ByteDance is in talks with Shanghai-based GPU maker Iluvatar CoreX to buy AI chips for inference workloads and is also considering Baidu’s Kunlunxin processors, Reuters reported on June 15, 2026. The deals would make Iluvatar CoreX ByteDance’s third major domestic AI chip supplier after Huawei and Cambricon, further reducing its reliance on Nvidia hardware amid US export controls.
This article aggregates reporting from 3 news sources. The TL;DR is AI-generated from original reporting. Race to AGI's analysis provides editorial context on implications for AGI development.
This is a classic ‘chips are the new oil’ moment for the Chinese AI stack. With Nvidia effectively pushed out of China’s high‑end accelerator market by US export controls, ByteDance sourcing inference GPUs from Iluvatar CoreX and potentially Baidu’s Kunlunxin shows how quickly domestic alternatives are filling the gap. For Doubao and ByteDance’s broader agentic AI ambitions, reliable local silicon is no longer optional; it’s the constraint that determines how aggressively they can scale models and users.
Strategically, this is big for both ByteDance and China’s AI hardware ecosystem. Iluvatar has been positioned as one of the few domestic players whose performance can credibly challenge Nvidia in training and inference. Locking in a top‑tier customer like ByteDance de‑risks its roadmap and sends a demand signal that will attract more capital and talent into China’s GPU startups. For ByteDance, diversifying beyond Huawei and Cambricon reduces vendor risk and strengthens its ability to keep Doubao competitive against OpenAI, Anthropic, and Google.
Zooming out, this accelerates the decoupling of AI compute supply chains. As Chinese labs harden around local chips, Western export controls become less of a throttle and more of a forcing function for indigenous capability.