Google Cloud outage knocks out major banks and apps

Immediate fallout

Google Cloud’s global load balancers returned 502 errors on Tuesday, forcing a cascade of failures across networking, App Engine and Stackdriver. The outage hit a range of customers, from fintech firms to consumer apps. HSBC, Revolut, Bank of Scotland, Lloyds and credit‑reporting agency Equifax reported service interruptions. Snapchat, Spotify and the Pokémon Go backend also went dark.

The incident began at 12:34 p.m. PT, when Google posted an investigation notice on its status dashboard. By 12:53 p.m. PT the company said a fix was in progress. At 1:03 p.m. PT Google declared the load‑balancer issue resolved for the majority of users. A Google spokesperson confirmed the resolution and promised a post‑mortem.

Technical trigger

Google identified the root cause as a malfunction in its global load‑balancing layer that produced HTTP 502 responses. The layer routes traffic to virtual machines and containers across regions. When the layer failed, inbound requests could not reach downstream services, causing the observed downtime.

The status page noted that the problem affected both standard and pre‑emptible VM instances in the europe‑west2‑a zone. Persistent Disk devices in that zone briefly ran in single‑redundancy mode. Google’s engineers shut down a subset of capacity to protect hardware, then restored capacity after the fix.

Impact on financial services

Banking applications rely on low‑latency API calls to process transactions, authenticate users and retrieve credit data. A load‑balancer outage breaks that chain, leading to login failures, stalled payments and delayed credit checks. HSBC and Revolut, both heavy users of Google Cloud for real‑time fraud detection, reported customer‑facing errors lasting several minutes.

Lloyds and the Bank of Scotland saw their online banking portals become unreachable. Equifax, which provides identity verification APIs, experienced degraded response times that could affect loan approvals. The outage illustrates the risk of a single cloud provider becoming a single point of failure for critical financial infrastructure.

Parallel cooling‑related incidents

A separate incident earlier in the week highlighted a physical‑layer vulnerability. Record‑breaking UK temperatures—peaking at 40.2 °C (104.4 °F)—triggered cooling system failures in Google’s europe‑west2‑a data centre. Google powered down part of the zone to avoid hardware damage, resulting in VM terminations and reduced redundancy for Persistent Disk volumes.

Oracle faced a similar cooling failure in its UK South (London) data centre. The company powered down non‑critical hardware after temperatures rose above design limits. Both providers restored services within 24‑48 hours, but the events underscore how climate‑driven hardware stress can translate into cloud outages.

Industry reliability concerns

The twin failures—software‑level load‑balancer bugs and hardware‑level cooling loss—renew debate over cloud‑provider resilience. Financial regulators have begun to scrutinize dependency on a handful of hyperscale platforms. Some banks are now revisiting multi‑cloud strategies to mitigate risk.

Google’s promise of a detailed analysis suggests internal changes may follow. Past incidents have led to tighter monitoring of load‑balancer health checks and expanded capacity buffers in heat‑prone regions. Yet the rapid escalation from a single 502 error to a multi‑service outage shows that detection and isolation mechanisms remain imperfect.

What to watch

Track Google’s forthcoming post‑mortem for specifics on the load‑balancer defect and any changes to its redundancy architecture. Monitor the rollout of enhanced cooling monitoring in European data centres, especially as UK heatwaves become more frequent. Finally, watch for announcements from major banks about diversifying cloud footprints or adding on‑premise failover layers.