On November 28, 2025 a chiller plant failure at CyrusOne's CHI1 data centre in Aurora, Illinois — about 50 miles from Chicago — overheated the facility that hosts CME Group's exchange, sending temperatures above 100°F.
CME halted trading in global futures — gold, oil and interest rates — for more than 10 hours, affecting markets from Tokyo to London. S&P futures reopened down 0.1% when trading resumed.
CME chose not to fail over to its New York disaster-recovery site, believing the outage would be brief. The decision extended a manageable cooling fault into a global market event.
A chill, invisible failure — a failed chiller — that stopped the world's futures markets for half a day.
CME skipped its New York disaster-recovery site, betting the outage was brief, and turned a cooling fault into a 10-hour global market halt.
CME's decision to stay on the failing primary site rather than switch to the NY disaster-recovery site assumed the outage would be short — a judgment made on outage duration, not on risk.
CyrusOne's single chiller plant at CHI1 had no resilient redundancy, so one plant failure took the whole cooling load down with the facility hosting a globally critical exchange.
A 10-hour halt of global futures is the cost of placing an exchange's continuity on one facility's assumption that cooling problems are short-lived.
The outage was operational: a mechanical failure upstream, and a failover decision downstream that converted it into a market-wide stop.
A disaster-recovery site is only useful if you use it. When a host's cooling fails, betting on a short outage is a bet on someone else's machinery — a global market stop is the price of that bet.
Trading resumed after more than 10 hours, with S&P futures opening slightly lower. The failure exposed how much of global futures trading rests on the cooling of one facility.
FOLLOW THE EVIDENCE
The sources
- Insurance Journal — CME global futures halted by data centre cooling failure insurancejournal.com