NuAxiSemiconductor trade intelligence
Huawei unveils new Atlas 350 AI accelerator with 1.56 PFLOPS of FP4 compute and up to 112GB of HBM — claims 2.8x more performance than Nvidia's H20
What happened
Huawei unveiled the Atlas 350 AI accelerator at its China Partner Conference, featuring the Ascend 950PR chip, 1.56 PFLOPS FP4 performance, and 112GB of proprietary HiBL 1.0 HBM. The device targets the inference prefill stage and claims 2.8x the performance of Nvidia's H20. Bullish for Huawei's proprietary HiBL 1.0 HBM supply chain.
Impact and signal
Low impactImpact score: 25/100
Affected components and suppliers
HBMAI Accelerator
Source and time
Source: Tom's Hardware · 3/24/2026 · View original source
Open the NuAxi live feed →
Real-time pricing, supply signals and AI analysis for semiconductor traders