Alibaba had quite a bit to say at its Apsara conference the other night. It introduced a new AI chip and laid out a target to take Alibaba Cloud’s global data center capacity past 20GW by 2032.
For perspective, that’s roughly one-fifth of the entire world’s estimated data center capacity in 2025.
As someone who already thinks cloud is the most exciting part of this company, I liked what I heard.
I’ve been bullish on the infrastructure side of AI for a while, and Alibaba keeps giving me more reasons to think it could be one of the bigger beneficiaries. The opportunity to serve more customers while bringing down the cost of compute is a big part of that.
We’ve already started to see some encouraging progress. In the June quarter, AI Cloud and Compute Services revenue and external-customer revenue both grew 45% year over year. The segment’s adjusted EBITA margin reached about 11.6%, up from roughly 7.2% a year earlier on the restated comparison. As I wrote in my last update, the margin was the number I cared about most.
I want cloud to become a much larger contributor to Alibaba’s earnings, and seeing profitability improve alongside that growth makes me more confident it can.
With growth like that, I can understand why management wants to add so much capacity. There is still a lot to build between now and 2032, and we don’t have a price tag or a yearly build schedule for the 20GW target yet. I expect this will mean more capex, which I’ll come back to. But the chip announcements also give us something to think about on the cost side.