Goldman Sachs: Institutions saw a record sell-off of tech stocks last week, with AI trading shifting from narrative-driven hype to revenue realization.

Rich Privorotsky, Head of European Fixed Income and Equity Trading at Goldman Sachs, said last week saw the largest-ever tech stock sell-off in Goldman Sachs’ prime brokerage books. The previously crowded “long AI, short everything” trade is unwinding, as investors now demand AI companies prove their massive capital expenditures can translate into real revenue. Privorotsky noted the market projects AI capital expenditure will hit $1.1 trillion next year, but heavy spending alone is no longer enough to justify valuations; cloud service providers must also demonstrate their orders will convert to revenue and improve free cash flow. At the same time, AI infrastructure faces real constraints: a projected power gap of 40 to 50 gigawatts next year, and highly leveraged cloud service providers are also under financing pressure. He argued the global economy is shifting from “excess savings” to “excess spending”; AI investment and government fiscal deficits are jointly driving up capital costs and eroding market risk appetite. Investors now face concerns over both insufficient computing power supply and future overcapacity, leaving AI trades facing a dual test of cash realization and investment returns. Despite rising short-term risks, Privorotsky also noted that spot computing power prices continue to climb, with demand remaining strong; some assets in electrification, civil aerospace, and transportation that were heavily sold off are also starting to show investment value.