芝加哥以西约 45 分钟的 Aurora 数据中心是 CME Group 的核心基础设施，每日承担的衍生品名义交易量估值高达 25 千万亿（quadrillion）美元。该中心面积 45 万平方英尺，自 2009 年投入使用以来一直是全球最大期货交易平台 Globex 的主机所在，吸引高频交易公司在周边数十米内布设天线以争夺微秒级延迟优势。2016 年 CME 将该设施出售给 CyrusOne，并签署 15 年租约继续托管计算系统，形成外包运营模式。CyrusOne 现隶属于 2021 年以 114 亿美元收购其的 KKR + GIP 私募财团，后者 2025 年又达成 400 亿美元的 Aligned Data Centers 收购，显示其在数据中心资产上的扩张力度。
此次事故源于 冷却系统故障，导致 CME 所有主要期货与期权平台几乎全面停摆。尽管 CyrusOne 官方声称拥有冗余冷却单元，EY 专家指出此类 Tier-级数据中心“通常应具备高度冗余”，此次失效暗示潜在设计缺陷。CME 灾备计划本应切换至纽约区的数据中心，但因最初预估恢复时间短而选择留在 Aurora 重启撮合引擎，最终因误判导致全球市场多品类交易中断。
CyrusOne 的模式依赖“整库租户”——如 CME 这类占据数据中心几乎全部容量的客户，并在美国、英国、德国、新加坡共运营 约 50 座数据中心。业内猎头与顾问均将其视为“头部玩家”。然而本次事件暴露出外包关键基础设施的系统性脆弱性：高度集中化（单点 Aurora）与冷却冗余失效叠加，使全球股票、外汇、债券、大宗品交易链条受到冲击。CME 曾在 2016 年辩称选择 CyrusOne 可在必要时迁往其“29 座可用数据中心”，但本次 outage 显示实际切换能力具有显著限制。
About 45 minutes west of Chicago, the Aurora data center—a 450,000-square-foot facility—serves as the digital backbone of CME Group, processing an estimated $25 quadrillion in notional trade volume per day. Since CME built the site in 2009, it has become a latency battleground for high-frequency traders, prompting firms like DRW and Jump Trading to install antennas within meters of the center. CME sold the facility to CyrusOne in 2016 while signing a 15-year lease to continue hosting its Globex matching engines, effectively outsourcing critical operations. CyrusOne was acquired in 2021 for $11.4 billion by KKR and Global Infrastructure Partners, which later pursued a separate $40 billion deal for Aligned Data Centers—evidence of private equity's aggressive expansion into digital infrastructure.
A cooling-system failure triggered a near-total outage of CME’s futures and options markets, despite CyrusOne’s advertised redundant chillers. EY analysts noted that Tier-level data centers normally maintain extensive redundancy, implying a possible design flaw. Although CME’s disaster-recovery plan includes failover to a New York–area site, the exchange chose to restart in Aurora based on early signals that the cooling issue would resolve quickly—an assessment that proved overly optimistic and prolonged the global disruption.
CyrusOne’s model revolves around “single-tenant dominance,” with clients like CME consuming almost all available capacity across a portfolio of roughly 50 data centers in the US, UK, Germany, and Singapore. Industry specialists describe CyrusOne as a major global player, yet the outage exposes structural vulnerabilities inherent in outsourcing mission-critical financial infrastructure. Despite CME’s 2016 claim that the partnership would allow migration to “29 available facilities,” the incident demonstrates that practical failover options were far more constrained than advertised.