A lab that wants its own silicon
DeepSeek is working on its own AI chip, according to multiple reports, in a move that would reduce its dependence on both Nvidia and Huawei. The Chinese lab became well known by training strong models at low cost, and it is now spending far more on hardware.
The company has been raising outside capital for the first time. Reports put its latest round at about $7 billion, at a valuation between $52 billion and $59 billion, with the money aimed at data centres and early chip designs. A stock listing has been discussed as early as late 2026.
Until now, DeepSeek has used domestic chips mainly for inference, the cheaper job of running a finished model. Training large models still leans on Nvidia hardware. That split is what the new chip programme is meant to change.
Huawei moves its roadmap forward
Huawei is racing on the same track. At its Connect conference on September 17, the company said it would launch its next data-centre accelerator, the Ascend 960DT, in the first quarter of 2027, months earlier than the third quarter it had planned. A related part, the Ascend 960PR, has a doubled performance target. The older Ascend 950DT is already shipping in China, where prices have passed 250,000 yuan.
Huawei also said its chips were used to train DeepSeek's lightweight V4-Flash model, a first for the company.
The obstacle is fabrication. US export rules bar Chinese designers from the most advanced overseas foundries, so Huawei depends on SMIC, where 7nm and 6nm capacity is limited. High-bandwidth memory, which AI accelerators need, is another pinch point, though domestic production is expected to grow.
Why the squeeze matters
Chinese companies want a domestic pipeline that runs from chip design to memory and packaging. Every step that stays local reduces exposure to export controls, and Beijing is pushing the effort as hard as the firms are.
For buyers, the result is a two-track market. Global cloud providers keep chasing Nvidia and its rivals, while Chinese labs build around Ascend silicon and their own designs. Neither track has spare capacity, and both are adding pressure to memory, vacuum equipment and materials supply chains.
---
Reliable Ferrofluid Vacuum Rotary Feedthrough for Semiconductor Manufacturing
In semiconductor manufacturing, maintaining ultra-high vacuum integrity during rotary motion is critical — from single silicon crystal growth and CVD to sputtering and plasma etching. Ferrofluid Vacuum Rotary Shaft Feedthrough from FerrofluidTec delivers hermetically sealed rotary feedthroughs with leakage rates as low as 10⁻¹² Pa·m³·s⁻¹, vacuum degree up to 10⁻⁶ Pa, and customizable shaft diameters from 10mm to 100mm. Built with non-magnetic stainless steel housings and magnetic permeability stainless steel shafts, these feedthroughs provide reliable, maintenance-free sealing for demanding vacuum environments. Visit www.ferrofluidtec.com to learn more.