When a smart home company shuts down

The worked example: Belkin Wemo, 31 January 2026#

Belkin ended Wemo service on 31 January 2026. The app and cloud features went away, and with them the Alexa and Google integrations, which were cloud-to-cloud links with nothing left on the other end.

One group of devices kept working. Belkin's own notice states that Wemo products already configured with Apple HomeKit before that date continue to function through the Apple Home app without Wemo cloud services, and that after the date you can no longer set up a Wemo device with HomeKit at all.

That split is the whole lesson of this page, so be precise about the mechanism. When a HomeKit accessory is paired, controller and accessory exchange long-term keys and store them on the accessory itself. From then on, the Home app or a home hub opens an authenticated session to the accessory over your own LAN. Belkin's servers were never a hop in that path, so switching them off removed the setup route, the app and the cloud-to-cloud integrations, but could not touch a session already established between two devices in your house.

Pairing, though, is the moment a vendor gets to say no, which is why new HomeKit setups stopped and existing ones did not. It also explains why the Alexa and Google links died instantly: those are rung five paths in local vs cloud control, where the assistant's cloud calls the vendor's cloud. Remove one cloud and the chain breaks at both ends.

The rest of the 2025 and 2026 record#

Wemo is the cleanest example, not the only one. Five other cases from the same period show the different shapes a shutdown takes.

Two of these deserve comment because they are not the story people expect.

The Nest thermostat case is the good version of a bad outcome. The device kept its primary function, because the control loop always ran on the thermostat rather than on a server. What was lost was remote access, app scheduling and integration. That is a direct argument for buying thermostats that hold their schedule locally.

The Apple architecture migration was not a shutdown at all, and it still cost people their smart home. Support for the previous version of Apple Home ended on 10 February 2026, and anyone who ignored the upgrade prompts for years lost automations and shared access. A platform does not have to fail to break your house. It only has to move.

The MyQ case is different again: nothing failed and nothing was retired. Chamberlain decided third-party access was not welcome, and garage doors that had been part of people's automations stopped being part of them. There is no engineering defense except never having depended on it. See blinds, shades and garage doors for alternatives, most of which drive the opener's dry contacts rather than its cloud.

Not every corporate event is a threat. LG acquired an 80 percent stake in Athom, the maker of Homey Pro, and no product discontinuations followed. Acquisition is a signal to watch, not a verdict.

The warning signs#

Shutdowns are announced with a few weeks of notice. The decision behind them is visible for a year or more. Watch for these, in roughly increasing order of seriousness.

1. The app is rewritten and the old one is deprecated. Migration is when a vendor decides which hardware generations are worth carrying forward. Anything not carried forward is on a list.
2. A subscription appears, or the free tier is restructured. Google replaced Nest Aware with Google Home Premium on 1 to 2 October 2025. Repricing is normal. Repricing plus feature removal is a company trying to make a line pay for itself.
3. Developer APIs are deprecated or start charging. Samsung is ending free SmartThings API access in October 2026, with a $4.99 a month personal plan for individual developers. Home Assistant's SmartThings integration falls under that plan and, as of September 2026, Home Assistant has not said how it will handle it. Apple, Google and Amazon still offer free smart home API access.
4. Firmware updates stop. Check the release notes page. Eighteen months of silence on a connected product is not maturity, it is a maintenance budget that went to zero.
5. Support articles get retitled "legacy" and hardware quietly leaves the store with no successor.
6. The integration you rely on breaks and nobody fixes it. A cloud-to-cloud link that stays broken for months is a link nobody is paid to maintain.

How to buy against it#

The goal is not to predict which company fails. It is to make the answer not matter much. Six rules, in order of how much they buy you.

1. Prefer devices that pair to something you own. A Zigbee or Z-Wave device joins your coordinator. The manufacturer has no session to revoke and no server in the path. This is why the enthusiast tier still buys these radios in 2026 despite Matter existing: the trust model has no company in it.

2. Where you buy Matter, use multi-admin. A Matter device can hold credentials for several fabrics at once. Commissioning the same device into a second controller, ideally a local one such as Home Assistant or Hubitat, means that losing one platform does not lose the device. It is ten minutes per device and it is the single best insurance available to a mainstream buyer.

3. Check what the device does if you never open the vendor app again. Not what it does during an outage: what it does permanently. A lock that stores codes on the lock, a thermostat that holds its schedule, a bulb that responds to a bound remote. If the honest answer is "nothing", you are renting it.

4. Prefer local pairing over account-based pairing. If setup requires creating an account with the manufacturer, the manufacturer owns the on-ramp. Matter's commissioning flow is designed to avoid that, though plenty of vendors still push you through their own app first.

5. Keep the exit cost low deliberately. Standard radios, standard mounting, devices that another hub can adopt. The exit cost calculator estimates what leaving would cost in hardware and hours, which is a more useful number than any brand reputation. Switching platforms without rebuying is the procedure if you have to.

6. Weight the things that are hard to replace. A $20 plug going dark is annoying. In-wall switches, a lock on the front door and anything wired into HVAC are expensive to change, so spend the extra scrutiny there and relax about the plugs.

What Matter changes, and what it does not#

Matter genuinely helps here. Control is local by design, credentials sit in fabrics on your own network, and multi-admin means no single controller vendor owns the device. If a Matter device maker disappears tomorrow, a device already commissioned into your fabric keeps taking commands, because the CSA's design put the control path on your LAN.

What Matter does not do is protect the things outside the control path. Firmware updates come from the manufacturer. Vendor-specific features that Matter never modeled, such as vendor lighting effects, live in the vendor's app and go when it goes. Cloud features attached to the same product go too. And a device that has not been commissioned yet may become uncommissionable, exactly as the Wemo case showed. What Matter still cannot do is the full list.