While the Western AI world is busy with Pentagon deals and billion-dollar valuations, China is quietly building its own infrastructure. And Huawei is at the center of it.
The numbers
Huawei expects around $12 billion in AI chip revenue for 2026 — at least a 60 percent increase over $7.5 billion the year before. The main driver is the Ascend 950PR processor, which entered mass production in March 2026 and has reportedly already secured most orders for the year.
An upgraded version, the Ascend 950DT, is planned for the fourth quarter.
Why this is happening
Two factors are driving the growth. First, US export restrictions are effectively pushing Nvidia out of the Chinese market. Second, DeepSeek optimized its V4 model for Huawei hardware — a signal to the entire Chinese AI industry that Nvidia is no longer necessary.
The dynamic is self-reinforcing: the more developers optimize for Huawei, the better the ecosystem gets, the less anyone misses Nvidia.
What this means for the global AI market
We’re watching the emergence of two separate AI ecosystems in real time. The West builds on Nvidia and TPUs, China on Huawei’s Ascend. Models from DeepSeek, Qwen, and other Chinese labs already run natively on Huawei hardware.
For Nvidia, this is a problem. Not because China is at the same performance level in the short term — the Ascend 950PR still trails the latest H200 and B200 chips. But because a market with over a billion potential users is setting its own standards.
And Reuters notes: the 60 percent projection comes from an FT report and couldn’t be independently verified. But even if the number is slightly lower — the direction is unmistakable.
Sources: