👩🍳 How we use AI at Tech in Asia, thoughtfully and responsibly.
🧔♂️ A friendly human may check it before it goes live. More news here
🧔♂️ A friendly human may check it before it goes live. More news here
Intel, Qualcomm held sale talks with AI startup Tenstorrent
Tenstorrent is a Santa Clara-based AI chip startup and it has held early sale talks with Intel and Qualcomm while also exploring a new funding round.
The company has also spoken with investment banks about its options and a deal could value it at more than US$5 billion.
The interest reflects a wider scramble for AI chip technology outside Nvidia and AMD, and Arm and SoftBank approached Cerebras before its recent IPO.
🔗 Source: Bloomberg
🧠 Food for thought
Implications, context, and why it matters.
Tenstorrent builds value through open standards, systems, and IP
- Tenstorrent sells AI accelerator cards, full systems, and intellectual property (IP) licensing alongside hardware 1.
- It aims at customers seeking less vendor lock-in, including sovereign AI efforts and automotive original equipment manufacturers (OEMs) such as Hyundai, which can place Tenstorrent IP in custom chips 1.
- The company skips the costly High Bandwidth Memory (HBM) common in many leading graphics processing units (GPUs). It uses GDDR6 with large on-chip static random-access memory (SRAM) to cut total cost of ownership 2.
- Another part of the valuation comes from high-performance RISC-V central processing units (CPUs), an open standard alternative to proprietary CPU instruction sets such as Arm, which Tenstorrent also plans to license 3, 4.
Buyers may want Tenstorrent’s IP while software slows adoption
- Intel or Qualcomm may value Tenstorrent for its architecture and IP rather than as a direct rival. The cited sources do not back the draft’s claim that Tenstorrent is modular or chiplet-based 4.
- That technology could feed future products in automotive or custom data center systems 1.
- Software remains the hard part. For hyperscalers, adopting a new instruction set architecture can cost about 100x the hardware spend because widely used software stacks must be ported 5.
- That helps explain Nvidia’s CUDA software platform dominance. It also means any deal would likely require years of integration and heavy spending before it could pressure the market leader 2.
Recent Tenstorrent developments
Stay updated on the go with our mobile app.
Get latest insights with smoother, more personalized experience through TIA mobile app.




