Hajra Naz
Something went wrong at Meta on September 21, and a lot of people in the US noticed. Facebook and Instagram both started acting up. By 9:11 PM Eastern, Downdetector — the site that collects outage reports straight from users — had logged over 17,000 complaints tied to Facebook alone. Instagram wasn’t spared either; more than 5,000 people flagged problems there too, all within that same evening stretch.
People couldn’t load their feeds. Uploads failed. Messages wouldn’t send. Basic stuff just stopped working for a while. That said, nobody really knows how bad it actually was. Downdetector only counts what users choose to report — it’s not pulling numbers from Meta directly. So the real total could be way bigger than what showed up on the tracker, or honestly, smaller too. There’s no way to know for sure.
Reuters reached out to Meta. So did other outlets. Nothing back. That’s pretty typical, actually — Meta tends to stay quiet until things are fixed and their engineers have had time to figure out what broke. In the meantime, users were left in the dark. No timeline, no cause, nothing to go on except waiting it out.
Read More: Facebook officially launches stand-alone Creator Studio app with AI tools
For a lot of people, this isn’t just an inconvenience. Facebook and Instagram are how they talk to family, run a small shop, handle customer questions, or just stay in touch. Small business owners selling through these apps felt it in their wallets during the outage. Creators who count on the platforms to reach their audience lost engagement windows they can’t really get back.
This isn’t the first time, either. Meta pours a ton of money into infrastructure and redundancy, but things still break sometimes. Back in 2026, there were other outages too, each with a different footprint — sometimes it was a data center issue, sometimes a bad software push, sometimes something tied to specific server regions.
Every time it happens, there’s an internal scramble to figure out the failure and patch whatever let it happen. And yet, despite running one of the largest server networks on the planet built specifically to avoid this exact scenario, it still slips through occasionally.
