Twenty years of smart home standards, and what each wave actually delivered

Wave one: X10 and the powerline era#

X10 was created in 1975 by Pico Electronics in Glenrothes, Scotland, and reached American homes through RadioShack and Sears from 1978. It signaled by putting a 120 kHz burst on the mains wiring at the zero crossings of the AC waveform, one bit per crossing, which after retransmission and line control worked out to roughly 20 bits per second.

What it promised: any module from any brand, from any controller, over wiring you already had. That promise was real, and modules from different manufacturers worked together for decades.

What it delivered: on and off, slowly, most of the time. Twenty bits per second is enough for an address and a command and nothing else, and the base protocol had no acknowledgment, so a controller could not tell you whether a command landed.

What limited it: physics. Signals attenuate crossing between the two legs of a US split-phase panel, so half your house may not hear the other half. Appliances inject noise onto the same wiring, neighbors on the same distribution transformer can trigger your modules, and only one command can be in flight at a time.

What it predicts: openness is necessary and not sufficient. If the physical layer is unreliable, people conclude that smart homes do not work, and that reputation outlives the technology.

Wave two: dedicated low-power radios#

Zensys, a Danish company, began developing Z-Wave in 1999 as a light control system and grew it into a home automation protocol. The Z-Wave Alliance formed in 2005 around early adopters including Danfoss, Ingersoll-Rand and Leviton. The Zigbee Alliance published Zigbee Specification 1.0 on 13 June 2005, built on the IEEE 802.15.4-2003 low-rate radio standard.

What they promised: reliable, battery-friendly mesh networking with acknowledgments and retries, plus certification so a certified device from one vendor would work with any certified controller.

What they delivered: essentially that. Z-Wave's US band at 908.42 MHz sits away from Wi-Fi, still its biggest practical advantage, at the cost of 232 nodes and four hops in the classic mesh. Zigbee got cheaper silicon, a shared 2.4 GHz radio, and eventually the largest installed base of any smart home mesh. Both are still supported and still selling in 2026.

What limited them: fragmentation and gatekeeping. Early Zigbee splintered into incompatible application profiles, which is why Zigbee 3.0 exists and why so many pre-3.0 devices were painful. Z-Wave ran on a single silicon supplier for most of its life, through Zensys, then Sigma Designs, then Silicon Labs, which bought the Z-Wave business in April 2018. That structural risk only closed in June 2025, when Silicon Labs and Trident IoT both certified under the new Z-Wave protocol certification. Both protocols also require a hub, and neither speaks IP, so every integration is a translation someone has to maintain.

What it predicts: protocols with independent standards bodies and multiple implementers keep working long after the interesting companies leave. Twenty years on, a Zigbee bulb from 2016 still joins a coordinator you bought last month.

Wave three: Wi-Fi, apps and the cloud#

From roughly 2012 onward the cheapest way to ship a connected product stopped being a radio module and a hub partnership. It became a Wi-Fi chip, a phone app and a rented server. Belkin's Wemo line, Insteon (launched in 2005 by Smartlabs and later cloud-connected), Wink (spun out of Quirky and released on 23 June 2014) and hundreds of white-label Tuya-based brands all took this route.

What it promised: no hub, no compatibility matrix. Buy the plug, scan the sticker, done, with remote access and voice for free because the device was already online.

What it delivered: exactly that, which is why Wi-Fi plugs still outsell everything else. It also delivered a per-device dependency on a server the buyer does not own, and a 2.4 GHz airtime problem that worsens with every device you add. Wi-Fi smart devices covers the second cost.

What limited it: unit economics. A $25 plug generates no recurring revenue but incurs recurring cost, forever. There are only three exits from that trap and all three have been taken. Charge for it: Wink announced on 6 May 2020 that continued use of its app, hub and devices would require a monthly fee, initially giving customers a week's notice, later pushed to July. Sell the company: Wink passed from Quirky's September 2015 bankruptcy to Flex for $15 million, then to i.am+ for $59 million in July 2017. Or stop: Insteon shut down abruptly in mid-April 2022 with no warning, after a sale process failed the month before. A group of its own users bought the company in June 2022 and restored service that October, the only genuine revival here and pure luck.

What it predicts: any product whose core function needs a vendor server has a life expectancy set by that vendor's balance sheet, not by its hardware. This is the single most reliable pattern in the whole history.

Wave four: the assistant wars#

Apple introduced HomeKit with iOS 8 on 17 September 2014 and did not ship the Home app until iOS 10 on 13 September 2016. Amazon announced Alexa and the Echo on 6 November 2014, with general US availability on 28 June 2015. Google announced Google Home in May 2016 and shipped the speaker on 4 November 2016.

What it promised: the assistant as the universal integration layer. You would not care what radio a device used, because you would talk to it, and the platform would handle the rest.

What it delivered: enormous reach and a useful abstraction, plus three parallel certification programs and three different feature ceilings for every manufacturer. It also relocated the integration from a protocol to a cloud-to-cloud API, which meant the platform owner could change or revoke it.

What limited it: the incentive to make the assistant the product rather than the plumbing. Alexa and Google Home routines still run in a data center. Amazon removed on-device voice processing on 28 March 2025, deleting the option not to send recordings to the cloud. Google's global outage on 18 August 2026 lasted about twelve hours and left Nest hubs unresponsive. The Local Control Index scores where each platform ended up.

What it predicts: platform-mediated compatibility is compatibility on loan. Which brings us to the clearest case study of the whole period.

Works with Nest, and what it taught everyone#

Google completed its acquisition of Nest Labs on 14 January 2014 for $3.2 billion. Nest had bought the hub startup Revolv, and on 15 May 2016 Revolv hubs were remotely disabled after a February notice. The hardware was fine. The service was switched off.

Works with Nest was the developer program that let third-party devices react to Nest presence and thermostat state. Google announced its discontinuation on 7 May 2019 and shut it down on 31 August 2019, pushing partners toward a narrower Google-controlled program. Working automations stopped on a date chosen by Google. The pattern continued: Dropcam and Nest Secure servers went dark on 8 April 2024, and Nest Learning Thermostat first and second generation units were cut off on 25 October 2025, left usable only as ordinary thermostats.

The lesson the industry took was not "do not depend on a platform". It was "we need a standard that nobody can switch off". That argument produced Matter, and the companies making it are the same ones that switched things off.

Wave five: CHIP and Matter#

In December 2019, Amazon, Apple, Google, Samsung SmartThings and the Zigbee Alliance announced Project Connected Home over IP. The alliance renamed itself the Connectivity Standards Alliance in 2021, the project became Matter, and version 1.0 was published on 4 October 2022.

What it promised: one IP-based application layer over Wi-Fi, Ethernet and Thread, with local control, multi-admin sharing across ecosystems and a single certification. No more three programs per product.

What it delivered: a well-designed data model and a working security architecture, plus the gap this site keeps documenting. The specification defines 92 device types across releases up to 1.6, while reporting from the alliance's own June 2026 event described Apple, Google and Amazon as barely at 1.3. Google shipped the Matter 1.2 appliance device types in September 2025, roughly 23 months after that specification. Joint Fabric became certifiable in 1.6 and nobody has shipped it, exactly as nobody shipped its predecessor Fabric Sync from 1.4. What Matter still cannot do is the current version of that list.

What limits it: implementation is voluntary and unenforced. Certification tests conformance, not behavior, and three shortcut paths exist under which only the original product passes lab testing. Matter also does not guarantee local control: the alliance's own FAQ notes that Matter-only devices need an internet-connected controller when you are away from home.

What it predicts: Matter fixes the certification tax on manufacturers, and does not fix the platform's incentive to differentiate. Expect the specification to keep running ahead of the apps.

Thread rose on Matter's back#

Thread's first specification arrived in 2015 and for seven years it was a technology with almost no consumer products. Thread 1.3 in October 2022 became the Matter 1.0 baseline, and within three years Thread radios were in Apple TVs, HomePods, Nest hubs, Echo speakers, eero routers and SmartThings hubs.

The bad news is that each ecosystem formed its own Thread network, and the credentials that define a network are not shared automatically. Thread 1.4, released on 4 September 2024, added credential sharing, but it is user triggered, single use and only mandatory for newly certified 1.4 border routers. Deployed hardware moved slowly: Apple TV and the Google TV Streamer reached Thread 1.4 around 10 June 2026, while HomePods and Nest Hubs still had not by September 2026. Why you have three Thread networks is the consequence.

Where things stand in 2026#

Matter shipped 1.5 on 20 November 2025 with cameras and closures, 1.5.1 on 31 March 2026 and 1.6 on 17 June 2026. SmartThings implemented 1.5 within about a month and remains the only ecosystem where a Matter camera works end to end. Zigbee 4.0 was announced on 18 November 2025, fully backward compatible with 3.0, which the alliance frames as protecting more than a billion deployed devices. Suzi, its sub-GHz long-range mesh on the Zigbee network layer, opened certification on 2 September 2026 with no shipping consumer products yet. Z-Wave Long Range reached 125 certified devices by CES 2026, with around 80% of the certification pipeline targeting it.

The two twenty-year-old protocols are both shipping major backward-compatible upgrades, and the new one is shipping features nobody has implemented. That is not an argument against Matter. It is an argument about which layer to bet a house on, and which radio for which job is the buying version.

Dead or degraded, with dates#

Every row is a cloud service or a platform program. None is a radio protocol. When a smart home company shuts down covers the warning signs and the buying rules that follow.

The lesson, stated plainly#

Open, local protocols outlive the companies that shipped them. Cloud-dependent products die on the vendor's schedule.

Zensys is gone, Sigma Designs is gone, and Z-Wave is not. The Zigbee Alliance changed its name and a Zigbee bulb from 2016 still pairs. Quirky went bankrupt, Wink changed hands twice and then started charging rent. Revolv, Dropcam, Nest Secure and Wemo were switched off by companies that are still perfectly healthy, which is the point: these products did not die because their makers failed. They died because keeping them alive was optional.

The rule is not "avoid the cloud". It is: know which of your devices stop being useful when a server goes away, and keep that set small and cheap. A Zigbee or Z-Wave sensor talking to a hub you own is a twenty-year asset. A Wi-Fi camera with no local recording is a subscription with a lens on it. Local vs cloud control and the outage audit tool sort yours into the two piles.