DeepSeek V4 and Huawei Ascend 950DT: AI Inference Cost Reduction
A groundbreaking trace-level analysis by Wall Street research firm SemiAnalysis has revealed that DeepSeek V4 and Huawei Ascend 950DT AI accelerator were co-designed from the ground up, overturning the prevailing assumption that the model was merely adapted to run on domestic Chinese chips after development. The revelation explains how DeepSeek was able to slash its V4-Pro API pricing by 75 percent, bringing million-token inference costs to just 0.20 yuan, roughly 50 times ch
A new analysis by SemiAnalysis shows that DeepSeek V4 and Huawei's Ascend 950DT were co-developed, challenging previous assumptions. This collaboration allowed DeepSeek to reduce its API costs by 75%, making inference significantly cheaper than competitors. The Ascend 950DT features a unique architecture designed for DeepSeek's needs, including advanced memory and processing capabilities. The chip's design led to improved performance and efficiency compared to other platforms. DeepSeek's market share increased rapidly, overtaking OpenAI in some metrics. Major Chinese companies have placed large orders for the Ascend 950DT, indicating strong market confidence. The chip is set to launch on Huawei Cloud in August 2026, marking a shift in AI infrastructure reliance.
This development highlights a significant shift in AI infrastructure, reducing costs and increasing reliance on domestic Chinese technology.
📌 Kaynak
Bu özet China kaynağından otomatik derlenmiştir. Tamamı için orijinal habere gidin.
Orijinal haberi oku →