Tenstorrent Raises $1.5 Billion to Scale Open-Weight AI Chips
Tenstorrent has raised $1.5 billion in Series E equity financing at a $6.2 billion post-money valuation. The company announced the financing on October 19, 2026. The round was led by Trace Capital and Innovation X. TechCrunch
Other backers included Cadence Ventures, Eclipse, Fidelity, SanDisk and T. Rowe Price. Trace Capital is a new firm launched by Quiet Capital co-founder Ben Mahdavi. The company is led by Chief Executive Jim Keller.
The $6.2 billion figure is post-money, meaning the value after the new money is counted. It is more than double the $2.6 billion valuation reported in late 2024. Reuters Five months before October 2026, Intel and Qualcomm had shown acquisition interest at valuations of as much as $5 billion.
Tenstorrent chose to remain independent and raise additional capital. That decision followed a demonstration in which its Galaxy servers ran DeepSeek models.
The company said the equity raise will fund volume manufacturing and global commercial expansion. Tenstorrent It plans to use the new capital to scale chips tailored for serving open-weight AI models — models whose weights are publicly available to download and run — and video generation.
The broader context here is a shift in where chip startups can compete. Training clusters still get most headlines and spending, but serving open-weight models at scale is a different engineering problem. It rewards memory bandwidth, or how fast data reaches the chip, interconnect efficiency, or how well chips work together, deterministic throughput, or steady and predictable speed, and cost per token rather than peak FLOPS alone, a measure of maximum math speed. Video generation adds parallel pressure, with long sequence lengths and sustained inference load, or continuous work answering requests. For a tech-literate buyer, the relevant questions are qualification, software maturity and supply continuity.
Looking at what this means for buyers and operators, volume manufacturing is the central test. Tape-outs, the first finished chip designs sent to a factory, and single-system demonstrations establish architecture. Volume establishes yields, firmware maturity, toolchain stability and support. Global commercial expansion implies data-center qualification cycles, systems integration work and sustained inference uptime. Scale is the test. Those are operational challenges as much as silicon challenges, and they determine whether an alternative accelerator can move from evaluation to production deployment.
In my view, the decision to decline acquisition interest in favor of a large private round fits that operational need. A sale at up to $5 billion would have transferred manufacturing and go-to-market risk to a larger buyer. A $1.5 billion raise preserves independence but requires Tenstorrent to solve distribution, support and software continuity itself. If execution holds, the upside is practical. More suppliers for open-weight serving would give operators leverage on price, availability and deployment flexibility, and add useful capacity in a market that continues to need it.
