AI capital spending versus oil and gas

AI capital expenditure reached $765 billion in 2026, surpassing oil and gas capex of $681 billion for the first time, with projections of nearly double that level by 2031. UBS projects hyperscaler AI infrastructure spending will total roughly $4.1 trillion between 2026 and 2028, more than three times the combined spending of the previous six years. Google, Meta, Microsoft, and Amazon are the four largest spenders, with three-year totals of $938 billion, $683 billion, $672 billion, and $628 billion respectively, while neoclouds like CoreWeave and Nebius are also expanding rapidly. Alibaba and Tencent together spent $18 billion on AI infrastructure in a single quarter, a pace that cut Alibaba's quarterly profit 75%. Supply constraints span memory, power equipment, and land; HBM from all three major makers is sold out through 2027 or 2028, and Nvidia is taking a minority stake in data center site developer Cloverleaf to shape infrastructure decisions before hardware is selected. Bloomberg reported a widening divergence between falling AI model access costs and rising infrastructure costs, and investor Chamath Palihapitiya argued that a doom narrative, bipartisan political backlash, and rising Treasury yields are combining to choke data center expansion, with frontier labs most exposed.