UBS: AI Profit Inflection Point for China's Internet Giants Still 2-3 Years Away
nashnova research
UBS's head of China internet research Kenneth Fong sees AI profits shifting from upstream hardware to downstream platforms within two to three years — but the giants' cash burn is already showing up in quarterly results.
Who is capturing the profits right now?
Profit distribution across the AI chain depends on where the bottleneck sits. Fong's logic: compute capacity is scarce today, so upstream suppliers — chipmakers, server vendors — hold pricing power and capture the bulk of the profit pool.
This means → even though internet platforms own massive data and user bases, they are currently the buyers, not the collectors.
In plain terms = whoever controls the scarce resource sets the price — right now, that's the GPU sellers.
When does the inflection point arrive?
Fong expects capacity constraints to ease in two to three years, at which point upstream scarcity fades and pricing power shifts to downstream platforms.
His core reasoning: internet companies control distribution, data, and users — three assets that become real bargaining chips only once capacity is abundant.
This reflects a classic industry-cycle pattern: infrastructure builders profit first, then the companies that run on that infrastructure.
What is the cost the giants are paying now?
Tencent's Q2 capex nearly tripled to RMB 52.8 billion; free cash flow turned negative for the first time, at RMB −13.8 billion.
Alibaba's Q2 capex hit RMB 67.7 billion; free cash outflow more than doubled year-on-year to RMB 44.7 billion.
In plain terms = both companies are spending heavily on AI compute, and the spending rate now exceeds the earning rate — net cash is flowing out.
What are investors worried about?
UBS notes that macro headwinds combined with aggressive AI spending have made investors cautious on China's tech giants.
This means → the market does not doubt the long-term AI story — it worries that near-term profits are being compressed by capex, with payoff still two to three years away.
This reflects a key validation point: if capacity constraints do not loosen on schedule, the platform layer may fail to deliver returns on this round of investment.
市场有风险，内容仅供研究参考，不构成投资建议。