Why AI hyperscalers are now the epicenter of a bear case for stocks

The key second-order shift is not just “AI capex is high,” but that the financing regime for hyperscaler spend may be breaking. Once the market starts treating frontier compute as a balance-sheet activity rather than an operating expense, equity dilution becomes part of the cost of AI, which compresses long-duration growth multiples across the whole complex. That is especially bearish for the names most levered to a “capex forever” narrative because they have the least flexibility if customers pause or financing windows close.

The most vulnerable subset is the picks-and-shovels cohort with the cleanest beta to incremental data-center buildout, since these stocks trade on the assumption that spend can only accelerate. If hyperscalers are forced to issue equity or slow projects, the market will first de-rate the suppliers whose order books are most exposed to a single end-market, then reassess whether backlog is truly high-quality or just deferred pricing. Intel is the odd exception: if capital gets rerouted away from external AI supply chains toward onshore capacity and inference efficiency, it can benefit from substitution and localization, but only if execution is credible over the next 2-4 quarters.

On the macro side, stronger labor data reduces the immediate odds of easier policy, which matters because the whole AI equity story is being funded by a historically low discount-rate assumption. That creates a nasty setup: higher-for-longer rates plus rising capex intensity is a multiple-compression cocktail, particularly for megacap growth where expectations already assume durable margin expansion. The defensive consumer/health-care rotation is rational not because it offers excitement, but because it offers optionality: cash returns, lower duration, and less dependence on external financing conditions.