Power grid investment boosts China’s AI sector: Alibaba chair

Alibaba chairman Joe Tsai said China’s recent AI progress has been helped by heavy investment in the power grid, commitment to open-source models, and the country’s manufacturing supply chain.

He was speaking at the China Development Forum 2026 in Beijing.

Tsai said China has spent about US$90 billion a year on power transmission in recent years and added that newly installed power generation capacity last year was 10x that of the US, which he argued lowers energy constraints for AI.

He said open-source models have reduced barriers in AI development and that China’s industrial base generates large volumes of data for training industrial AI systems.

Tsai said Alibaba is rolling out AI applications using its Qwen model, while Industry and Information Technology Minister Li Lecheng and Xiaomi CEO Lei Jun outlined continued government planning and R&D investment that would exceed 200 billion yuan over the next five years at the event.

🔗 Source: South China Morning Post

🧠 Food for thought

Implications, context, and why it matters.

China’s power surplus and Alibaba’s integrated AI set it apart

Joe Tsai’s remarks on energy leave out part of China’s edge; a China-focused electricity expert told Fortune that the nationwide reserve margin has never fallen below 80%–100%, which suggests at least double the capacity needed 1.
The same expert told Fortune that U.S. regional grids usually run near a 15% reserve margin; in July 2024, a voltage fluctuation in northern Virginia disconnected 60 data centers at once, creating a 1,500-megawatt surplus and forcing emergency grid adjustments 1 2.
Tsai also did not mention agentic capabilities in the consumer-facing Qwen App, meaning it can take actions on a user’s behalf; the app is in public testing in China and can order food, make in-chat payments, plus plan and book travel 3.
Those features rely on tight links inside Alibaba’s ecosystem, including Taobao (Alibaba’s e-commerce marketplace) for shopping and Alipay (Alibaba affiliate Ant Group’s digital payments service) for payments; one analysis said this structure helps Alibaba connect to a wide set of consumer services, while the claim that no single U.S. company matches it is not established in the provided sources 4 5.

The AI race tracks national infrastructure choices

Different ways of building infrastructure can shape how fast AI moves from labs into daily use 1.
China’s planned overbuild of energy capacity lets it treat AI data centers as a way to “soak up oversupply,” according to the electricity expert quoted by Fortune 1.
The U.S. approach leans on private capital; the same expert said many investors look for a 3-5 year return, which clashes with grid projects that can take a decade, and Goldman Sachs warns these bottlenecks could limit industry growth 1.
At the company level, Alibaba can weave AI into commerce plus payments, which may speed up rollout of transaction-ready consumer AI compared with systems that depend on partnerships across separate platforms 4 5.