The AWS outage is proof someone is breaking the internet

A three-hour global outage at Amazon Web Services was enough to halt banks, games and government services – exposing the fragile foundations on which our digital world rests, says Chris Stokel-Walker

It’s been Groundhog Day on the internet as apps, services and websites mysteriously went offline. From high street banks and HMRC to Roblox’s online games, few elements of our lives were untouched by a massive outage that affected Amazon Web Services (AWS).

The e-commerce giant Amazon happens to control huge amounts of internet cloud infrastructure – the nuts and bolts that keep all the apps, websites and more that we use day to day alive. We tend only to take notice of how the world wide web works when it doesn’t.

Which is more often than you might think. Similar AWS outages caused chaos in 2023, 2021 and 2020. Internet infrastructure outages are often more reliable than many buses.

To its credit, Amazon worked quickly to solve the problem. Within just over three hours of people’s web browsers throwing up errors, Amazon had “fully mitigated” the issue, which seems to be rooted in how the company served up the internet’s phone book, the domain name system, or DNS. Kids can return to Roblox. Adults can log back onto Zoom meetings. Businesses can make payroll on accounting software Xero – all three of which were impacted.

Yet the outage, and the speed at which everything ground to a halt, has highlighted a fatal flaw in how we have set up the web.

It was initially conceived as a self-sustaining network, where its users would help contribute collectively to its continued existence. They were often based in university laboratories and non-profit organisations. Since the late 1980s, things have changed significantly. Big Tech has entered the picture, and the web – and our dependency on it – has grown.

So much so that these days, we rely on a handful of companies to keep us online. AWS controls around one-third of the cloud infrastructure market. Microsoft holds another 20 per cent or so, through its Azure systems, while Google Cloud controls around an eighth of all our digital connectivity.

All do admirable jobs in keeping us online day in, day out. The fact that the internet falls over so catastrophically so rarely is vindication of that – and of their much-vaunted claims to offer “six nines” or better uptime (no, that’s not a teen in-joke like the 6-7 meme: it’s the decimal points at which infrastructure is measured – working without a hitch 99.999999 per cent of the time).

But it’s what happens the 0.000001 per cent of the time that becomes the issue. That concentration of so much of our digital lives in the hands of so few companies ought to give us pause, because it means that when those small, rare hiccups occur, they balloon into huge ones.

The fact that these outages have hit government services, including the UK taxman, might make the government think twice about whether too much power is in too few hands here. This outage was almost certainly caused by a cock-up, not a conspiracy. It seems likely that a small configuration error or update will have compounded into a big problem. But it’s worth thinking about the way in which a US-based outage can impact the entire web.

The European Union has started talking – long before this outage, and in part because of the America First policies of Donald Trump – about developing a Europe-specific tech stack, so they aren’t in thrall to the whims and issues that ensue from an unpredictable president. Given that small errors can happen even without the whims of a vindictive world leader telling companies what to do, it seems sensible to try and think more deeply about how much we want to route through the United States.