Huawei Unveils More Than 10 AI Chips to Rival Nvidia

Huawei introduced more than 10 chipsets for AI computing infrastructure on Thursday at its Huawei Connect conference, Nikkei Asia reports. The lineup spans the next generations of the company's AI accelerators, its CPUs, and chips that move data quickly between processors — the three pieces any large training cluster needs at once.

Read the categories, not the count

The count itself shifts depending on which line you read: Nikkei's headline puts the figure at 11 AI-related chips, while the story text says more than 10 chipsets. Both come from the same report, and the gap is a matter of what gets counted rather than a contradiction. The useful signal is the breadth — Huawei is not filling one gap in its stack but refreshing several parts of it together.

Supply is the constraint, not the announcement

Acting CEO Eric Xu framed the limit plainly. According to derStandard, he said the company cannot cover demand inside China and therefore does not plan to expand onto international markets at full scale. Nikkei's subhead states the same constraint from the product side: domestic demand for Ascend accelerators runs ahead of what Huawei can supply.

For buyers outside China, that is the operative sentence in the whole announcement. New silicon that cannot be shipped abroad changes the competitive picture in China first and everywhere else much later.

Why the interconnect matters

derStandard's subhead highlights faster communication between processors rather than any single chip's peak throughput. That emphasis is worth noting, because in large training runs the links between accelerators often decide how much of the theoretical performance survives contact with a real workload. Whether Huawei's new connectivity parts deliver on that is not something either report establishes.

What neither source supports

Both articles are freely readable only as previews. That leaves individual model names, all technical specifications, availability dates and every direct performance comparison with Nvidia, Intel or AMD unverified here. Until those details are published, the announcement documents Huawei's roadmap and its capacity problem — not a closed performance gap.