Tenstorrent’s $1.5 billion round values the AI chip startup at $6.2 billion

AI chip company Tenstorrent has closed a $1.5 billion Series E round at a $6.2 billion post-money valuation, according to its announcement on October 19, 2026, which Reuters reported as more than double its previous valuation.

Who backed the round and where the money goes

Trace Capital and Innovation X led the round. According to a report on the company’s press release, other participants include ARK Invest, Fidelity Management & Research Company, Morgan Stanley, Sandisk, Tether Investments, Virtu Financial and XTX Markets, with T. Rowe Price taking part through advised accounts. The same report says Trace Capital was founded in 2026 by a former co-founder of Quiet Capital, and that Innovation X dates from 2016 and reports more than $4.5 billion under management.

Tenstorrent says it will use the funds for three things: deploying its Galaxy servers at scale, continuing to serve open-weight models for large language models and video generation, and advancing the AI and CPU intellectual property it licenses to chip designers.

The company describes a business that sells AI compute at every level, from a single card or workstation to racks and superclusters, alongside licensable IP. It says it licenses its TT-Ascalon RISC-V CPU and Tensix AI cores to chip designers, naming Samsung and LG among them, and offers an open-source software stack so developers can run models locally and enterprises can deploy them on-premises. CEO Jim Keller summed up the pitch in one line: “We built one architecture and sell it at every size.”

A hardware, software and licensing mix rather than a single chip

Tenstorrent is not pitching a single GPU rival. It sells finished systems, software and licenses from the same design. That spreads risk: a customer who does not want its servers can still buy cores or a CPU design. Keller previously worked on Apple’s A4 and A5 chips, AMD’s Zen and Tesla’s self-driving chip, according to the company’s background section.

The Galaxy line is the hardware piece. Listings tracking the company say the Galaxy Blackhole system became generally available on April 28, 2026, and that cloud provider Cirrascale now offers Galaxy servers, with claims of roughly half the cost of GPU alternatives. Those cost figures come from the vendors and have not been independently measured.

For readers following the market, the context is Nvidia, which crossed a $5 trillion valuation in 2025. Even at $6.2 billion, Tenstorrent is a small fraction of that scale, and investors are paying for the chance that open software and licensable designs find a place next to it.

Strategy: Staying independent while capital is plentiful

The round says something about how Tenstorrent wants to grow. Bloomberg reported that the company chose to remain independent after preliminary takeover discussions earlier this year with Intel and Qualcomm. Neither company has publicly confirmed such talks in the material reviewed here, so this remains the outlet’s reporting. Separately, one tracking site cites a June report of a possible Qualcomm approach at a much higher price, which is unconfirmed.

The jump is also steep against the last round. Tenstorrent’s Series D in December 2024 raised more than $693 million at a $2 billion pre-money valuation, led by Samsung Securities and AFW Partners, according to trackers of the deal. Reuters says the new figure is more than double what came before. Bloomberg, as relayed by Investing.com, cites a different prior valuation of $2.6 billion, so the exact multiple depends on which number is used.

The investor list is mixed. It combines venture and growth funds with large asset managers and trading firms, and the company names Samsung, LG Electronics, Hyundai Motor Group and Bezos Expeditions among its earlier backers. Several of those are also customers or licensees, which ties the funding to demand for its IP.

What the announcement does not establish

The release does not give revenue, shipment numbers or customer counts, so the valuation cannot be checked against sales. Its claims about performance, including video generation results that one investor said beat industry benchmarks, are not independently verified. Reuters and Bloomberg also differ on where the company is based (Santa Clara versus Toronto); the company lists offices in Santa Clara, Austin, Toronto, Belgrade, Tokyo and Bangalore.

Signs to watch are whether Galaxy deployments grow beyond early cloud listings, whether more chip designers sign for the CPU and AI core licenses, and whether the company publishes independent benchmark results for its open-weight model serving. Another chip-software move this week came from Amazon, which says PyTorch code now runs on Trainium with one setting changed, a reminder that software compatibility is where much of the competition against Nvidia is fought.