The CEO's forecast
During the call with analysts on the results for the second quarter of fiscal year 2027, Alibaba Group CEO Wu Yongming identified 2027 and 2028 as the two years in which China will face the greatest shortage of computing capacity. The statement was reported by IT之家 based on an article by Jiemian News. It is a forecast made by the company's top management, not a market figure already measured or verified by independent sources.
The timeline indicated by Wu is twofold: the greatest tension would be concentrated in the next two years, while from 2027 onward the composition of Alibaba Cloud's supplies would change structurally, tilting predominantly toward its own in-house Pinguotou chips.
The shift to domestic chips
According to the statements, after 2027 the chip supply for Alibaba Cloud will consist mainly of Pinguotou processors, alongside purchases of other China-made chips intended for internal inference needs. The company therefore intends to maintain a share of external procurement, but within the scope of domestic semiconductors.
Wu specified the priority assigned to flagship chips: they will be reserved first for training large models, to support the models' own iteration. Only at a second level will the company take into account the needs of partners' customers. For inference, by contrast, Alibaba will aim to purchase a broader variety of chips, including other Chinese brands.
Models already being trained on in-house chips
On the same day, IT之家 recalls, Wu had said that the Qwen model is already training the next generation on Pinguotou chips and that Bailian already provides inference services based on domestic semiconductors. These are company statements, as reported by the source, and there is currently no external confirmation of the scale and performance of these uses.
The CEO argued that Alibaba's full-stack AI infrastructure will draw a clear advantage from the domestic computing capacity supply chain. The combination of in-house developed chips and in-house developed models, in his view, would translate into a more favorable cost per unit of compute and greater efficiency in developing artificial intelligence applications.
What remains to be verified
The statements outline a declared strategy, not results already achieved: there is a lack of public data on Pinguotou's production capacity, on the volumes actually available and on the share of external procurement that Alibaba intends to maintain. Nor is the cost impact quantified; it is indicated only as an expected improvement.
The shortage forecast for 2027 and 2028 remains the CEO's assessment: the material offers no elements to measure its extent or to establish which segments of the Chinese market will be most affected. The news is therefore relevant as a signal of industrial direction, but its concrete consequences remain to be observed.
