SoFi CEO enters prepaid forward contract on 1.5 million shares
Investing.com -- China is making steady progress toward self-sufficiency in artificial intelligence, but full independence, especially in critical enabling technologies like AI compute, remains a long-term goal, according to UBS.
“China has varied levels of self-sufficiency in AI,” UBS analysts wrote, citing strong domestic capabilities in the application and intelligence layers, driven by dominant local internet platforms and low-cost model developers like DeepSeek.
However, they added that the enabling layer, which includes semiconductors and compute infrastructure, still presents hurdles.
UBS estimates China’s overall semiconductor self-sufficiency will reach 27% by the end of 2025, up from the “mid-teens” during the pandemic, as the country builds capabilities across segments such as DRAM, analog chips, and semi-cap equipment.
The bank notes that progress has continued despite export restrictions.
The most challenging component is said to be AI compute, where global leader NVIDIA (NASDAQ:NVDA) holds over 80% market share.
UBS wrote, “NVIDIA’s GB200 chip can generate around 40 images whereas Huawei’s 910C chip can generate only 13 images,” illustrating the performance gap.
Still, Chinese companies held a one-third share of their domestic AI compute market in 2024, and UBS expects this to climb to 90% by 2029.
This localization drive will come at a cost, according to UBS.
“Strong revenue growth will be generated at the expense of compromising performance,” UBS said, warning of weaker pricing power and higher costs for Chinese players.
Nonetheless, domestic AI compute revenues are forecast to surge from $6 billion in 2024 to $81 billion by 2029, says the bank.
UBS recommends exposure to both global leaders like NVIDIA and Chinese firms such as Alibaba (NYSE:BABA) and Tencent (HK:0700), as “the AI compute market is big enough for all the players to grow.”