2025 was a reminder that "the cloud" is still someone else's data center, and that data centers have bad days. Across the three major providers, several multi-hour regional events took large swaths of the internet offline. Every one of those events had an SLA behind it, and every one should have produced credits for the customers in the blast radius. Most of that money was never claimed.
This recap looks at the year's pattern of outages, what the SLAs owed, and how to be faster in 2026. To see current outage activity as it happens, Cloud Downtime tracks live provider status.
The shape of a 2025 outage
The individual incidents varied, but the pattern was consistent. A control-plane or networking failure in a single region cascaded, dependent services went dark, and recovery took hours rather than minutes. Because so many customers concentrate workloads in a handful of popular regions, a single regional event rippled across thousands of businesses at once.
That concentration is exactly what makes SLA credits worth chasing. When one region breaks, everyone in it has the same claim at the same time, and the tier math is identical for each of them.
What the SLAs owed
Recall the standard structure of a compute-class credit table:
| Measured monthly uptime | Typical credit |
|---|---|
| 99.0% to below 99.99% | 10% |
| 95.0% to below 99.0% | 25% |
| Below 95.0% | 100% |
A multi-hour outage matters more than it sounds. A single four-hour outage in a 30-day month leaves uptime around 99.4%, which clears the 99.99% target and lands squarely in the 10% credit tier. Stack a second incident in the same billing month, or a longer single event, and you can drop into the 25% tier, which more than doubles the payout.
On a $40,000/month service line, that is $4,000 at 10% and $10,000 at 25%, per affected service. When a regional event hit several of your services in one month, each generated its own credit. The real dollar math shows how these figures scale with spend.
The recurring themes from 2025
Reading across the year's incidents, a few lessons repeat:
- Single-region concentration is the risk multiplier. Teams that ran everything in one region felt every regional outage at full force. It also, per many SLAs, weakened their claims, because some credits require multi-zone deployment to pay out. Our fine-print guide covers that trap.
- Providers do not always post a clean public timeline. Some 2025 incidents had vague or delayed status updates. That does not block a claim: your own monitoring data is valid evidence, and the SLA measures your resources, not the provider's press releases.
- Recovery time is not claim time. The service came back the same day. The claim window did not reset. Roughly 60 days for AWS and Azure, 30 for Google Cloud, from the incident. A lot of 2025 credits quietly expired unclaimed.
Why so many 2025 credits went unclaimed
The blocker was never the money being small. It was three familiar failures:
- No one was watching for the breach in real time, so the incident was noticed late or not at all.
- Evidence aged out while teams focused on recovery instead of capturing status history and monitoring graphs. Our evidence checklist fixes this.
- The uptime math felt like a chore, so the claim never got calculated. It is arithmetic: total minutes minus impacted minutes, over total minutes.
How to be faster in 2026
The playbook is short and pays for itself the first time a region hiccups:
- Log every incident the day it happens, with timestamps, so the claim window is on someone's calendar.
- Capture evidence immediately, before dashboards roll off.
- Compute measured uptime for the affected month and find the tier.
- File within the window with the exact figures, framed as a service credit. Our filing guide and how-it-works overview cover the submission steps.
The takeaway from a rough year
2025 proved the credits are real and recurring: regional outages happen, they breach SLAs, and they owe customers a percentage of the affected bill. The businesses that collected were simply the ones who noticed in time, kept their evidence, and filed before the window closed.
To avoid repeating the unclaimed-money story in 2026, automate the watching. Cloud Credits handles the billing-accuracy side of provider money, Cloud Downtime shows live outage data, and Next Signal (our sponsor) detects breaches, captures evidence, and drafts the claim automatically. Next time a region goes down, run the credit math the same day and file.
