Goldman Sachs: China’s AI Infrastructure Outlook Revitalized as Enterprise Token Usage Soars, with Alibaba Poised to Benefit

Goldman Sachs’s latest analysis underscores a renewed investment narrative around China’s AI infrastructure, driven by an explosive uptick in enterprise demand for large-scale models. Full-stack cloud providers such as Alibaba are entering a period of accelerated growth as companies ramp up their adoption of advanced AI frameworks and corresponding computing resources.

Industry metrics reveal that Chinese enterprises averaged 10.2 trillion token requests per day in the first half of 2025—a 363 percent jump from the latter half of 2024—highlighting a rapid deployment of generative AI in commercial settings. According to Frost & Sullivan, this surge reflects the fastest pace of enterprise-scale AI integration the market has witnessed.

Goldman Sachs assigns Alibaba a favorable position based on its sophisticated in-house model development, commanding 47 percent of China’s public cloud sector, and its diversified access to domestic and foreign chip suppliers. These advantages underpin the bank’s decision to raise Alibaba’s 12-month price target from USD 163 to USD 179 while retaining a “Buy” recommendation. Within this valuation, Alibaba Cloud’s per-ADS estimate climbed from USD 36 to USD 43, and its growth outlook for fiscal 2026’s Q2 through Q4 was upgraded to a 30 – 32 percent range.

Analysts led by Ronald Keung point to China’s shift toward proprietary inference chips and a “multi-chip strategy” as a critical inflection point. By reducing reliance on imported semiconductors, cloud providers can secure a more resilient hardware foundation. Goldman Sachs forecasts that capital expenditure among Chinese cloud service operators will rise by 39 percent year-over-year in the third quarter of 2025, further cementing the industry’s capacity to support sustained AI-driven revenue gains.

The report likens China’s current phase to the U.S. market in late 2022, when infrastructure investments were rewarded following ChatGPT’s debut. As in that cycle, strategic spending on data-center build-out and computational power appears primed to attract investor interest ahead of mass enterprise adoption.

Beyond infrastructure, China’s AI model development and application landscape are advancing swiftly. On September 12, Alibaba unveiled Qwen3-Next, a next-generation architecture claimed to deliver tenfold performance improvements over its predecessor at one-tenth the development cost; its 80 billion-parameter Qwen3-Next-80B-A3B variant reportedly operates ten times faster than the 32 billion-parameter model launched in April. Baidu followed on September 9 with ERNIE X1.1, which the company states rivals GPT-5 and Gemini 2.5 Pro in factual accuracy, instruction compliance, and agent functionality. Meanwhile, Tencent introduced HunyuanWorld-Voyager for 3D environment generation and the high-resolution HunyuanImage 2.1 for text-to-image creation.

Generative AI is also being embedded as intelligent assistants within mainstream services. Meituan’s “Xiaomei” now supports voice-driven food orders and restaurant reservations, and Gaode Map by Alibaba offers “Teacher Xiaogao,” a personalized travel companion leveraging native AI guidance.

Despite these advances, monetization remains an ongoing challenge. As of August 2025, global annual recurring revenue (ARR) from AI applications stands at roughly USD 30 billion, whereas China’s domestic AI ARR totals only USD 1.5 billion—about 5 percent of the worldwide figure—indicating significant room for commercial growth.