Chinese Tech Giants Report $40 Billion Combined AI Investment for 2026

Capex Plans Double Year-Over-Year

China's four largest technology companies collectively committed approximately $40 billion in artificial intelligence-related capital expenditure for calendar year 2026, roughly double their combined AI spending in 2025, according to disclosures in earnings reports and investor presentations released during the February reporting season.

Alibaba Group led with a projected $15 billion in AI infrastructure spending, followed by ByteDance at $12 billion, Tencent Holdings at $8 billion, and Baidu Inc. at $5 billion. The investments cover GPU procurement, data center construction, model training compute, and AI talent acquisition.

Self-Sufficiency Imperative

The spending surge is driven in part by U.S. export restrictions that have progressively limited Chinese access to the most advanced AI chips. The expanded controls announced in January, which restrict mid-range inference chips, have intensified urgency to build domestic computing capacity.

"Chinese tech companies are essentially pre-ordering all the compute they can get," said Kai-Fu Lee, chairman and CEO of Sinovation Ventures and former head of Google China. "The risk is not over-investment — it's being caught without sufficient computing power as AI capabilities advance."

Domestic GPU Adoption Accelerates

Huawei's Ascend 910B AI accelerator, manufactured by Semiconductor Manufacturing International Corp. (SMIC) using 7-nanometer process technology, has become the primary alternative to restricted Nvidia chips. Huawei is projected to ship approximately 200,000 Ascend units in 2026, up from an estimated 80,000 in 2025.

However, the Ascend 910B delivers approximately 70% of the performance of Nvidia's H100 at similar power consumption, and significantly trails the H200 and B200. Software ecosystem limitations — Huawei's CANN framework is less mature than Nvidia's CUDA — add friction to adoption.

Data Center Construction Boom

Data center construction is underway across China's western provinces, where electricity costs are lower and land is abundant. Alibaba is building a 500-megawatt campus in Zhangbei, Hebei. Tencent is expanding its Guizhou facility to 300 MW. ByteDance has broken ground on a 400-MW complex in Inner Mongolia that will be powered primarily by wind and solar energy.

China's total data center capacity is projected to reach 55 gigawatts by end-2026, up from 35 GW in 2025, according to estimates from the China Academy of Information and Communications Technology.

AI Model Competition

The infrastructure investments support an intensifying competition in large language model development. Alibaba's Qwen 2.5 model, released in January, scored competitively with Meta's Llama 3 and OpenAI's GPT-4o on standard benchmarks. DeepSeek, a startup backed by quantitative trading firm High-Flyer, released its V3 model with strong performance-per-compute efficiency, using a mixture-of-experts architecture trained on Huawei accelerators.

Baidu's Ernie 4.0 Turbo and Tencent's Hunyuan Pro are targeting enterprise customers in finance, manufacturing, and government services, where data sovereignty requirements favor domestic AI providers.

Investor Concerns

Some investors have expressed concern about the pace of spending. Alibaba's stock fell 2.4% following its capex disclosure, and analysts at Jefferies warned that the "AI arms race" could pressure margins for multiple quarters before generating meaningful revenue returns.

Cloud AI revenue across the four companies totaled approximately $8 billion in 2025, representing a modest return on the $20 billion invested. Management teams uniformly argued that AI infrastructure investment is a prerequisite for long-term competitive positioning and that revenue lags were expected.