tech
DeepSeek and Zhipu AI Begin Self-Developing AI Chips
Want to read in a language you're more familiar with?
TL;DR
- DeepSeek and Zhipu AI are developing custom AI chips for inference, joining OpenAI and Anthropic.
- This move aims to reduce costs associated with continuous inference and GPU dependency.
- For Chinese firms, in-house chip development also mitigates risks from US export controls on advanced GPUs.
- Chip development is capital-intensive, with estimated design costs around $500 million and no guarantee of success.
- The industry is seeing a bifurcation of training and inference hardware, with specialized chips for each function.