ByteDance is now developing its own AI chip and will be working with Samsung Electronics to manufacture it, according to Reuters.
The sources who spoke to Reuters said ByteDance expects to receive sample chips by the end of March. This venture is for the company to “secure supply of advanced processors” as Reuters put it. Reuters also says ByteDance plans to produce at least 100,000 units this year. One of the sources added that the company hopes to increase output over time to as many as 350,000 units.
Negotiations with Samsung, according to Reuters’ source, also cover access to memory chips, which are in short supply as companies build out AI infrastructure. One source told Reuters that memory supply makes the potential agreement particularly attractive. High bandwidth memory such as HBM3 and HBM4 is in heavy demand for AI servers. Only Micron, Samsung and SK Hynix manufacture these chips at scale, according to Sammobile.
ByteDance denied the Reuters report. A spokesperson said in a statement that “The information about ByteDance’s in-house chip project is inaccurate,” without giving further detail. Samsung declined to comment.
Is This Part Of ByteDance’s Increased AI Spending Plans?
The chip project, codenamed SeedChip, is one of the many under ByteDance’s AI investment plans. Reuters reported that the company plans to spend more than 160 billion yuan (that is about $22 billion), on AI related procurement this year. More than half of that budget will go towards buying Nvidia chips, including H200 models, and advancing its own processor, according to one of the sources.
ByteDance has worked on chip development since at least 2022, when it began hiring chip specialists. In June 2024, Reuters reported that the company was working with US chip designer Broadcom on an advanced AI processor, with manufacturing expected to be outsourced to Taiwan’s TSMC.
The new discussions with Samsung suggest a possible change in manufacturing partner. Sammobile reported that Samsung Foundry could supply sample chips by March 2026 and that ByteDance aims to scale production to 356,000 units in time.
More from Artificial Intelligence
- Harvard Says AI Adds Onto Workloads More Than It Reduces Them, Here’s Why
- Top AI Startups In Ireland
- Which Countries Have The Biggest Data Centre Carbon Footprints In 2026?
- AI In VoIP: What It Is And How Does It Work?
- How To Identify Whether Your Startup Is AI-Ready
- New Report: 100% Of Businesses Now Use AI, But Only 25% Can Actually Scale It
- AI Trends In Photography – Transforming The Workflow
- Introducing Moltbook: The “Reddit” For AI Chatbots
ByteDance set up a division called Seed in 2023 to develop AI models. The company promotes these models across its businesses, which span short video, ecommerce and enterprise cloud services. The SeedChip project links hardware and software under one roof.
At a January staff meeting, ByteDance executive Zhao Qi told employees that the company’s AI investment would benefit all divisions, according to a person briefed on the meeting who spoke to Reuters. Zhao, who oversees the Doubao chatbot and its overseas version Dola, admitted that ByteDance’s AI models lag behind global leaders such as OpenAI but pledged continued backing for AI development this year.
Can ByteDance Close The Gap With Global And Domestic Rivals?
Chinese tech groups face tighter US export controls on advanced chip sales to China. Reuters reported that these restrictions have added urgency for companies to design their own AI processors.
ByteDance is not alone. Alibaba recently introduced its Zhenwu chip for large scale AI workloads, according to Reuters. Baidu sells AI chips to external clients and plans to list its chip unit Kunlunxin soon.
Global companies including Google, Amazon and Microsoft have also developed in house AI chips to lessen reliance on Nvidia, which dominates the market for advanced AI processors, Reuters reported.
ByteDance’s move follows this venture as companies compete to control supply and costs.
