“The cloud” sounds like something more reliable than a drive sitting on your desk. In reality, it is always a real building at a real address, and its reliability promises are usually about equipment failure, not about the building itself being destroyed.
Real Cases: Amazon and Yandex Both Lost Data Centers in 2026
In the spring and summer of 2026, Amazon Web Services (AWS) data centers in Bahrain and one of the UAE regions came under fire during military action in the region. AWS is the largest cloud provider in the world, used by thousands of US companies — and yet, according to media reports, it took Amazon months to even attempt to fully restore access, and some customer data was declared permanently lost. A similar picture played out in Russia in October 2026: a series of drone strikes knocked out three of Yandex’s data centers within a few days. About 64,000 Yandex Cloud customers were affected, and full recovery of analytics capacity is estimated to take months. Two different providers, two different regions, the same summer-to-fall stretch of 2026 — and both cases are real, confirmed by independent sources, not a hypothetical scenario.
Neither company can be called careless. Every AWS region is built from at least several physically separated zones — separate buildings with their own power and networking — and AWS itself officially recommends spreading data across at least two zones at once. Yandex Cloud uses a similar model: a region is made up of several availability zones, physically located across four different data centers. Both companies built their infrastructure specifically so that one site failing wouldn’t take everything down with it — which is exactly why it’s so telling that several zones were hit at once in both cases: this isn’t a single building’s outage anymore, it’s an event beyond what that kind of redundancy is designed to protect against.
What a Data Center Tier Rating Actually Means
The industry rates data center reliability on a Tier scale from I to IV, with each level carrying a stated annual uptime percentage and an approximate annual failure risk:
| Tier | Annual uptime | Approx. annual failure risk |
|---|---|---|
| Tier I (data center) | 99.671% | ~0.3% |
| Tier II (data center) | 99.741% | ~0.26% |
| Tier III (data center) | 99.982% | ~0.02% |
| Tier IV (data center) | 99.995% | ~0.005% |
| Home laptop | ~80-94% | ~6-20% per year, rising with age |
The higher the tier, the more redundancy is built in — duplicate power sources, cooling systems, and network paths — so that one component failing doesn’t stop the entire facility.
Even Tier IV Was Never Built for This
Even Tier IV, with its 99.995% uptime, was never designed around events on the scale of a military strike or a building’s direct physical destruction. The uptime percentage in this system describes how reliable the infrastructure inside the facility is — power, cooling, network paths — not protection from external, catastrophic events. So even a data center with the highest possible reliability rating doesn’t guarantee the data inside it stays accessible under every circumstance — which is exactly what the AWS and Yandex cases showed.
Why a Data Center Is Still More Reliable Than a Home Drive
Even the lowest tier, Tier I, with its 99.671% uptime and roughly 0.3% annual failure risk, is still a specific, measurable commitment: backup power, temperature control, physical security, and staff monitoring the equipment around the clock. A home laptop is a completely different picture — roughly 80-94% uptime, with a 6-20% annual failure risk that climbs as the drive ages. Even the lowest-rated data center is an order of magnitude more reliable than an ordinary drive at home. This isn’t about whether the cloud can be trusted — it’s simply a difference in approach: building-level redundancy versus a single unprotected drive.
The Real Lesson: Don’t Keep Only One Copy
The right takeaway from the Amazon and Yandex cases isn’t “don’t trust the cloud” — it’s “don’t keep only one copy of your data,” no matter how reliable the place storing it looks. If the only copy of your important files lives in a single data center or on a single home drive, any incident at that one location, however unlikely, can be the last thing that ever happens to that data. The sensible approach is to keep more than one copy, in different places and on different types of storage, so that no single incident can wipe out everything at once.
If you’re not sure where your backups actually live, or you need help setting up backup the right way — message me, I’ll help.
