AI in action · Markets
AI Build-Out Stocks Are Down 23% Since the S&P 500 Record
The companies that build and power AI data centres have fallen a median 23% since the S&P 500's record close on October 6, Yahoo Finance calculated. Strong results have not helped: Bloom Energy and Comfort Systems dropped 11% and 9% on October 23.
For most of 2026 the safest-looking way to own the AI boom was to own the companies selling it picks and shovels: the firms that build data centres, wire them, cool them and power them. In the past three weeks that trade has gone into reverse faster than the chip stocks themselves. Yahoo Finance's Jared Blikre calculated on October 23 that data-centre builders and equipment makers, a group including Caterpillar, GE Vernova, Vertiv and Eaton, had fallen a median 23% since the S&P 500's record close on October 6. Power producers were down a median 14%, Nvidia 11% and the big technology companies doing the spending only 5%.
The ordering is the point. The firms furthest from the end customer, whose revenue depends on the hyperscalers continuing to write cheques, have been hit hardest, while the spenders themselves have held up best. A dozen of the largest builder and power names have together lost more than $300 billion of market value since the October 6 peak, according to the same analysis.
Results are not stopping the slide. Bloom Energy, which sells on-site fuel-cell power to data centres, reported sales up 138% and raised its full-year growth outlook to about 122%, then fell 11% on Friday, its second daily drop in a row; its cash flow also turned negative in the quarter. Comfort Systems, a mechanical and electrical contractor, said its backlog had nearly doubled to $17.8 billion and lost 9%. Credo, which makes high-speed connectivity chips, also fell 9%. A day earlier Lam Research, the chip-equipment maker, dropped 11.8% after beating profit and revenue forecasts, the Associated Press reported, on a session when the S&P 500 had its worst day since July.
The demand case has not obviously weakened. Apollo's chief economist Torsten Sløk told Yahoo Finance that demand for computing power is significantly greater than supply, and that lead times for transformers now stretch up to 26 months. Jefferies' Aniket Shah called power the biggest constraint on the build-out. What has changed is the price investors will pay for that demand, with worries about how the spending is financed and a 10-year Treasury yield that only slipped back below 5% on October 22.
This week offers the suppliers a direct read-through. Vertiv reports on Wednesday alongside Microsoft and Alphabet, and Amazon follows on Thursday. If the hyperscalers confirm or raise their capital spending plans, the gap between strong supplier results and falling supplier share prices becomes harder to justify; if they trim them, the 23% drop may turn out to have been the market moving first.
Sources: Yahoo Finance (Jared Blikre) — The AI sell-off is crushing stocks of companies leading the data center build-out (Oct 23, 2026); AP via WXXV 25 — Sharp drops for AI stocks overwhelm markets worldwide as oil prices jump (Oct 22, 2026)