Chinese regulators are weighing whether to approve purchases of Nvidia's AI chips, according to a report from the Seoul Economic Daily, a shift that signals Huawei's domestic accelerator lineup is not meeting the country's compute requirements at scale.

The report, titled "China Weighs Approving Nvidia Chip Purchases as Huawei Falls Short," points to a growing gap between the artificial intelligence compute that Chinese cloud providers and model developers need and what Huawei can currently deliver. Neither the article nor its sourcing specifies which Nvidia products Beijing would clear — the export-control-compliant H20 line, older inventory, or another category — nor does it set a timeline for a decision.

The deliberation matters because it reverses, at least conditionally, Beijing's earlier posture. Chinese authorities had discouraged domestic firms from buying Nvidia's downgraded China-market chips and had pushed companies toward Huawei's Ascend series as part of a broader import-substitution drive. That pressure intensified after Washington tightened export controls, culminating in the H20 restrictions that removed even Nvidia's compliant SKU from the Chinese market earlier this year.

Huawei's shortfall is the operative fact. The company's Ascend 910B and 910C processors are the backbone of China's domestic AI compute buildout, but production constraints — particularly the supply of advanced memory and the mature-node capacity at SMIC needed to fabricate the chips in volume — have limited shipments relative to demand from the largest cloud and internet players. The Seoul Economic Daily report does not quantify the gap, but the fact that regulators are reconsidering Nvidia approvals is itself an acknowledgment that substitution is behind schedule.

For Nvidia, even partial re-entry into China would carry real revenue weight. China accounted for roughly a fifth of Nvidia's data center revenue before successive rounds of US export controls eroded its addressable market, and the company has repeatedly flagged the lost sales in its filings and earnings calls. Whether Washington would permit resumed shipments of any Nvidia AI silicon to China — and under what license conditions — remains the other half of the equation, and the report addresses only the Beijing side of the approval chain.

The commercial logic cuts both ways. Chinese AI labs face a genuine compute bottleneck as they race to train and serve large models, and Huawei's constrained output forces rationing among buyers. At the same time, Beijing has strategic reasons to keep pressure on its domestic champions, and an official nod to Nvidia purchases could soften the urgency that currently drives orders toward the Ascend ecosystem.

No figures on order volumes, approval scope, or affected Nvidia product families appear in the source material, and any characterization of the size of a potential Chinese order would be speculative. What is established is the direction of the deliberation: regulators are actively discussing allowing Nvidia purchases, and Huawei's inability to satisfy demand is the stated reason.

A formal approval would reshape competitive dynamics across China's AI compute market in short order — Nvidia's installed software base remains a pull for Chinese developers — while continued hesitation would keep Huawei, SMIC and the memory suppliers in the Ascend supply chain under pressure to close the capacity gap.