DeepSeek's Huawei Order Still Leaves Nvidia Doing the Training
Yahoo Finance ·
DeepSeek's Huawei Order Still Leaves Nvidia Doing the Training Renato Neves, CFA Fri, September 4, 2026 at 8:23 AM EDT 1 min read NVDA This article first appeared on GuruFocus . DeepSeek plans to deploy at least 160,000 of Huawei Technologies' top accelerators at a data center it is building in Inner Mongolia, a step that would advance China's efforts to replace Nvidia Corp. ( NASDAQ:NVDA ). Huawei's next-generation Ascend 950DT would power the models, according to people familiar with the matter, in what would become one of the largest known clusters of Huawei AI chips. DeepSeek wants more of Huawei's best silicon than Huawei can currently produce, and shortages of components including top-end memory will hold 950DT output to the low hundreds of thousands this year. Huawei is also serving other customers and shipping small volumes overseas. Filling the order could take more than a year, the people said. There's a second limit worth noting. DeepSeek doesn't plan to use the 950DT for training, even though Huawei built and marketed the chip for exactly that workload. It tried training earlier models on Huawei parts and has kept using Nvidia accelerators for that step. DeepSeek is in talks to raise billions to fund the buildout.
DYAX Investor Sentiment
Bullish (Long) 49% · Bearish (Short) 51%
417 participants
Related News
- A New Report Suggests September Is the Worst Month to Buy Stocks. History Offers a Clear Answer for How to Handle This
- Industry-Leading Email Signature Platform Crossware Is Now Available Through Ingram Micro's Xvantage Platform as Part of the Microsoft Marketplace Private Offers
- Airbus (AIR FP) reportedly reaches a pre-agreement with Spanish unions, Democrata reports
- Daily US Equity Opening News: NVDA in talks for USD 2.5bln Thinking Machines investment; MU plans HBM capacity increase
- Where Will Sandisk Stock Be in 3 Years?
- This Week In Cloud AI - Opti9 Expands Canadian Reach with Strategic Hut 8 Acquisition
