Heading 1

The Blue Sweep changes the AI question.

It is no longer just what compute costs or how it gets financed. It is whether the build-out can be permitted, powered and delivered in 2027 and beyond. Moratoriums, local opposition and tighter regulation could all cap supply. That redraws the winners and losers.

The logic runs in two directions. Picks and shovels suppliers may have over-ordered against power that won’t arrive. Compute that already exists gets scarcer, and more valuable. So the question for a portfolio is simple: does your exposure sit with those building capacity, or those already owning it?

We ran 27 names through the Quant Insight AI risk score. Chips score 1, the average S&P 500 stock 0.

Each score measures a stock’s exposure to the macro forces behind AI — real yields, metals and risk appetite — and counts only where the equity market has actually traded that name as AI, rather than where a narrative says it should.

What the screen shows is not where most investors would expect the risk to sit.

Power and cooling now carry more AI risk than chips. Vertiv scores 1.51 and Comfort Systems 1.03, both above the chip benchmark. The biggest chip makers are the least geared. NVIDIA 0.41 and Broadcom 0.39, while AMD, Marvell and Micron carry roughly three times as much.

Neoclouds still building out are the most geared group in the screen, at 1.3 to 1.8. Hyperscalers average 0.26 and barely trade with AI at all. Oracle is the exception, trading with the builders rather than the owners.

Power generators are geared through real yields. That is the cost of financing the build-out, not the metals going into it.

If supply gets throttled, the risk sits with those promising compute tomorrow, not those owning it today.

‍