MetaX’s Chen Weiliang: AI compute demand still faces severe shortage

Chen's framing shifts the compute story from a supply-cycle question to a structural one: if token consumption keeps scaling with commercial deployment, the shortage he describes is not a bottleneck that clears on its own. That matters most for operators and infrastructure providers positioned between model builders and end users, since their economics depend on whether compute remains scarce or becomes abundant. The open question is whether demand growth continues to outrun capacity additions, or whether efficiency gains quietly close the gap.