$30M Raised to Redo the CPU
Tendrils secured $30M from Benchmark and others to revolutionize general-purpose computing, proposing a new CPU architecture with thousands of parallel cores and a bespoke language, Vine, to escape the shackles of x86 legacy. This audacious undertaking, backed by significant VC, immediately ignited the HN community's skepticism. Debates rage over its feasibility, comparisons to past failed ventures, and the perennial question of whether hardware or software truly bottlenecks progress.
The Lowdown
Tendrils has announced a $30 million funding round, led by Benchmark, to embark on an ambitious mission: to fundamentally redesign the CPU. The company posits that contemporary CPUs are hobbled by decades of backward compatibility constraints, leading to inefficiencies like complex cache coherency protocols and elaborate reorder buffers. They argue that true progress requires moving beyond these legacy architectures.
Tendrils' approach is radical, focusing on a parallel-first design from the ground up:
- Thousands of Parallel Cores: Instead of a sequential design, their new processor will feature an immense number of cores operating simultaneously.
- Departure from Traditional Memory: This architecture necessitates moving beyond the conventional notions of shared mutable memory and linear instruction sequences.
- Vine Language: To facilitate this paradigm shift, Tendrils is developing an open-source language called Vine, engineered specifically for these new parallel-first architectures.
- Significant Funding: The $30M round includes investments from Benchmark, Visionaries Club, UK Sovereign AI Fund, FR8, Antimatter, Cloudberry VC, Conviction, Embassy Ventures, and SV Angels, alongside angel investors. Jack Altman has also joined their board.
By tightly integrating hardware and software development, Tendrils aims to carve a new path for computing, believing that a complete overhaul of the CPU is essential for future innovation.
The Gossip
Ghost of CPUs Past
Many commenters expressed deep skepticism, immediately recalling a long list of past ambitious CPU projects that failed to dislodge x86 or gain significant traction. Names like Mill, Transputer, Itanium, and Transmeta were frequently cited as cautionary tales, highlighting the immense difficulty and historical failures in attempting to 'redo' the CPU and the software ecosystem challenges. The prevailing sentiment is that history offers little hope for such a monumental shift.
Parallel Processing Puzzles
The technical core of Tendrils' vision—thousands of parallel cores and moving beyond shared mutable memory—sparked questions about its novelty and practicality. Several users pointed out that this description sounds strikingly similar to a GPU, leading to inquiries about how Tendrils' offering would differentiate itself or if it's simply a repackaged GPU for general-purpose use. Others questioned the viability of a new language (Vine) and the compiler challenges, noting that software adoption is historically the hardest part of any new hardware architecture.
The VC Velocity
A significant thread of discussion revolved around the $30M funding round itself, with many commenters expressing cynicism about venture capital's role in such ambitious, unproven hardware ventures. Some suggested that the funding might be more about hype and 'laundering money' than a solid technical vision, especially given the lack of a demonstrable product. The mention of a potential fast follow-up raise before any tangible results further fueled this perception of VC-driven speculation over substantive innovation.