and compiler ecosystem will mature. Tendrils has not published benchmarks in this announcement.
The round was led by Benchmark, with participation from Visionaries Club, the UK Sovereign AI Fund, FR8, Antimatter, Cloudberry VC, Conviction, Embassy Ventures, SV Angels and a group of angel investors. The company also announced Jack Altman as a new board member.
Benchmark is a well-known venture firm that typically backs early-stage technology companies; its lead role gives the round a high-profile anchor. Visionaries Club is an angel network that frequently backs founders in deeptech and infrastructure. The presence of the UK Sovereign AI Fund suggests an element of public or strategic interest in projects that could underpin future AI infrastructure. Smaller VCs and angels in the syndicate bring a mix of early-stage capital and domain expertise that Tendrils will likely rely on during the hardware development cycle.
If you're researching potential backers in this space:
The company framed its effort as an intentional break from legacy assumptions. Tendrils argues that starting from a parallel-first hardware model requires rethinking programming languages and runtime systems, which is why the team is developing Vine in parallel with the processor. The announcement also included an invitation to potential hires, signalling that the firm is still building core engineering capabilities.
Tendrils’s raise comes at a time when investors and governments are taking renewed interest in foundational compute infrastructure. Domain-specific chips and software-hardware co-design have gained traction as cloud-scale AI and data workloads push traditional architectures. Yet hardware startups face long development timelines, high capital needs and the challenge of building an ecosystem of tools and software that can run efficiently on new architectures.
This deal underlines that appetite for deeptech hardware remains, particularly where the pitch intersects with AI and parallel computing. For the UK and European ecosystem, participation from the UK Sovereign AI Fund and local VCs highlights continued public and private interest in backing infrastructure projects that could underpin future competitiveness in compute and AI.
| Investors | Investment Focus | Startup Investments | Round Size | Connect |
|---|---|---|---|---|
![]() Benchmark Capital( ) San Francisco, US | Generalist | |||
![]() Visionaries Club( ) | ||||
![]() UK Sovereign AI Fund( ) | ||||
![]() Convicton VC( ) NYC, US | Generalist | |||
| All investors | All investor sectors | All funded startups | All funding rounds |
Click here for a full list of 7,589+ startup investors in the UK