Internet platforms with vast data and large user bases will capture a larger share of artificial intelligence profits in two to three years, even though macro headwinds have temporarily fuelled investor caution over aggressive AI spending by Chinese tech giants, UBS analysts said. Investors have been cautious over a weak macro environment in the second half of the year, and greater AI spending for hardware and infrastructure would drag down short-term profits, but the industry’s power balance was set to flip soon, according to Kenneth Fong, head of China internet research at UBS. “This is all about the cycle. Now the capacity constraint is on the upstream... so [they] capture a big chunk of the whole profit pool,” Fong said in an interview on Tuesday on the sidelines of a UBS event in Shenzhen. “But two to three years down the road after the capacity constraint eases, the pricing power will shift to the downstream, where they have the distribution capability, data and users,” Fong noted. “The internet company will start to work again.” To strengthen their AI war chest, Chinese tech giants have been ramping up their capital expenditures in the past quarter, at the expense of their free cash flows. Tencent Holdings nearly tripled its second-quarter capex to 52.8 billion yuan (US$7.85 billion), while it posted a negative free cash flow of 13.8 billion yuan for the first time. Alibaba Group Holding, which owns the South China Morning Post, saw its free cash outflow in the June quarter more than double to 44.7 billion yuan year on year, on the back of a 67.7 billion yuan quarterly outlay. Still, the total spending of Chinese technology firms was eclipsed by their US counterparts, accounting for only one-seventh of that of American giants, as the former faced restricted access to advanced foreign chips and operated at smaller scales, according to Fong. Major Chinese technology firms were both defensive and offensive with their AI spending, as they feared missing out on AI advancements while seeing strong returns in cloud and existing business integrations, Fong said, adding that they are using one year to one-and-a-half years’ annual cash flows for their investments. “Even if AI fails to materialise, they basically give up one year of profit, but at least they can make sure they are still in the game,” Fong said. However, compared with US AI firms, Chinese AI labs and tech giants have stood out with a clear advantage in cost efficiency, according to Xiong Wei, China internet analyst at UBS Securities. The training costs for China’s models are estimated to be less than 10 per cent of those of global leaders, while the average API price of major China models was less than 20 per cent of their comparable global peers, Xiong said at the UBS briefing. “On this basis, domestic model developers are not operating at a loss,” Xiong said. “Rather than burning cash for adoption, Chinese developers are maintaining healthy gross margins while utilising technological innovation to enhance profitability, which continuously optimises their training and inference efficiency even as model usage grows rapidly.”