ByteDance is developing an artificial intelligence chip and has held discussions with Samsung Electronics on manufacturing support, according to a Reuters report citing two people familiar with the matter. The chip is designed for AI inference workloads and is part of ByteDance’s effort to secure long-term access to advanced processors as U.S. export restrictions tighten.
The company expects to receive initial chip samples by the end of March and is targeting production of at least 100,000 units in 2026, with plans to scale output to about 350,000 units. Talks with Samsung have also covered access to high-bandwidth memory, a critical component that remains in short supply as global AI infrastructure investment accelerates.
The project, internally codenamed SeedChip, reflects ByteDance’s broader push to build AI capabilities across hardware, large language models, and data center infrastructure. The company plans to spend more than $22 billion this year on AI-related procurement, including significant purchases of Nvidia’s H200 chips.
ByteDance denied the accuracy of reports related to its internal chip program, while Samsung declined to comment on the discussions.
The development follows other chip-focused initiatives in the sector, such as Cisco’s launch of the Silicon One G300 switch chip and router, which aim to improve AI data center performance while competing with Broadcom and Nvidia solutions.
