Venture firm Benchmark led funding for Tendrils Compute, an early-stage AI chip startup focused on inference acceleration.

Benchmark has led a funding round for early-stage chip startup Tendrils Compute, according to multiple sources familiar with the matter. The company is already discussing a fast follow-up raise that would value the firm at more than $1 billion. Neither round has been previously reported. Benchmark declined to comment.

The Tendrils deal marks the latest in a wave of major venture capital bets on chip startups, representing a revival of the sector that gave Silicon Valley its name but has long been overlooked as investors focused on software and networking. Startups are betting that specialized chips designed for specific workloads will play a critical role in infrastructure expansion—a departure from previous chip generations dominated by a small handful of manufacturers.

The small Tendrils team, led by Nils Cremer and based in Cambridge, U.K., is pursuing a distinctive approach. The company is developing general-purpose chips based on interaction nets, an obscure graph-based computing model, betting that this architecture will enable scaling without the coordination and cost challenges associated with contemporary multi-core CPUs. As AI agents execute multiple steps of inference and tool calls, CPU speed may become a new bottleneck—a problem Tendrils aims to solve.

Benchmark has positioned itself at the forefront of the renewed chip enthusiasm. In 2016, Eric Vishria co-led the Series A round in Cerebras Systems, achieving a 20.5x multiple on the investment when the company went public in spring 2026. It was Benchmark's first hardware investment in a decade, and Vishria nearly skipped the meeting, he told TechCrunch. That success, combined with relentless demand for AI compute, has convinced venture capitalists that chip startups offer substantial returns.

Recent funding activity demonstrates this enthusiasm. In August, Etched, which builds specialized chips and racks for AI inference, doubled its valuation to $21 billion in a $700 million fundraise. Fractile, a fellow inference-focused chipmaker, is in talks to raise at a $6.5 billion valuation following a deal to supply chips to Anthropic. OLIX, a U.K.-based inference chip startup, raised $312 million at a $3.3 billion valuation in August. Two other companies announced nine-figure rounds in September: inference chip startup Euclyd and Delos Data, which builds software and networking equipment to help information travel more quickly around data centers. Volantis Semiconductors raised $88 million from Lachy Groom and Abstract Ventures for another inference-focused chip platform.

"They have a massive tailwind from capex build-out," said Darian Shirazi, managing partner at Gradient. "They probably will all do well."

Yet chip development remains notoriously capital-intensive and cyclical, with long development timelines and substantial risks. Cerebras went public in May at $185, traded as high as $386, but now sits close to its offering price. Venture capitalists have adopted creative approaches to distribute risk and cost. Coatue and chipmaker MatX are in talks to form a joint venture worth billions of dollars to help the company finance memory purchases and manufacturing capacity, according to The Information.

Investors acknowledge the emerging chip ecosystem is noisy and difficult to parse. The sector spans three distinct types of chip startups, each addressing different aspects of the infrastructure buildout that venture capitalists are increasingly willing to fund.