At the 2026 Apsara Conference, Alibaba Cloud’s annual event running in Hangzhou from September 22 to 24, the Chinese group unveiled its new artificial intelligence accelerator, the Zhenwu V900, developed by its semiconductor unit T-Head.
According to specifications released by the company, the new chip delivers three times the performance of its predecessor, the M890, with 216 GB of memory (up from 144 GB) and inter-chip bandwidth of 1,200 GB/s (up from 800 GB/s), and can be linked into superclusters of up to 500,000 units, compared with 128 for the previous model. Mass production and commercialization are expected in the first quarter of 2027. CEO Eddie Wu called it «China’s most powerful AI chip».
The group also signaled that the next generation of its Qwen model family (the current flagship, Qwen3.8-Max, has roughly 2.4 trillion parameters) will scale up to a range of 5 trillion to 10 trillion parameters: Qwen4 is «in training», with the 4.5 and 5 versions expected to follow.
Alibaba’s 20-gigawatt plan and its expansion beyond China
The centerpiece of the event, though, was the infrastructure plan: Alibaba is aiming to exceed 20 gigawatts of global data center capacity by 2032, opening its first cloud regions in Turkey, Finland, and the Netherlands over the next 12 months, with further expansion planned in Malaysia, Germany, the United Arab Emirates, France, and Hong Kong.
It is the first time the group has pushed so explicitly onto the turf of Amazon Web Services and Google Cloud outside Asia. The plan builds on a $53 billion, three-year commitment to AI investment announced in 2025. Citigroup analysts, according to Benzinga, estimate that the 20-gigawatt network could eventually support more than $160 billion in external cloud revenue over time — an estimate from the investment bank, not a figure confirmed by the company.
Wall Street cheers, then thinks twice
The stock market’s reaction went from enthusiastic to cautious. On Tuesday, September 22, shares rose about 3% both in Hong Kong, where Alibaba stock (9988.HK) led the market’s gains, and on its New York ADR (NYSE: BABA). But in Wednesday’s premarket session, September 23, the stock reversed course, falling roughly 3.6% and dropping to around $112. Analysts remain broadly positive overall — consensus is «Buy»/«Strong Buy», with an average price target near $190 — but the market is clearly weighing the sheer scale of the spending commitments announced.
Why are investors pumping the brakes?
The sticking point is profitability. In the quarter ended June 30, 2026, revenue from the AI Cloud and Compute Services segment grew 45% year-over-year to 48.4 billion yuan (roughly $6.8 billion), the fastest pace in 22 quarters, with AI-product revenue alone growing at triple-digit rates for a twelfth consecutive quarter. But the group’s net profit collapsed 75% year-over-year to 10.4 billion yuan (roughly $1.5 billion), while adjusted (non-GAAP) net profit fell 38%, driven by AI infrastructure spending.
A plan that promises to further double that spending, extending it abroad, revives a question that has hung over Chinese and American tech giants alike for months: how long will it take for that capacity to convert into revenue and margins, and how much capital will be needed in the meantime.
Editor’s note
This article was originally published in Italian on money.it by Money.it Research Office on September 23, 2026 as «Alibaba svela il chip AI Zhenwu V900 e punta a 20 gigawatt di data center. Ma il rally a Wall Street dura poco». It has been translated and adapted for an international audience by the Money.it International desk.
