Huawei reportedly postponed Ascend 910C AI chip mass production

Huawei was about to begin the mass production of the Ascend 910C chips this month, but the company has reportedly postponed its plan. A new input reveals that the tech giant might initiate the respective activity by the end of this year.

X tipster @Jukanlosreve says that Huawei has postponed the Ascend 910C mass production to the end of this year. The company has delayed the AI chip manufacturing due to certain technical issues appeared in the chipset.

In the current scenario, the Kunlun P800 and the Ascend 910B chip are in the appropriate stock. The company may begin the batch delivery of these processors for high-end Chinese AI products.

The Chinese phone maker was about to begin the mass shipment of the Ascend 910C last month. Since the US restricted the Nvidia H20 chips in China to slow down the country’s AI growth, Huawei tried to step in as a helping hand in the situation.

Apart from the 910C, the company also started working on a new-gen AI chip called Ascend 910D, which is said to be more efficient than the Nvidia H100 processor.

Huawei reportedly postponed Ascend 910C AI chip mass production (Image Credits: X)

However, the latest report hints at a pause on Huawei’s AI chip targets. It looks like the company is still struggling with the supply chain, as a result of which, it has postponed the Ascend 910C AI processor’s mass production till 2025-end.

The tipster noted that many other Chinese chip brands are either out of stock or still in the development phase. These brands currently have a high degree of uncertainty in terms of release schedule.

He also revealed that the high-profile customers are working on a new rule that says, “sell only what you already have in stock; if you fail to deliver, you’re in trouble.” The regulation will become active starting this month.

X tipster further described the definition of the Chinese AI chips. He said that such components should meet three requirements. These are as follows:

• Card-to-card interconnectivity
• Single chip computing power over approx 300 TFLOPS (FP16)
• OAM form factor