Amazon Web Services, the cloud provider behind much of modern life, suffered a massive failure in its US‑East server region in October. The chain reaction crashed banking apps, grounded parts of airline systems, froze food‑delivery platforms and turned smart home gadgets into lumps of plastic.
Customers of the smart bed company Eight Sleep had to go to bed that night sleeping at strange angles, because their US$3000, internet-enabled beds had gone offline. The company quickly rolled out an “outage mode”, connecting beds via Bluetooth instead of the cloud. But the very existence of such a feature says everything about the era we live in – where even a decent kip can depend on the stability of a far‑off data centre in northern Virginia.
AWS explained that a routine software update to an internal networking subsystem – the digital plumbing that keeps cloud servers talking to each other – had triggered an avalanche of failures. The outage came three years after the pandemic accelerated cloud adoption, and one year after the CrowdStrike meltdown of 2024, which crashed Windows machines across the planet, airlines and hospitals included, thanks to a single faulty security update.
Both events revealed the same brittle truth – modern computing isn’t distributed, it’s concentrated. We may speak of “the cloud” as something soft and amorphous, but in reality, much of the internet lives in a handful of data centres run by a few tech giants. When one misconfigures a switch or ships a bad line of code, the consequences ripple across finance, logistics, media – and now, evidently, bedtime.
The problem is set to get worse as artificial intelligence takes over more of the computing load. AI depends on continuous cloud access to train, reason and act. As so-called AI agents start handling scheduling, trading, creative work and infrastructure management, downtime becomes more than an inconvenience. A few minutes without connectivity could stall fleets of delivery drones or paralyse automated financial systems. The smarter and more autonomous our services become, the dumber the world looks when they go dark.
I was in San Francisco last week at Dreamforce, a major software conference where AI agents were front and centre. Soon, your own personal AI agent will talk to agents run by the airlines, banks, your Google calendar and the local supermarket, working in the background to take care of your life admin.
That’s great, but cloud resilience quietly grows as an existential issue. Businesses have long accepted the trade‑off of convenience for control. It’s cheaper to rent computing power from Amazon or Microsoft than to build redundant data centres. But the AWS fiasco highlighted how that concentration of power is itself a systemic risk, the digital equivalent of putting every egg in one gigantic basket that turns off if Virginia sneezes.
What’s the answer? Geeks talk about multi‑cloud strategies, edge computing and local processing. They’re all sensible ideas in theory. We need to build sovereign clouds to reduce our reliance on Big Tech to run critical digital infrastructure.
Yet in practice, the gravitational pull of the big three – AWS, Microsoft Azure, and Google Cloud – only increases. The AI boom is accelerating that dependence. Every chatbot, image generator and recommendation model sits atop the same stacks of rented computers.
The real answer lies in remembering what made the original internet resilient – decentralisation. Cloud computing has mirrored the corporate logic of consolidation: build once, scale infinitely, trust the uptime dashboard. But real resilience comes from diversity – more providers, more local capability.
Otherwise, the next time the cloud goes down, the lights will flicker, the chatbots will stop responding, and somewhere, a thousand smart beds will once again tilt humankind awake to its own technological overconfidence.
