Monday, October 20th 2025

AWS Outage Takes Half of the Internet Down, Services Now Recovering

Amazon Web Services (AWS) outage has just taken down half of the known internet. However, services are now recovering. Just as Monday started on the Pacific Daylight Time (PDT), AWS reported incidents of multiple AWS services being down, with a simple confirmation that "We are investigating increased error rates and latencies for multiple AWS services in the US-EAST-1 Region." Online services like Amazon.com, Alexa, ChatGPT, Epic Games Store, Epic Online Services, social media like Snapchat, and games like Fortnite, have been down since.

A few hours later, AWS confirmed that it had isolated the issue. "We have identified a potential root cause for error rates affecting the DynamoDB APIs in the US-EAST-1 Region," they stated, adding, "This issue also impacts other AWS services in the US-EAST-1 Region. Global services or features that rely on US-EAST-1 endpoints, such as IAM updates and DynamoDB Global tables, may also be experiencing issues. During this time, customers may be unable to create or update Support Cases. We recommend customers continue to retry any failed requests." Subsequently, AWS confirmed that the "underlying DNS issue has been fully mitigated, and most AWS service operations are now succeeding normally."

Update 19:45 UTC: AWS is still experiencing some outage, but full recovery is in progress. The team stated "We continue to observe recovery across all AWS services, and instance launches are succeeding across multiple Availability Zones in the US-EAST-1 Regions."
As AWS works towards a full resolution, customers who haven't resolved their issues are advised to flush their DNS caches. At TechPowerUp, our sovereign infrastructure is intentionally designed to operate without relying on external cloud providers. When an AWS, Azure, or GCP outage occurs, any service dependent on them is affected. If our infrastructure fails, it is solely due to our own error. Hence, this AWS outage hasn't affected our server. We provide a technical issues forum thread to update the community during an outage, which you can check out here.
Sources: AWS Status, via The Verge
Add your own comment

54 Comments on AWS Outage Takes Half of the Internet Down, Services Now Recovering

#26
Aquinus
Resident Wat-man
WirkoHow many nines does AWS advertise?
Depends on the service. S3 doesn't start reimbursing until it falls below 99.9%, but only at 10% and it's credited to your account for forward charges. They don't give you your money back. The longer it's down, the more they'll reimburse. Other services probably have different SLAs.
aws.amazon.com/s3/sla/
Posted on Reply
#27
phanbuey
AquinusLMAO, not affected because we're on us-east-2 at work?

I wouldn't call AWS' infrastructure shitty by any stretch of the imagination. Cloud service providers are offering a trade off and services. You can run your own servers and everything, but then you are on the hook for running them, maintaining them, securing them, and scaling them to demand, not to mention sourcing your own platform services for things like queueing, async processing, distributed file storage, etc. A smaller software company might not have the resources of a huge ops team, a flexible data center, and a suite of platform services with tooling to match.

To be completely fair, cloud solutions don't make much sense if all you're doing is running servers through EC2 with AWS or GCE with Google. They shine when it comes to their distributed services. Take S3 for example. You literally get multi-region replication out of the box without having to care about any of the infrastructure to do it. Same thing with Dynamo. You can provision everything with CloudFormation and use that to deploy from your CI/CD pipeline, all without having to care about infrastructure. There is a strength to that and only a fool would dismiss those tradeoffs.

So sure, you pay the Google or Amazon tax when you use their services, but on the other hand they let me scale horizontally, on demand, without having to care about the underlying infrastructure while providing distributed systems for things like file and data storage that work seamlessly with the rest of their platform. There is value to that.
I would disagree here - the alternative to EC2/GCE is to:
1. buy servers, switches, firewalls,
2. go to the datacenter, go through the background check, rent rack space for $$$$,
3. install your switches, firewall, server, storge (DAS or SAN depending on arch),
4. pay windows cals and /or install a VM hypervisor, provision virtual machines,
5. pay a sysadmin, network tech, and general support IT guy, to set it all up and to keep those running and locked down -
6. pay for licensing for managed switches, firewall, security software, redundant managed PDUs, and
7. pay security consultant to do a pentest/security report (if you're in finance or healthcare this is essentially mandatory).
8. Do the same thing at another COLO and configure a hot failover for redundancy (if you can afford it)

or you can

1. Click a few buttons on a WI, set up hot az failover, elastic storage.
2. Set up your OS, SGs, Subnets
3. Download AWS ISO/HIPAA/FERPA/HITRUST certs and attest that you have configured your services properly.

Also - good luck if you're a startup - if you have to shutter the business in 6 months and you're locked into a 12-36 month datacenter hosting agreement.
Posted on Reply
#28
EsaT
R-T-BHonestly it's amazing this does not happen more often, to be quite frank.
There's Finnish saying which fits to this:
Ships sail (and float) with luck.
Posted on Reply
#29
Aquinus
Resident Wat-man
phanbueyI would disagree here - the alternative to EC2/GCE is to:
1. buy servers, switches, firewalls,
2. go to the datacenter, go through the background check, rent rack space for $$$$,
3. install your switches, firewall, server, storge (DAS or SAN depending on arch),
4. pay windows cals and /or install a VM hypervisor, provision virtual machines,
5. pay a sysadmin, network tech, and general support IT guy, to set it all up and to keep those running and locked down -
6. pay for licensing for managed switches, firewall, security software, redundant managed PDUs, and
7. pay security consultant to do a pentest/security report (if you're in finance or healthcare this is essentially mandatory).
8. Do the same thing at another COLO and configure a hot failover for redundancy (if you can afford it)
That's the part where I said you're on the hook for running, maintaining, securing, and scaling the infrastructure yourself if you decide to not use a cloud service provider. I think we're on the same page here.

Also, if you're just using the WI and not CloudFormation in AWS for a real deployment, you should be slapped for not using IAC.
Posted on Reply
#30
W1zzard
R-T-BDoesn't TPU use cloudflare for their DDOS protection? I understand you can turn that off easily, but yeah, not completely independent of the cloud as claimed.
We don't. We use Cloudflare as CDN for static content only, not pages, and that can be turned off in a few seconds and the change propagates within a few minutes
Posted on Reply
#31
Shihab
Onasifor the vast majority of services the cost to benefit analysis on that just doesn’t make sense.
Only because “cost” and “benefit” are only concerned with stuff that it [or someone could delude themselves into believing it] can be described as hard numbers. Very few things in life fit that criteria…

Remember when “cost/benefit analysis” meant “move all your manufacturing to China?”
Posted on Reply
#32
Bloste
TIL Autodesk uses AWS for their Licensing and Entitlement services. Everything (100+ AEC Collection licenses) stopped working at 9:30 CET. Just great.
Posted on Reply
#33
Aquinus
Resident Wat-man
BlosteTIL Autodesk uses AWS for their Licensing and Entitlement services. Everything (100+ AEC Collection licenses) stopped working at 9:30 CET. Just great.
We're learning about all of the companies that don't have an easy to execute DR plan.
Posted on Reply
#34
remixedcat
BlosteTIL Autodesk uses AWS for their Licensing and Entitlement services. Everything (100+ AEC Collection licenses) stopped working at 9:30 CET. Just great.
I think cisco meraki uses aws for some of the meraki dashboard shards. glad I switched to ubiquiti cuz all my network management is local and immune from cloud dependancy
W1zzardWe don't. We use Cloudflare as CDN for static content only, not pages, and that can be turned off in a few seconds and the change propagates within a few minutes
I use that for my sites too and dev mode or purge cache works pretty quick
Posted on Reply
#35
Princess Garnet
WirkoHow many nines does AWS advertise?
Since none of them is a Final Fantasy, the answer is not enough of them.

(Sorry.)

But yeah, I saw more people reporting on this than I noticed it, so I might not have noticed it otherwise. But it also seemed to largely happen overnight for my time zone.
Posted on Reply
#36
Unregistered
dgianstefaniOne of the downsides of this approach is that when AWS has a problem, literally half of the internet has issues.
"Half the Internet" is one of the most absurd exaggerations I've heard. The outage was only a portion of AWS (AWS East); AWS holds under a 35% market share; and many servers on the Internet are still privately maintained, and don't use cloud services whatsoever.
dgianstefaniThere's not much freedom to compete when a couple of dominant players have effectively cornered the market. Economies of scale and available experienced engineers basically force viable alternative solutions to be niche or expensive
People keep repeating this falsehood no matter how often history proves them wrong. Intel had effectively corned the processor market much more than AWS has cloud services. GM and Ford once held over 85% of the auto market, now today their share is 35% and new startup Tesla is worth several times what they are combined. General Electric once held a monopoly on nearly all home appliances made -- how are they doing today?

And these are manufacturing sectors where economies of scale play a far, far larger role than cloud services.
Posted on Edit | Reply
#37
R-T-B
W1zzardWe don't. We use Cloudflare as CDN for static content only, not pages, and that can be turned off in a few seconds and the change propagates within a few minutes
Fair enough. Just playing journalist again and asking the hard questions. Should've known you'd be ready lol.
Posted on Reply
#38
Unregistered
ChaitanyaMany large firms have learned their lessons from previous AWS outages and this one will furthur cement things having services and data on premise or self hosted premises is the way forward.
So you agree with me and the vast majority of economists that the invisible hand of the market is working well.
Posted on Edit | Reply
#39
Wirko
ShihabOnly because “cost” and “benefit” are only concerned with stuff that it [or someone could delude themselves into believing it] can be described as hard numbers. Very few things in life fit that criteria…

Remember when “cost/benefit analysis” meant “move all your manufacturing to China?”
The term "cost/benefit" was cleverly coined so as to exclude "risk". If you run a business in the rest-of-North-America, South America, Africa, Asia or Europe and need to make an uneasy decision whether to trust your (financial, health, HR etc) data to AMZN/GOOG/MSFT/ORCL, there certainly is more than $cost/$benefit on the table.
Posted on Reply
#40
TheDeeGee
My iRobot Roomba app was affected.

Could still start it manually, but then it will not ignore zones it's not allowed in.

Always online stuff like that seriously needs to be banned.
Posted on Reply
#41
ymdhis
TheLostSwedeThings that aren't connected to the internet?
Not true, the joystick on my C64 doesn't work right. The last I regularly used it was in the early 00s, I took it out for testing around 2010 and it barely worked. I suspect the mechanical contacts on the switches need cleaning, in so many years it may have accumulated some rust or something else. I'll have to rinse it with either contact cleaner or wd40, depending on how the switches will look like.

The computer still works though.
Posted on Reply
#42
TheinsanegamerN
TheDeeGeeMy iRobot Roomba app was affected.

Could still start it manually, but then it will not ignore zones it's not allowed in.

Always online stuff like that seriously needs to be banned.
Why did you buy an always online trinket if you dont like always online trinkets?
ymdhisNot true, the joystick on my C64 doesn't work right. The last I regularly used it was in the early 00s, I took it out for testing around 2010 and it barely worked. I suspect the mechanical contacts on the switches need cleaning, in so many years it may have accumulated some rust or something else. I'll have to rinse it with either contact cleaner or wd40, depending on how the switches will look like.

The computer still works though.
How much old stuff still works? My NES and N64 have both failed from old age, and the N64 remains unfixable at the moment. Likely a bad co processor that you cant replace unless you have a BGA soldering machine.

I've had tons of controllers fail over the years.

Then there's software. Windows 95 software didnt require an internet connection, how much of that just runs on a modern PC? I'd guess very little. Sure there's compatibility patches and mods and things like DOSBOX, but the OOTB experience with old software is pretty dire, just like modern internet connected software.
Posted on Reply
#43
FreedomEclipse
~Technological Technocrat~
This outage is actually very interesting. It tells us how many businesses are running their web services through AWS.

There are too many eggs in one basket.
Posted on Reply
#44
R-T-B
TheinsanegamerNMy NES
Mine still works. Amazing considering how much I abused the thing.
Posted on Reply
#45
TheDeeGee
TheinsanegamerNWhy did you buy an always online trinket if you dont like always online trinkets?


How much old stuff still works? My NES and N64 have both failed from old age, and the N64 remains unfixable at the moment. Likely a bad co processor that you cant replace unless you have a BGA soldering machine.

I've had tons of controllers fail over the years.

Then there's software. Windows 95 software didnt require an internet connection, how much of that just runs on a modern PC? I'd guess very little. Sure there's compatibility patches and mods and things like DOSBOX, but the OOTB experience with old software is pretty dire, just like modern internet connected software.
You come vacuum the house for me then.
Posted on Reply
#46
remixedcat
FreedomEclipseThis outage is actually very interesting. It tells us how many businesses are running their web services through AWS.

There are too many eggs in one basket.
And exposes their attack surface too. And reveals where they are hosted.
Posted on Reply
#47
FreedomEclipse
~Technological Technocrat~
TheDeeGeeYou come vacuum the house for me then.
Be careful what you wish for.

Posted on Reply
#48
mechtech
I wondered why late last night net was so slow? Maybe cause this?
Posted on Reply
#49
Bigshrimp
Too much concentration of infrastructure in one basket. When that one basket goes, so does all of what is contained in it. We need more baskets (service providers) not less. Call it a vulnerability if you will.
Posted on Reply
#50
Rob94hawk
Is that why all my ring cameras were down?!🤣
Posted on Reply
Add your own comment
Nov 17th, 2026 13:08 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts