DeepSeek Optimizes New AI Model for Huawei Chips, Cutting Nvidia Reliance Under US Curbs
Updated
Updated · The New York Times · May 12
DeepSeek Optimizes New AI Model for Huawei Chips, Cutting Nvidia Reliance Under US Curbs
6 articles · Updated · The New York Times · May 12
DeepSeek said its latest AI model, released last month, was optimized for Huawei chips for the first time—a small but notable shift away from Nvidia hardware.
That move shows Chinese AI firms are redesigning systems around U.S. export controls rather than waiting for access to advanced American semiconductors to return.
Huawei’s gain matters because most leading AI models worldwide still run on Nvidia chips, making domestic alternatives a key test of China’s push for technological self-reliance.
The announcement also gives Beijing a confidence boost ahead of this week’s Trump-Xi summit, where Nvidia chip restrictions are expected to remain a live issue in trade talks.
As Nvidia's China market vanishes, can Huawei's new chips truly power the nation's AI ambitions?
Is the world fracturing into two separate, competing AI ecosystems led by the US and China?
DeepSeek V4’s “Day 0 Adaptation” on Huawei Ascend Chips Signals China’s AI Independence Drive
Overview
DeepSeek V4's launch marks a turning point for China's AI industry, as it shifts away from reliance on Nvidia hardware and embraces domestic solutions like Huawei's Ascend chips. This move is highlighted by the unprecedented 'Day 0 adaptation,' where multiple Chinese chipmakers achieved immediate compatibility and performance with DeepSeek V4 at launch, setting a new benchmark for local chip capabilities. Driven by U.S. export controls that limit access to advanced Nvidia chips, DeepSeek faces high costs and supply constraints, but expects prices to drop as Huawei's hardware scales up, signaling a new era of self-reliant AI development in China.