A forecast says China’s internet conglomerates will take a bigger share of artificial intelligence (AI) profits within 2 to 3 years.

On Sept. 2 (local time), the South China Morning Post (SCMP) reported that UBS sees AI profits currently going to upstream players that control chips and infrastructure. It expects pricing power to move to internet platforms with data, users and distribution capabilities if supply constraints ease.

Kenneth Fong (케네스 퐁), head of UBS China internet research, said at an event in Shenzhen, "Right now, capacity constraints are upstream, so they take a large part of the overall profit pool." He added, "If constraints ease in 2 to 3 years, pricing power will move downstream to those with distribution capabilities, data and users." He forecast that the monetisation phase for internet companies will begin again.

The forecast comes alongside a recent expansion in aggressive AI investment by China’s big tech companies. Tencent Holdings nearly tripled second-quarter capital expenditure to 52.8 billion yuan, and free cash flow recorded a negative 13.8 billion yuan for the first time. Alibaba Group Holding posted a June-quarter free cash outflow of 44.7 billion yuan, more than doubling from a year earlier. Quarterly spending during the same period was 67.7 billion yuan.

Markets, however, are taking a cautious stance on large-scale AI spending by Chinese technology companies, citing a slowdown in the macro environment in the second half and near-term earnings pressure. UBS said expanded investment in hardware and infrastructure could immediately erode profits. UBS assessed that, despite these burdens, major companies are choosing defence and offence at the same time. It said they feel they cannot afford to fall behind in the AI race and are also confirming the possibility of recouping investment when AI is added to cloud and existing businesses.

Fong said Chinese technology companies are spending cash flow equivalent to 1 to 1.5 years annually on AI investment. He said, "Even if AI does not materialise as expected, it is effectively at the level of giving up 1 year of profit," but added, "At least it can keep you from being pushed out of the competitive arena."

Even by current investment size, Chinese companies fall well short of U.S. rivals. UBS sees total spending by Chinese technology companies at about one-seventh of that of major U.S. technology companies. Access to advanced overseas chips is restricted and business scale is relatively smaller. As a result, UBS said upstream hardware and infrastructure remain advantaged in the near term, but the profit-sharing structure could change if supply bottlenecks are resolved.

Cost efficiency was presented as a strength of China’s AI camp. Xiong Wei (슝웨이), a China internet analyst at UBS Securities, said training costs for Chinese models are estimated at less than 10 percent of those of global leading companies. He also assessed that the average application programming interface price of major Chinese models is less than 20 percent of that of global competing models.

Xiong said that, based on this cost structure, Chinese model developers are not relying only on loss-making competition. He said, "Chinese developers are maintaining healthy gross profit margins rather than burning cash to expand adoption." He added that they are improving profitability through technological innovation while continuing to improve training and inference efficiency even as model usage rises rapidly.

Ultimately, UBS’s view is that the decisive factor in China’s AI race lies in mid- to long-term platform dominance rather than short-term capital expenditure size. For now, cash flow at companies such as Tencent and Alibaba is under pressure. UBS forecast that if supply constraints ease, internet platforms with large user bases, data and service distribution channels could become central to AI monetisation.