What’s the point of a “smart” bed if it won’t withstand a rough night in the cloud? Last October 20, 2025, a huge AWS blackout highlighted a glaring weakness in Eight Sleep’s marquee Pod3 system: with no internet connection, its AI-driven temperature control, sleep monitoring, and automation fail completely. For thousands of owners, the breakdown was more than a gentle annoyance it was a literal uncomfortable night, with beds stuck at scorching presets, at rigid inclines, or with blinking lights in the dark.
The blackout was a result of a fault at AWS’s US‑EAST‑1 region, which froze access to DynamoDB endpoints with a fault in DNS resolution. This snowballing point failure spread out to dependent services, affecting over 3,500 companies in 60+ countries and generating over 16 million reports of outages. For Eight Sleep, whose architecture directs all control signals as well as biometric processing of data through AWS, connectivity loss simply made the Hub, or control for its water‑cooled system of coils, cease to work. Even the embedded touch interface, which was intended as a standby, was inconsistent across most users.
This was a classic concentration risk of IoT design. As cloud architecture experts warn, tightly coupled device behavior with a particular cloud region increases a particular outage’s “blast radius.” Because the Pod3 couldn’t work offline, owners were powerless, unable to override cooling/heating, change bed orientation, or record sleep metrics until AWS came back online. One social media posting by tech buff Alex Browne epitomized the ridiculousness perfectly: his bed, set to warm +9°F over room, stuck in “sauna mode” for the evening.
Eight Sleep’s cloud dependency is typical. Almost every home device that’s internet capable thermostats, lighting fixtures, etc. uses cloud-first control paths for convenient local functionality as well. This architecture opens itself to AI-driven personalization based on individual behavior as well as cloud-based, automated local updating of software, but it creates a tenuous dependency chain. As Alan Woodward, professor of computing at the University of Surrey, explained to us, “The more complex you make the functionality, the more difficult it is to understand what will happen when some parts of that whole chain of computing fails.”
In an engineer’s view, the Pod3 combines ballistocardiography-driven sleep tracking with body-temperature control via circulating water channels. Temperatures range dynamically between 55°F and 110°F, with biometric input optimizing for stage of sleep. But with all decision logic resident on the cloud, a local request for a change in temperature had had to go via the internet to AWS servers and home once more. When that circuit was interrupted by a DNS issue, the device had no coded fallback behavior beyond maintaining its previous state.
Resiliency for designing for IoT needs more than redundant servers. Best practices are installing local control loops for key functions, multi-region failover, and “graceful degradation” modes of operation in which non-essential features shut off but key ones such as temperature control remain operating. For smart beddings, that might involve moving a thin controler that would execute established thermal schedules, override on demand over Bluetooth or local-web UI.
The blackout also highlights the budding need for locally controlled smart home systems. Home Assistant and Hubitat, for example, run requests locally on the home network, so simple automation is retained in the event of an internet outage. For a device that will cost between $2,000–$5,000, it’s not a nice-to-have but a hedge both against cloud outages as well as factory shut downs.
Temperature control for smart beddings is a precision engineering endeavor. It must maintain stable microclimates with continuously monitored skin temperature, heart rate variation, as well as room conditions, with algorithmic regulation of water circulation, as well as heater/chiller outputs, in near real time. For existing device architecture of Pod3, such algorithms are hosted remotely. It would reduce latency, increase reliability, and keep pace with future IoT resilience standards by bringing at least a part of that logic to the device’s firmware.
Eight Sleep CEO Matteo Franceschetti has also committed to making Pod experience “outage-proof,” which would translate to a future with offline access. Technologically, such a course may involve firmware upgrades to enable local processing of biometric data as well as local device control in situations where cloud access is lost. Such a rethinking would not only protect users from future AWS outages but would also strengthen larger issues about control, ownership of internet-enabled devices, as well as about privacy.
For internet of things professionals, thePod3 outage provides a classic case study in cloud-enabled intelligence vs. local control. For consumers, it’s a warning that in a smart home, luxury turns to liability overnight if there’s no resilience with that convenience.
