DeepSeek builds its own semiconductor for AI inference
Chinese AI company DeepSeek is currently working on the development of its own semiconductor, according to Reuters. Sources indicate that this hardware is specifically designed for inference, the process in which a pre-trained model interprets user input and provides answers.
DeepSeek is joining a broader trend among leading AI organisations to produce specialised hardware in order to optimise performance and efficiency. Similar initiatives are widespread. OpenAI recently introduced a custom chip developed together with Broadcom, and Alphabet continues to see strong demand for its Tensor processing units (TPUs).
In addition, Amazon is exploring the possibility of selling its own custom chips to external data centres.
DeepSeek plays a crucial role in China’s strategy to achieve international leadership in the field of AI. The company previously drew global attention by developing a high-quality model that was able to function with remarkably little computing power.
That achievement illustrated the ability of Chinese technology companies to compete with Silicon Valley, even despite strict US restrictions on the import of high-end hardware.
Business AM
Subscribers 0
Fans 0
Followers 0
Followers