Huawei set back back with new AI chip “910C”
The 910c performs the NVIDIA performance compared to the H100.
Listen to the story
According to a Reuters report, Huawei is ready to start a large -scale shipment of its new Artificial Intelligence Chip, The 910C. Some units have already landed with customers, the report said, even the Chinese tech Major quickly took steps to fill the zero left by the new stringent US export restrictions imposed by the administration of US President Donald Trump on AI Chips.
This development comes at a significant turn for China’s ongoing AI ambitions, such as models such as Dipsek R1 not only challenges established players such as microsoft-supported openi and Google, but also give them a run for money, at least as far as resource allocation is concerned. Many of these companies rely on Nvidia chips-the recent H20-all heavy-lifting which requires AI training and charge.
NVIDIa’s most powerful chips, such as H100 and A100 -The upcoming B200, has been banned to reduce the country’s technological progress on the AI chip exports of Trump administration in China, especially in areas where there is a possible military application. H20 was previously allowed for sale as a “obedient” option, but now, even those people require a special export license from the US government before the sale.
Huawei’s 910C has a large task in hand to help reduce China’s dependence on foreign chipmakes, removing the lack of any short-term supply for AI startups and cloud computing firms that are running to secure options for American-made chips.
Technical details are rare, but allegedly, 910C has been performed for H100, something that he can get by mixing two 910B chips in a single integrated package.
This dual-chap approach should allow Huawei to double computing power and memory capacity-at least in principle-without investing time and resources in building a new architecture, without investing time and resources, which can be a risky proposal at this time. More incremental improvements at 910B, as well as better support for diverse AI workloads, will be expected. More details are awaited.
