ByteDance plans to develop its own CPU to deal with the rising cost of chips and the demand for computing power.

date
15:24 28/05/2026
avatar
GMT Eight
According to reports, ByteDance is developing its own central processing unit (CPU) to address its growing demand for artificial intelligence infrastructure.
According to reports, ByteDance is developing its own central processing unit (CPU) to address its growing demand for artificial intelligence infrastructure. The report indicates that the continuous increase in chip prices and long-term supply shortages have limited the company's expansion plans. Sources say that the company is exploring two chip architecture solutions for its self-developed CPU, one based on SoftBank's Arm architecture and the other using the open-source RISC-V architecture. Currently, both ByteDance and Arm have not commented on the matter. Sources reveal that ByteDance is currently purchasing CPUs from Intel and AMD, but both suppliers have significantly raised their prices, with quarterly increases ranging between 10% and 35% in recent months, further accelerating ByteDance's progress in developing its own CPU. ByteDance, the parent company of TikTok, plans to deploy its self-developed CPU in its servers and data centers to support internal operations, while also preparing for the large-scale launch of products, including the Coze platform. Insiders say that ByteDance has been in contact with multiple external partners to seek assistance. These partners are expected to not only participate in chip design but also help secure capacity at manufacturing plants. The project is still in its early stages. The above report points out that this move reflects the industry's acceleration towards the "inference" stage, deploying AI models at the operational level of agent-driven workloads. Compared to NVIDIA's dominant position in the current AI wave with graphics processing units (GPUs), CPUs have a greater advantage in such tasks. Alphabet's Google, Amazon, Microsoft, and other global hyperscale data center operators have also begun developing their own customized CPUs to reduce procurement costs and optimize performance for specific workloads. This is making mainstream CPU manufacturers like Intel and AMD, strong challengers to NVIDIA's dominance in the AI chip field.