The fee is real, but it leans on a single listing

The five biggest banks are forecast to report a 27% year-on-year rise in quarterly investment-banking fees to $11.1 billion, the highest since 2021. The number is impressive, but open the model and the concentration shows: US IPOs raised a record $104.8 billion in the quarter, and the SpaceX listing carried it. SpaceX's float alone generated $500 million in fees for 23 banks, with Goldman and Morgan Stanley each taking $100 million. When one mega-deal carries a quarter, that revenue line is less diversified than it looks.[1]

Global M&A topped $3 trillion in 2026, up more than 40% year on year, with AI 'picks-and-shovels' infrastructure firms dominating activity. Is that a durable fee stream or a reflection of the AI capex cycle? Nudge the discount rate or the appetite for AI a notch and the IPO calendar thins. The fee is real today; its persistence depends on how tethered demand is to a single theme.[1]

The valuation of staying private

Goldman's new 'alternative investments platform' sits exactly in that gap: wealthy clients and family offices will take direct stakes in individual private companies rather than in broad private-equity funds. The logic is clear — the most successful startups stay private far longer, and early investors capture most of the gains before public investors get in line. But for the analyst opening the model, the real question is pricing: without a public reference, on what basis will these direct stakes be marked?[2]

Yesterday I asked 'where did the cash go' and tracked the synergy claim of a $110 billion merger being tested in court; the lesson there was that the story is free and the capital isn't. Today the same lesson moves to private markets: Goldman is leaning into wealth management for steadier fees, but a valuation kept private stays untested when the IPO window shuts. What I'll watch next quarter is the price at which these direct stakes actually find a buyer in secondary hands — because the model, in the end, weighs the price too.[2], [3]