Huawei’s efforts to match US computing power in artificial intelligence are facing challenges due to bug-ridden software. Customers of Huawei’s AI chipmaker are reporting performance issues and difficulties in switching from Nvidia products.
Always here for the terrible Dad jokes in FT comments section.
On piece our about Huawei's customers and R&D staff complaining about its buggy software for its rival Nvidia AI chips. https://t.co/jsaO5LjgjJ pic.twitter.com/kD98443Oqw
— Eleanor Olcott (@EleanorOlcott) September 4, 2024
Huawei has emerged as a frontrunner in developing a domestic alternative to Nvidia after Washington tightened export controls on high-performance silicon last October.
However, industry insiders say the chips still lag behind Nvidia’s for the initial training of models. They cite stability issues, slower inter-chip connectivity, and inferior software developed by Huawei called Cann. Huawei’s own employees are among those complaining about Cann.
Finally a story that shows Huawei is not 10 feet tall.
Huawei’s buggy software hampers China’s efforts to replace Nvidia in AI https://t.co/WtYbglJBXC via @ft
— Dennis Wilder偉德寧 (@dennisw5) September 3, 2024
Huawei’s bug-ridden software hampers China’s efforts to replace Nvidia in AI https://t.co/A4qXGKn76S via @ft
— Demetri (@Dimi) September 3, 2024
One researcher said it made the Ascend product “difficult and unstable to use” and work on testing it was being hampered. Another Chinese engineer briefed on Baidu’s use of the Huawei processors said the chips crashed frequently, complicating AI development work. To address the problem, Huawei has been sending engineers to help customers transfer training code previously written on Cuda into Cann.
Bugs hobble Huawei’s AI development
Baidu, iFlytek, and Tencent are among the tech companies that have received teams of engineers. Huawei can leverage a huge workforce to accelerate the shift.
According to the company, more than 50 percent of its 207,000 employees work in research and development, including the engineers dispatched to install technology for customers. Huawei has also set up an online portal for developers to give feedback on how its software can be improved. After the US tightened export controls in October, Huawei raised the price of the Ascend 910B, its chip used for training, by 20 to 30 percent.
Huawei’s customers have also expressed concern about supply constraints for the Ascend chip, likely due to manufacturing difficulties, with Chinese companies prevented from buying state-of-the-art chipmaking machinery from the Dutch company ASML. Despite the challenges, Huawei has seen strong demand for its AI chips. It reported a 34 percent increase in first-half revenues on Thursday, without providing a breakdown of sales for its different businesses.
More than 50 foundational models have “been trained and iterated” on the Ascend chip, according to a Huawei executive.
