Toronto-based AI chip designer Tenstorrent, led by veteran architect Jim Keller, has raised $1.5 billion in a Series E round at a post-money valuation of $6.2 billion. The round more than doubles the $2.6 billion valuation set in its December 2024 Series D and gives the company fresh capital to scale its open compute ecosystem against Nvidia and AMD.
🔑 Key takeaways
- Series E of $1.5 billion announced October 19, 2026, co-led by Trace Capital and Innovation X.
- Post-money valuation of $6.2 billion, up from $2.6 billion at the December 2024 Series D.
- Full-stack offering: Galaxy servers, Tensix Neo accelerators, Ascalon CPUs and open-source software on the RISC-V instruction set.
- Headcount of 1,150 employees, after talks with Intel and Qualcomm about a possible acquisition ended with the company staying independent.
A round that doubles the valuation in under two years
Announced on October 19, 2026, Tenstorrent’s Series E lifts its post-money valuation to $6.2 billion. The previous round, a $700 million Series D closed in December 2024, had set the valuation at $2.6 billion. That is a step-up of roughly 2.4x in a little under two years, a pace that reflects investor appetite for alternatives to the accelerators sold by Nvidia and AMD.
| Round | Date | Amount raised | Post-money valuation |
|---|---|---|---|
| Series D | December 2024 | $700M | $2.6B |
| Series E | October 2026 | $1.5B | $6.2B |
The round is co-led by Trace Capital, a firm founded in 2026 by Ben Mahdavi, formerly of Quiet Capital, and by Innovation X, a late-stage fund created in 2016 that manages more than $4.5 billion in assets. Existing and new investors include Agostinelli Group, ARK Invest, Brett Barna, Cadence Ventures, Eclipse, Fidelity Management & Research Company, Maverick Silicon, Morgan Stanley, Protagonist, Sandisk, T. Rowe Price, Tether Investments, Troluce Capital, Virtu Financial and XTX Markets. The mix of asset managers, investment banks, market makers and a storage industry player suggests that capital is looking beyond the best-known names in AI hardware.
Keller framed the raise as validation of the model: “The future of compute is open, scalable and owned by the people who build on it.” He added that investors share that view and that the company sells its architecture at every scale, from workstations to superclusters.
Ben Mahdavi, founder of Trace Capital, pointed to the team’s decades of practical experience across silicon, software and intellectual property, and to the company’s continued growth.
A full platform, from Galaxy servers to RISC-V cores
Tenstorrent positions itself as an alternative to Nvidia and AMD in AI chips, selling a complete stack. Its enterprise Galaxy servers can be linked together to form superclusters for training and running frontier models, while liquid-cooled TT-QuietBox workstations target developers.
These platforms are powered by Tensix Neo accelerators and Ascalon central processors, both built on the open-source RISC-V instruction set architecture, which lets licensees adapt designs without paying a fee for the instruction set itself.
A stress test with DeepSeek models
Speaking to Bloomberg, Keller identified a May demonstration as a turning point. Galaxy servers ran high-performance DeepSeek models, a result the company presents as proof that its hardware can serve leading open-weight models.
“Few companies design both the chip and the software that runs on it. Tenstorrent does both, and its video generation results are beyond what we expected from the industry benchmarks.”
George Bousis, Co-founder and Managing Partner, Protagonist
Protagonist, an investor in the round, said it had doubled its stake in the Series E. The comment underlines the company’s central argument: an integrated design of chip and software, judged on real workloads rather than on spec sheets.
An open software stack as a commercial argument
Hardware alone is not enough to compete with established ecosystems. Tenstorrent has therefore built an open-source software stack, whose main components are:
- TT-Forge: a compiler that connects machine learning frameworks such as PyTorch to its RISC-V chips.
- TT-Metalium: a software development kit for hardware optimisation and custom kernel development.
- TT-LLK: low-level kernel optimisations.
The approach already has commercial uses. According to the company, Virtu Financial runs AI trading agents on-premises on Galaxy servers, while colocation provider Equinix makes the systems available to enterprise customers.
Tenstorrent also licenses its Tensix AI cores and its TT-Ascalon RISC-V CPU to third-party chip designers, including Samsung and LG Electronics. This model extends the reach of the architecture without requiring the company to manufacture every end product.
“Tenstorrent is getting the funding it needs to keep innovating, and that is good news for enterprises. Nvidia is no longer the only option.”
Holger Mueller, Analyst, Constellation Research
Funding, headcount and competition: the near-term strategy
Series E proceeds will go to four priorities: deploying Galaxy servers at scale, continuing to serve open-weight models including large language and video generation workloads, advancing the AI and CPU IP licensed to chip designers, and expanding headcount globally.
Headquartered in Toronto, the company has offices in Santa Clara, Austin, Belgrade, Tokyo and Bangalore, and employs 1,150 people according to Bloomberg. Its founder is a well-known semiconductor architect who contributed to Apple’s A4 and A5 chips, AMD’s Zen architecture and Tesla’s full self-driving chip.
Earlier in 2026, Tenstorrent held preliminary talks with Intel and Qualcomm about a possible acquisition before deciding to remain independent. Keller says the company can compete on price and accept orders too small for larger rivals to pursue, a positioning that could open underserved segments of the market.
Several risks remain. Galaxy superclusters must prove they can deliver beyond demonstrations, the software stack must attract enough developers to challenge entrenched environments, and licensing partnerships must translate into meaningful volumes.
Conclusion: three scenarios for the months ahead
The $1.5 billion raise gives Tenstorrent the means to move from technical demonstration to industrial deployment. The base case assumes a gradual ramp-up of Galaxy servers among enterprise customers and licensing partners, with headcount growth matched to development needs.
- Upside scenario: significant supercluster orders and new licensing partners confirm the relevance of an open approach against Nvidia.
- Central scenario: the company consolidates its niche customers and partnerships without changing the balance of the AI accelerator market.
- Downside scenario: production delays or a thin software ecosystem limit growth, exposing the $6.2 billion valuation to a reassessment.
In every case, Tenstorrent’s trajectory will be a useful gauge of whether open architectures built on RISC-V can win ground against proprietary platforms in AI compute.
Sources
- SiliconANGLE, October 19, 2026
- Techmeme, October 19, 2026
- Tenstorrent press release, Series E
- AI Weekly, Series E
- Yahoo Finance, Series E
- The Next Web, Series E
Cet article est publié à titre informatif et éducatif. Il ne constitue en aucun cas un conseil en investissement. Faites vos propres recherches (DYOR) avant toute décision.
