Cloud Outage Analysis
GCP Outage History - Google Cloud Outages & Real Losses
Updated October 2026 · 9 major incidents documented · Cost sources re-audited October 2026
Looking for live status?
If you are asking whether Google Cloud is down right now, go to status.cloud.google.com, the Google Cloud Service Health dashboard. It is the operator's own live record of open incidents by product and region, and no third-party page can be more current than it. If you are signed in, Personal Service Health in the Cloud Console goes one step further and tells you whether an open incident actually touches your projects, which the public dashboard cannot.
This page is the other half of that question: what has already happened, how long each event ran, and what an outage of that length costs. Reviewed 8 September 2026.
GCP Outage History, In Brief
Google Cloud has had at least 9 major documented outages since 2016, and 4 of the 9 were global or all-region events - a consequence of GCP's global control plane. The most recent was September 1, 2026, when routine hardware maintenance in us-central1 (Iowa) went wrong: a technician was told to disconnect all fiber paths at once instead of one router at a time, collapsing inter-datacenter capacity in the us-central1-b zone and degrading 19 products for about 4h 11m. It came just twelve days after a similar regional network failure in us-west1 on August 20, 2026 (a failed automated reroute during fiber maintenance, 33 products, ~2h 22m). The most recent global event, and the most severe in years, was June 12, 2025, when an invalid automated quota policy update with blank fields crashed Service Control and broke IAM authorization globally for about 3 hours (us-central1 up to about 2h 40m), hitting dozens of products and cascading to Cloudflare, Spotify, Discord, and OpenAI. The longest single event was April 2023, when a water leak and battery fire at a Paris data center took the entire europe-west9 region offline for more than 24 hours, with some services degraded for weeks. The June 2, 2019 network-congestion outage lasted about 4 hours and took YouTube, Gmail, Snapchat, and Discord down across US regions, and the famous April 11, 2016 event dropped all Compute Engine connectivity worldwide for 18 minutes. GCP has fewer major incidents than AWS us-east-1, but they scale worldwide fast. Full chronology and SLA credit math below.
GCP Outage & Downtime Statistics
9
Major outages since 2016
~1/yr
Frequency (2019-2026)
4 / 9
Global / all-region scope
24h+
Longest (Apr 2023, Paris zone)
~4h 11m
Most recent (Sep 2026, us-central1)
Frequency reflects the 9 major incidents documented in the chronology below, not every status-page event. Scope counts documented major outages: 4 of 9 (Jun 2025, Nov 2021, Mar 2019, Apr 2016) were global or all-region; the others were confined to one region or zone (Sep 2026, Aug 2026, Apr 2023, Jun 2023) or a cluster of US regions (Jun 2019). Google has had smaller incidents not listed here.
GCP Outage History
| Date | Region | Services | Duration | Impact | At risk / $100M revenue |
|---|---|---|---|---|---|
| Sep 2026 | us-central1 (Iowa, us-central1-b) | Inter-datacenter fiber (maintenance); 19 products incl. GCE, GKE, Cloud SQL, Spanner, BigQuery, Cloud Run, VPC | ~4h 11m (07:41-11:52 PDT) | Fiber paths unplugged at once; 100% traffic drop at peak, VMs unreachable across the zone | $47,700 |
| Aug 2026 | us-west1 (Oregon) | Inter-datacenter network (fiber maintenance); 33 products incl. GCE, GKE, Cloud SQL, BigQuery, Cloud Storage, IAM | ~2h 22m (08:00-10:22 PDT) | Regional congestion after failed auto-reroute; timeouts/errors across 33 products | $27,100 |
| Jun 2025 | Global | Service Control / IAM; dozens of products (BigQuery, Firestore, Vertex AI, Gmail, Drive) | ~3 hours (us-central1 up to ~2h 40m) | Global API auth failures; cascaded to Cloudflare, Spotify, Discord, OpenAI | $34,200 |
| Apr 2023 | europe-west9 (Paris) | GCE, Cloud Run, Cloud SQL, BigQuery (data-center water leak + fire) | Zone offline 24h+ (~4-5h peak) | Entire French region down; some services degraded for weeks | $274,000 |
| Jun 2023 | us-east5 | IAM auth; Cloud SQL, Spanner, Bigtable, Pub/Sub | ~49 minutes | Regional auth failures across data services | $9,400 |
| Nov 2021 | Global (Cloud LB) | Cloud Load Balancing (config error) | ~2 hours | Spotify, Snapchat, Discord, Etsy hit by 404s | $22,800 |
| Jun 2019 | US multi-region | Network congestion; Compute, GCS, YouTube, Gmail | ~4 hours | Elevated packet loss; YouTube, Gmail, Snapchat, Discord | $45,700 |
| Mar 2019 | Global (all regions) | Cloud Storage (internal blob storage) | ~4 hours | GCS + Gmail/Photos attachments degraded (<6% error rate) | $45,700 |
| Apr 2016 | Global (all regions) | Compute Engine (external connectivity) | 18 minutes | All GCE instances lost inbound connectivity after a bad IP config rollback | $3,400 |
Dates, durations, root causes, and affected services are taken from Google's published incident reports (status.cloud.google.com) and contemporaneous reporting (The Register, DCD, ThousandEyes). The last column is ours, and it is arithmetic you can redo: $100,000,000 of annual revenue divided by the 8,760 hours in a year is $11,416 an hour, multiplied by the event's duration (24 hours used for the Paris event, which ran longer). It is a gross upper bound - it assumes the outage took you fully offline for its whole length - and it is not an estimate of what the event cost Google, its customers in aggregate, or anyone else. Use the calculator to scale it to your own revenue and add margin, recovery and SLA credits. Updated October 2026.
What a Google Cloud Outage Cost: Nobody Has Published a Figure
Google's incident reports are detailed on cause, timeline and affected products, and silent on money: they carry no customer-impact figure, and Google has never published one for a Google Cloud outage. We have also found no named analyst or insurance-modelling estimate for any of the nine incidents above, including the June 2025 Service Control event that broke IAM globally. That matters when you are reading an outage-cost number somewhere else: for AWS there are exactly two such published estimates in fourteen years (Cyence on the 2017 S3 outage, CyberCube on October 2025, both set out here), and for Google Cloud there are none. A confident dollar figure attached to a GCP outage is somebody's model, not a disclosure - so the useful question is whose model, over which inputs. Ours is in the column above, and you can redo it in one division.
GCP SLA Credit Rules
| Monthly Uptime % | Compute Engine Credit | GKE Credit |
|---|---|---|
| 99.95%+ | No credit (SLA met) | No credit |
| 99% - 99.95% | 10% | 10% |
| 95% - 99% | 25% | 25% |
| <95% | 50% | 50% |
Credits applied as GCP billing credits. Claims within 30 days. Full credit vs loss analysis.
Frequently Asked
Is Google Cloud down right now?
When was the most recent major GCP outage?
How often does Google Cloud have outages?
What was the worst GCP outage?
How reliable is Google Cloud Platform?
How do GCP SLA credits work?
Related
GCP September 2026 Case Study
us-central1 fiber-disconnection outage: root cause, timeline & cost
GCP August 2026 Case Study
us-west1 fiber-maintenance outage: root cause, timeline & cost
GCP June 2025 Case Study
Root cause, timeline & cost of the Service Control outage
AWS Outage History
Compare AWS reliability
Azure Outage History
Compare Azure reliability
Cloudflare Outages
The Jun 2025 cascade downstream
Calculate Your Exposure
Your GCP outage cost