Tenstorrent, led by chip architect Jim Keller, is the open option: RISC-V control cores, an open-source software stack and hardware sold at every size from a single card to a supercluster. It also licenses its AI and CPU designs to companies such as Samsung and LG. In October 2026 it raised $1.5 billion at a $6.2 billion valuation.
Key Takeaways
- What it is — Open hardware. Grid of Tensix cores with RISC-V control processors; designs also licensed to other chipmakers.
- Best for — Teams that want to own and modify the full stack, and buyers too small to get Nvidia allocation.
- Watch out — Lower memory bandwidth than HBM parts, and production scale is still being proven.
At a glance
| Company | Tenstorrent |
|---|---|
| Current chips | Wormhole and Blackhole accelerators, sold in Galaxy servers |
| Architecture | Grid of Tensix cores with RISC-V control processors; designs also licensed to other chipmakers |
| Memory | GDDR-based accelerators, which keeps cost down at the expense of HBM-class bandwidth |
| Scale-up | Galaxy servers scale from single cards and workstations to racks and superclusters over Ethernet |
| Software | Open-source stack (TT-Metalium, TT-Forge) running open-weight models |
| Where to get it | Direct hardware sales and clouds such as Cirrascale |
| Pricing | Published list prices for cards, workstations and Galaxy servers; cloud pricing through partners |
What changed in 2026
- Oct 2026. Tenstorrent raises a $1.5 billion Series E at a $6.2 billion valuation to scale Galaxy servers.
Scores
Scored 1 to 5 on the same five axes as every vendor in the AI chip companies analysis.
| Performance | Scale-up | Software | Access | 2026 momentum | Total / 25 |
|---|---|---|---|---|---|
| 3 | 3 | 3 | 4 | 4 | 17 |
Verdict
The choice for builders who value an open stack and ownership over peak performance. Pilot it on a workload you can benchmark end to end.
Alternatives
In other lanes:
- Google TPU (hyperscaler custom silicon)
- AWS Trainium (hyperscaler custom silicon)
- AMD Instinct (merchant gpu)
- Cerebras (inference specialist)
- Groq (inference specialist)
- SambaNova (inference specialist)
- Microsoft Maia (hyperscaler custom silicon)
Head-to-head comparisons
- Google TPU vs AWS Trainium vs Nvidia GPU: The three ways to train and serve frontier-scale models in 2026, scored on scale-up domain, software, access and price.
- Cerebras vs Groq vs SambaNova: The SRAM-first inference specialists compared on speed, token price, deployment options and company risk.
New to custom AI silicon? Our partner site CTAIO explains how TPUs, pods and GPU clusters differ in TPU vs GPU.
Frequently asked questions
What is Tenstorrent?
Tenstorrent, led by chip architect Jim Keller, is the open option: RISC-V control cores, an open-source software stack and hardware sold at every size from a single card to a supercluster. It also licenses its AI and CPU designs to companies such as Samsung and LG. In October 2026 it raised $1.5 billion at a $6.2 billion valuation.
How do you get access to Tenstorrent?
Direct hardware sales and clouds such as Cirrascale. Pricing: Published list prices for cards, workstations and Galaxy servers; cloud pricing through partners.
What software does Tenstorrent run?
Open-source stack (TT-Metalium, TT-Forge) running open-weight models.
Is Tenstorrent a good alternative to Nvidia?
The choice for builders who value an open stack and ownership over peak performance. Pilot it on a workload you can benchmark end to end. It fits best for: teams that want to own and modify the full stack, and buyers too small to get Nvidia allocation. The main caveat: lower memory bandwidth than HBM parts, and production scale is still being proven.
Sources
Facts verified 2026-10-31. Vendor performance figures are as stated by the vendor.
Ready to Find the Right AI Tools?
Browse our data-driven rankings to find the best AI tools for your team.