Huawei is accelerating its AI chip roadmap as it deepens its challenge to Nvidia, unveiling new details at its Huawei Connect conference in Shanghai this week. Rotating chairman Eric Xu confirmed the company has pulled forward its Ascend 960DT chip release to Q1 2027 — nine months ahead of the original schedule — while the Ascend 960PR follows one quarter early in Q3 2027, with the Ascend 970 and 980 targeted for 2028 and 2029 respectively.
The urgency is deliberate: Huawei says it cannot currently produce enough AI chips to meet domestic demand, and as a result has no plans for major international expansion, focusing almost entirely on satisfying China's market first. The company also unveiled its new Atlas 960 SuperPoD computing cluster and Peerium architecture, designed to connect up to 1 million processors — pushing further into the massive interconnected clu
Xu was blunt about the motivation: "We cannot accept a destiny where our fate is determined by others in terms of willingness to sell chips to China or not" — a direct reference to US export controls that have restricted China's access to Nvidia's most advanced processors since 2019, with Huawei itself facing its own US trade restrictions since May of that year.
The momentum appears real rather than purely rhetorical: more than 40 AI models have already been trained on Huawei's computing platform, with over 5,200 developers now active monthly on Ascend chip software. Xu said the Ascend 950DT chip has shown strong results in testing, and he expects Chinese AI model developers to begin training on 950DT-powered systems throughout the coming year.
This accelerated timeline reflects a broader pattern we've traced all year — AI competition increasingly playing out at the infrastructure layer, not just the model layer. Just as the US tightens remote-chip-access
See What’s Next in Tech With the Fast Forward Newsletter
Tweets From @varindiamag
Nothing to see here - yet
When they Tweet, their Tweets will show up here.
