Alibaba touts “most powerful AI chip in China” for data center buildout

The scale of Alibaba’s spending deserves attention before the chip superlative does. KrAsia reports that the company had spent RMB 190 billion by June against a three-year pledge of RMB 380 billion. That is money committed to building capacity, rather than a promise resting entirely on future fundraising.

Capacity is still the constraint in the longer-term plan. Alibaba wants more than 20 GW globally by 2032. Goldman Sachs estimated 3 to 4 GW last year, implying roughly 2 GW of additions each year. The target describes the size of the future network, not an annual amount already being delivered.

The V900 adds another question: how well the hardware will work in commercial use. Alibaba’s claim of three times its predecessor’s performance needs to be distinguished from independently established results. Large clusters sound impressive, but customers need useful computing capacity, not simply a large card count.

That distinction also matters when reading Citi analyst Alicia Yap’s USD 160 billion cloud revenue forecast for 2032. It depends on the buildout. Treating the revenue projection as proof of the chip’s success would reverse the argument: the deployment has to justify the forecast.