Amazon Web Services’ US-EAST-1 region went down for just over three hours on Sunday, the same Northern Virginia region whose failure knocked Fortnite and Roblox offline for most of a day last October. This time, players of the biggest online games barely noticed. Epic Games’ status page logged no incident for Fortnite or any of its online services on Sunday, and AWS listed no games among the 11 services it said were affected. The damage instead fell on sign-ins to AWS’s own tools, account sign-ups and a small group of network load balancers.
Problems began at 3:05 AM PST, and AWS said it fixed the authentication failures at 5:20 AM, more than two hours later. A smaller group of customers kept losing connections until 6:21 AM, when AWS fully mitigated the underlying network problem, according to its Health Dashboard. Some posts summarising the outage put its length at about two hours, but that figure only covers the sign-in fix. AWS has not yet said what caused the failure.
Sunday’s failure hit work logins, not player logins
The clearest impact was on IAM Identity Center, the AWS service that companies use to sign their own staff into the AWS Console and other tools. Customers relying on it could not authenticate, which also cut them off from services that log in through it, including Amazon Connect, the contact-centre platform, and Kiro, AWS’s AI coding tool. AWS said at one point that there were “no workarounds for impacted customers”, and new customers trying to create AWS accounts were blocked as well.
That difference explains why players were largely spared on Sunday. When you log in to Fortnite or another online game, you sign in through the publisher’s own account system, not through the workforce login that failed. A studio’s engineers may well have been locked out of their AWS dashboards for a couple of hours, but the game servers they run kept serving matches unless they depended on the affected network components.
Those network components were the longer tail of the outage. A subset of Network Load Balancers in a single cell of one Availability Zone, known as use1-az1, kept dropping connections after the sign-in problem was fixed. AWS told affected customers that the quickest workaround was to create a new load balancer with targets spread across several Availability Zones. The full list of affected services also included NAT Gateway, CloudWatch, Step Functions, OpenSearch Service and VPC Lattice.
TalkEsport checked the places where a game outage would have shown up first. Epic’s incident history lists nothing for any of its games or services on 15 November. Searches on X for Fortnite, Valorant, Roblox, PUBG or Twitch going down during the outage window also turned up no player reports.
Fortnite and Valorant run on AWS, and October 2025 showed what a bad day looks like
Gaming’s dependence on AWS is not a secret. Epic Games runs Fortnite’s online services on AWS, according to Amazon’s own customer pages, and Riot Games has moved its infrastructure to AWS after years on its own data centres. Twitch, the biggest streaming platform for games, is owned by Amazon.
That dependence showed on 20 October 2025, when a DNS problem with DynamoDB in US-EAST-1 set off a failure that lasted around 15 hours, according to AWS’s own post-event summary. Fortnite and Roblox went down alongside Snapchat, Alexa and several banking apps, because so many companies run core systems in Northern Virginia. It is AWS’s oldest and largest region, and many companies use it by default.
Gamers have had a rough few months with infrastructure beyond Amazon too. On Friday, a Microsoft outage kept Outlook and Teams down for more than 16 hours, although Xbox held up. Players in the Middle East have dealt with server problems since missile strikes in March hit an AWS facility in the UAE, and AWS’s dashboard still lists that region as disrupted. A PHP flaw also knocked gaming and esports websites offline in October.
Sunday’s failure began before dawn in the US, but it was late morning in Britain and late afternoon in India, so plenty of players were online when it hit. For Indian players, it ran from 4:35 PM to 7:51 PM IST. Epic’s own last incident came a day earlier, when Easy Anti-Cheat was degraded on 14 November for what the company called “a small window of time”.
