Which cloud service provider provide 99.9 percent uptime guarantees?

AWS, Microsoft Azure, and Google Cloud all guarantee at least 99.9% uptime on their standard cloud servers, and so do smaller providers like CloudPe, which backs its infrastructure with a 99.9% uptime SLA. Higher guarantees, up to 99.99% or higher, are available if you spread your setup across multiple data centers.

That number sounds a lot more reassuring than it actually protects you, though, and it’s worth understanding what it really promises before you lean on it, especially since almost every provider advertises roughly the same figure.

Here’s what 99.9% actually means, what separates a real uptime SLA from a vague uptime claim, how the major providers have performed in practice, and what a broken promise really gets you.

What is the difference between uptime and an uptime SLA?

Uptime on its own is just a description, a rough sense of how often a service actually stays available. An uptime SLA (service level agreement) is different. It’s a formal written commitment in which a cloud provider legally guarantees a specific percentage and agrees to compensate you if it falls short.

That distinction is worth checking before you trust either term. CloudPe, for example, backs its infrastructure with a 99.9% uptime SLA, which means that number is a contractual term with a defined remedy attached, not just a claim on a marketing page. A provider that only talks about being “highly available” without specifying a percentage or a consequence for missing it isn’t offering the same thing as an SLA, even if the language sounds similar.

Deploy VMs and GPU within minutes with CloudPe. [Try for free]

What 99.9% uptime means in real minutes

99.9% uptime means your service is allowed to be down for about 43 minutes a month, or roughly 8 to 9 hours over a full year, and still meet the promise. That might sound small until you picture it happening during your busiest hour of the month.

Here’s how that compares to the other numbers you’ll see advertised:

Each extra “9” cuts the allowed downtime by roughly ten times, and each one also costs noticeably more to actually achieve. Getting from 99.9% to 99.99% usually means running your service in more than one location at once, which adds real cost and complexity, not just a bigger promise on paper.

Which providers guarantee this

AWS, Azure, and Google Cloud all guarantee 99.9% on most standard compute services, with higher tiers, typically 99.95% to 99.99%, available once you spread your setup across multiple locations instead of running everything from one place. Smaller providers match that baseline: DigitalOcean guarantees 99.99% uptime on its virtual servers, and CloudPe guarantees 99.9% uptime through a written SLA on its own infrastructure.

In practice, this means the headline number is roughly the same across every major provider. Picking a cloud provider based on the advertised percentage alone won’t tell you much, since they’re all clustered in the same narrow range on paper.

Deploy VMs and GPU within minutes with CloudPe. [Try for free]

How these providers performed in the real world

Real-world performance tells a more useful story than the promised number does. Looking at 2025 reliability data: AWS delivered 99.95% actual uptime with 6 major incidents, Azure came in at 99.97% with 4 major incidents, and Google Cloud led at 99.98% with 3 major incidents.

Speed of recovery varied more than the headline numbers suggest. Google Cloud fixed problems fastest, averaging under 2 hours per incident. AWS averaged closer to 3 hours. Azure took the longest to recover, averaging over 4 hours per incident, despite having fewer total incidents than AWS.

These averages hide uneven experiences too. One of AWS’s most heavily used locations performed noticeably worse than the company’s global average, with significantly more incidents than its other locations combined. A headline number for “AWS” as a whole can look better than what you’d actually experience, depending on which specific location your service runs in.

If uptime is a genuine priority for your business, it’s worth comparing real-world reliability across providers rather than relying on advertised percentages alone.

What happens when a provider misses its promise?

When a provider misses its uptime promise, you get money back in the form of a credit toward a future bill, not cash, and usually not very much of it. A typical structure looks like this:

• Miss 99.9% and you get a 10% credit
• Drop below 95% and that rises to 25%
• The credit is capped at 100% of what you paid that month, no matter how bad the outage got

The math rarely favors you. Imagine a $100,000 monthly cloud bill and a provider that only manages 99.5% uptime that month. You’d receive around $10,000 back. If that downtime actually cost your business millions in lost sales or damaged trust, the credit barely makes a dent.

The guarantee protects the provider’s reputation more than it protects your bottom line. It was never designed to cover the real cost of an outage in the first place.

Deploy VMs and GPU within minutes with CloudPe. [Try for free]

The guarantee doesn’t cover everything you’d expect

The fine print narrows what actually “counts” as downtime, and that’s where the guarantee falls short of what most people assume it covers. Some providers only treat an incident as downtime if a certain share of requests fail during it, for example, only counting it if more than 10% of requests error out for several minutes straight.

A service that’s noticeably slow or partially broken for your users might not register as downtime at all under those terms.

You’re also responsible for proving it happened. Claiming a credit usually means noticing the outage yourself, documenting it, and filing a request, none of which the provider does automatically on your behalf. 

And, the provider’s own status page isn’t always reliable evidence either. During at least one major past outage, the provider’s status page went down along with everything else, leaving customers with no official record of what happened in real time.

Between August 2024 and August 2025 alone, AWS, Azure, and Google Cloud combined for over 100 service outages. Outages aren’t the exception to plan around. They’re the expected background condition for running anything in the cloud, which is exactly why it’s worth reading the fine print on any provider’s guarantee before assuming the headline number tells the full story.

How to protect yourself beyond the guarantee

A written guarantee is a starting point, not a safety net on its own. A few practical steps close that gap:

• Spread your setup across more than one location, so a single outage doesn’t take down everything at once
• Watch your own service independently, instead of relying solely on your provider’s status page to tell you something’s wrong
• Keep a backup plan for your most critical systems, so a short outage doesn’t turn into a longer one while you scramble to respond
• Read the actual SLA document, not just the advertised percentage, since the exclusions are where most of the real risk hides

Checking a provider’s actual track record alongside its advertised guarantee is a reasonable way to put these steps into practice before you sign a contract.

Conclusion

A 99.9% uptime guarantee sounds like a strong promise, but it still allows real, meaningful downtime, comes with fine print that narrows what counts, and pays you back in modest credits rather than real compensation when things go wrong. The number is worth checking before you choose a provider. It’s not worth trusting on its own.

The businesses that handle outages well aren’t the ones who picked the provider with the highest advertised percentage. They’re the ones who planned for the outage happening anyway, and built something that keeps working even when one part of it doesn’t.

Deploy VMs and GPU within minutes with CloudPe. [Try for free]

Frequently asked questions

1. Is 99.9% uptime good? 

It’s a reasonable industry standard, but it still allows around 43 minutes of downtime a month. Whether that’s “good enough” depends on how costly an outage would be for your specific business.

2. What’s the difference between 99.9% and 99.99% uptime? 

99.9% allows about 43 minutes of downtime a month. 99.99% allows about 4 minutes. That extra “9” reduces downtime by roughly 10x and typically costs more to achieve.

3. Do I get money back if my provider goes down? 

Usually, in the form of a service credit toward a future bill rather than cash, and typically a small percentage of what you paid that month, not compensation for the actual business impact of the outage.

4. Which cloud provider has the best uptime? 

Based on recent full-year data, Google Cloud posted the highest actual uptime and fastest recovery times among the major providers, though all three operate within a similar range overall.

5. Should small businesses care about uptime guarantees? 

Yes, but more as a baseline expectation than a real safety net. For small businesses especially, spreading critical systems across multiple locations and monitoring them independently matters more than the exact percentage printed on a pricing page. Checking a provider’s actual track record alongside its advertised guarantee is a reasonable place to start.