Huawei expects 60% chip revenue jump

Huawei’s AI-chip business is starting to look less like a backup plan and more like a real domestic industry. That’s the point of the new forecast — Huawei expects revenue from its AI chips to rise at least 60% this year, to about $12 billion, as Chinese customers buy more local hardware instead of waiting on Nvidia. The bigger story is not just one company’s sales jump. It’s that China’s AI compute market is being pushed into its own lane, with its own chips, clusters, and software tradeoffs. (money.usnews.com) ### Why is Huawei suddenly talking this way? Because demand is there now. Chinese internet groups, data-center operators, and state-linked buyers have spent the last year looking for a workable answer to tighter U.S. export controls. Huawei’s Ascend line has become the obvious domestic option, so the company can talk about growth in a way that would have sounded aspirational a year or two ago. The forecast surfaced in Financial Times reporting and was echoed by Reuters on May 1. (money.usnews.com) ### What changed on Nvidia’s side? The key disruption is U.S. licensing. Nvidia said in May 2025 that Washington had told it a license was required to export H20 chips to China. That led to a $4.5 billion charge tied to excess inventory and purchase obligations, and Nvidia said it also could not ship another $2.5 billion of H20 rev(money.usnews.com)eal opening for local rivals. (nvidianews.nvidia.com) ### Why does that help Huawei so much? Because AI infrastructure buyers hate uncertainty. If you are building a cluster for model training or inference, “maybe the licenses clear later” is not a comforting procurement plan. Huawei does not need to beat Nvidia at everything to win those orders. It just needs to be available, supportable, and good enough for t(nvidianews.nvidia.com)he quiet shift underneath this story. (money.usnews.com) ### What is Huawei actually selling? The relevant family is Ascend — Huawei’s in-house AI compute platform. Huawei’s enterprise materials describe Ascend as a full-stack offering built around processors plus Atlas cards, servers, and clusters for training and inference across edge, server, and cloud deployments. Huawei has also bee(money.usnews.com)ell a whole compute environment. (e.huawei.com) ### Is this just about hardware? No — and that’s the catch. Chips matter, but software ecosystems matter almost as much. Nvidia’s moat has long been CUDA and the huge amount of optimization work built around it. A more Huawei-centered market means developers inside China may increasingly tune models, inference pipelines, and orchestration tools for Ascend-specific environments. That can work well locally, but it als(e.huawei.com)t needs to run across regions or vendors. (e.huawei.com) ### Does this mean Huawei has “won” China? Not cleanly. China’s AI hardware market is still fragmented, and local alternatives have had real limits in supply, software maturity, and interoperability. Nvidia also has not disappeared — it has been trying to preserve a China business where rules allow. But the direction of travel is clearer now: every new export-control shock gives domestic suppliers another chance to move from second choice to standard choice. (blogs.nvidia.com) ### What should software teams take from this? Assume compute is becoming regional. The old idea that one dominant accelerator stack would cover everyone is getting weaker. Teams that depend on AI infrastructure now have to think more about model portability, vendor diversity, and where their performance assumptions come from. The cheapest or fastest setup in one market may not even be available in another. (money.usnews.com) ### Bottom line Huawei’s 60% growth target is a sales story on the surface. Underneath, it’s a market-structure story. China is building an AI compute ecosystem that increasingly has to stand on its own — and that changes the rules for chip vendors, cloud buyers, and software teams alike. (money.usnews.com)