On September 16, media reports revealed that ByteDance is working on two self-developed AI chips to reduce its reliance on costly NVIDIA processors and strengthen its competitive position in the AI market. According to insiders, ByteDance is collaborating with TSMC (Taiwan Semiconductor Manufacturing Company) to achieve mass production of these chips by 2026.
The two chips are designed to target different aspects of AI technology: one is a training chip focused on handling the massive data computations required for building AI models, while the other is an inference chip optimized for accelerating model-based predictions and decision-making processes.
By developing its own AI hardware, ByteDance aims to cut costs significantly, with estimates suggesting savings of billions of dollars. The company has reportedly placed an order with TSMC for several hundred thousand units of these chips, signaling its strong commitment to integrating in-house AI solutions into its operations.
This move highlights a broader trend among tech giants to invest in proprietary AI hardware to gain greater control over their technology stacks and reduce dependence on external suppliers like NVIDIA.
All Comments (0)