Chinese internet platforms with large user bases and developed distribution networks could begin capturing a significant share of profits from artificial intelligence within two to three years. The key condition is the easing of restrictions on access to advanced chips and infrastructure. I draw this conclusion based on fresh estimates from UBS analysts presented at an industry event in Shenzhen.
Shift in Market Power
Currently, the bulk of margins in the AI sector are captured by suppliers of hardware and related services, as the market faces an acute shortage of computing power. However, as Kenneth Fong, head of UBS research for China's internet sector, emphasizes, should export barriers be relaxed, "pricing power" will shift to platforms possessing vast data and monetization channels. For players like Tencent and Alibaba, this means a return to a phase of actively extracting value from their ecosystems.
Warning Signs in Financial Reports
The forecast came amid a sharp surge in AI spending by Chinese tech giants. Alibaba nearly tripled its capital expenditures in the second quarter, bringing them to 52.8 billion yuan ($7.86 billion). Notably, the company's free cash flow turned negative for the first time, reaching -13.8 billion yuan ($2.05 billion). For the quarter ending in June, the outflow amounted to 44.7 billion yuan ($6.65 billion)—more than double the figure from the previous year. Total quarterly expenditures reached 67.7 billion yuan ($10.07 billion).
UBS notes that the market views such investments with caution amid slowing macroeconomic dynamics in the second half of the year. Fong's assessment is telling: Chinese tech companies' annual AI spending is equivalent to their cash flow over one and a half years. This creates significant pressure on short-term profitability.
Betting on Efficiency, Not Volume
However, even with weak returns, these investments are necessary to avoid falling out of the competitive race. The total AI spending of China's tech sector amounts to only about one-seventh of that of their American counterparts. This gap is explained by limited access to advanced foreign chips and smaller business scale, which for now keeps the advantage with "hardware."
China's strength remains cost efficiency. According to estimates by Xiong Wei, an analyst at UBS Securities, the cost of training local models does not exceed 10% of the level of global leaders, while the average API price of major Chinese models is less than 20% of international competitors. In the medium and long term, the decisive factor will not be the volume of capital expenditures, but the dominance of platforms with large audiences and data.
As a reminder, in August Alibaba raised $10.2 billion for AI development through the placement of 710 million new shares.
My comment: Chinese platforms are in a unique position—they can win the race through total cost savings and scale, rather than an arms race in data centers. However, the two-year horizon looks optimistic: without the removal of chip restrictions, monetization may be delayed, and the current cash flow of the giants already demonstrates the limits of their patience.