The funding comes with a hardware shift
Deepseek paused the current round after Liang’s remarks about the company’s dependence on Nvidia chips circulated widely. The startup is also developing its own inference chip, aiming to reduce reliance on both Nvidia and Huawei.
In April, it released preliminary versions of V4-Pro and V4-Flash. The planned data center, with at least 160,000 Huawei chips, points to a substantial infrastructure buildout even as Deepseek works on an alternative chip of its own.
What the round does not answer
CATL declined to comment. Tencent and Deepseek did not respond to requests for comment.
I think the more interesting tension is between Deepseek’s open-model ambition and the scale of infrastructure it is assembling. The company wants to keep model weights freely available, but its IPO plans now sit alongside a major funding round, a large Huawei-based data center and a push to build its own inference chip. The announcement gives no details on the round’s size or terms, leaving investors to judge how much of the plan depends on outside capital—and how quickly Deepseek can lessen its hardware dependence.
Daily AI news
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
Only what matters — every day
Follow on X