www.it-connect.fr
Proton Suffers Global Outage After Frankfurt Datacenter Cooling Failure
On the night of August 26 to 27, 2026, Proton suffered a major outage affecting the Swiss company’s services: Proton Mail, Proton VPN, Proton Drive, Proton Pass, Proton Calendar, and more. Why? The cooling system at the Frankfurt datacenter failed. It took about 2 hours and 20 minutes to bring everything back online. Here’s what we know.
A 2h20 outage at Proton
Proton runs its own servers, notably in Switzerland, Germany, and Norway. This independence is essential for Proton, especially to maintain full control over its infrastructure and its customers’ data. Rare event: one of its datacenters experienced a major outage.
According to Proton’s status page, here is what happened (events shown in Paris time):
- 00:09: Proton reports that an investigation is underway regarding the availability of its services.
- 00:18: Proton VPN moved from major outage to partial outage, with the service still working for some users.
- 00:38: The root cause of the incident is identified. It is a failure of the cooling system at its Frankfurt datacenter. The datacenter is isolated and users are routed to other data centers.
- 01:37: the first services begin responding again for some affected users.
- 01:57: a fix is applied, and monitoring begins.
- 02:27: the outage is declared resolved.
Personally, I don’t do video calls in the middle of the night and I don’t read my emails either (although that can happen), so I didn’t notice this outage. But for those using Proton services and who were hit during the day, it may have been disruptive.
Indeed, according to the incident page, the outage affected Proton Mail (web interface, incoming and outgoing mail, etc.), Proton VPN, Proton Calendar, Proton Drive and Proton Docs, Proton Pass, as well as SimpleLogin, Proton Wallet and Lumo. Although this morning around 09:00, I used Proton VPN and it was not working properly, there may be a link with this outage.
Servers in a room at 50-60 °C: an extreme overheating event
In a thread posted on X, Andy Yen, Proton’s CEO and cofounder, described a total and unprecedented failure of this datacenter’s cooling system. According to him, the servers went down as they overheated, which prevented the failover systems from determining whether the site was actually unavailable.
This behavior forced the technical teams to perform a manual failover. All of this in a room where the air temperature reportedly quickly reached 50 to 60 °C. The cooling systems were redundant, and yet they all failed at the same time: unlikely, and yet it happened.
"We therefore had to choose between restoring cooling and saving the hardware that is extremely expensive by powering everything down.", we can read.
What matters to users is that no data was lost. There was still an impact on email with delays, but mail servers normally queue undelivered messages (SMTP retry mechanism SMTP). Finally, note that Proton promises a post-mortem on its status page once this incident is fully over.
