Skip to content

Killed by Google: The Product Graveyard

Abstract

Google has launched and closed hundreds of consumer products, and the closures became a subject of their own. The volunteer-run site Killed by Google, started in September 2018 after Google announced the end of Inbox by Gmail, listed 307 discontinued Google apps, services and devices in September 2026, with a median life of about four years. The pattern began with Larry Page’s “spring cleans” of 2011 and reached its most-remembered moment with Google Reader in 2013. It has been explained, by Google employees among others, as a result of a promotion system that rewards launching over maintaining. Its cost fell on users in three ways: habits and data moved out of closed services, devices such as the Revolv home hub disabled by a server shutdown, and links and APIs that stopped working. By the time Google launched Stadia in 2019, customers and developers were pricing the graveyard into their decisions, and Google’s cloud business published written stability promises in 2021 to answer the same doubt from paying customers.

Google Stadia Controller
The Stadia controller. When Stadia closed in January 2023, Google refunded the hardware and released a tool to switch the controller to Bluetooth so it would not become useless. Image: ubahnverleih, public domain (CC0), via Wikimedia Commons.

Spring Cleaning

For its first decade Google let engineers start products freely and seldom stopped them. Larry Page, who returned as chief executive in April 2011, set out to put “more wood behind fewer arrows.” Google Labs, the showcase for experiments, closed in July, and the first round of what Google called spring cleaning was announced on 2 September 2011 by Alan Eustace, a senior vice-president, as “a fall spring-clean”: Aardvark, Google Desktop, Fast Flip, the Maps API for Flash, Google Pack, Web Security, Image Labeler, Notebook, Sidewiki and Subscribed Links. “Technology improves, people’s needs change, some bets pay off and others don’t,” the post began. More rounds followed through 2012 and 2013. Google Buzz, the social network that had exposed Gmail users’ contacts at launch, closed in December 2011; Google Health, a personal health-record service, on 1 January 2012; Google Wave in April 2012.

These were mostly small, and the users affected were few. The one that was not came on 13 March 2013, when Google announced that Google Reader, its RSS reader, would close on 1 July. Reader had been the working tool of a large share of journalists, bloggers and developers for eight years; a petition collected more than 100,000 signatures within days, and the rival Feedly took half a million users in 48 hours (see Dead End: Google Reader). Google gave declining use as its reason. Reader’s users took away a different lesson: a free Google service could hold years of someone’s reading habits and still be closed with three and a half months’ notice.

The Graveyard

On 16 September 2018, days after Google said it would close Inbox by Gmail, a front-end developer named Cody Ogden created a public repository for a site he called Killed by Google, “part guillotine, part graveyard for Google’s doomed apps, services, and hardware.” Each entry has a small headstone, an opening and a closing date, and an epitaph. The data is an open file on GitHub that anyone can correct, and other sites and news reports began to cite its running count.

In September 2026 the file held 307 entries: 215 services, 68 apps and 24 pieces of hardware, a few of them with closure dates announced but not yet reached. Measured from launch to closure, the median product lived about four years and one month. The busiest year was 2019, with 37 closures, including Google+, Inbox, Allo and the URL shortener; 2011 (26) and 2021 (32) were the other peaks.

Messaging became the standard example. Google Talk (2005), Wave (2009), Buzz (2010), Hangouts (2013), Allo and Duo (2016), Google Chat and Google Messages overlapped, replaced one another and were merged or closed on schedules users could not predict. Hangouts, the longest-lived, ran from May 2013 to November 2022; Allo, a chat app with Google’s assistant built in, lasted from September 2016 to March 2019.

Launch, Promo, Abandon

Google has rarely explained its closures beyond usage and focus. Employees and former employees offered a structural explanation. In October 2022, after Stadia’s closure, a current Google employee wrote on Twitter that “you launch a service or a major overhaul and you put in your promo package. No one ever get promoted for ‘maintaining’ or ‘fixing something broken’,” and a former employee described a cycle known internally as LPA: launch, promo, abandon. Under that reading, a product’s best engineers leave after the launch they needed for promotion, the product stagnates, its usage falls short of what a company of Google’s size can justify, and it is closed. The claim comes from individual accounts rather than from Google’s documents, but it matches the graveyard’s shape: many products launched to fill the same need, few maintained for long.

Bricks

A closed web service loses its data; a closed service behind a device can take the device with it. In 2014 Google’s Nest division bought Revolv, a start-up whose $300 home-automation hub had been sold with a “lifetime subscription.” In February 2016 Nest announced that the hub and its app would stop working on 15 May. A customer, Arlo Gilbert, wrote that Google was “intentionally bricking hardware that he owns” and asked which hardware it would choose to brick next. The Federal Trade Commission opened an inquiry into whether disabling a product people had paid for was an unfair practice; it closed the inquiry in July 2016 after Nest offered full refunds, while noting that consumers would not reasonably expect a hub to stop working. The case became a standard citation in the argument over whether owners of connected devices own them at all.

When Google closed its game-streaming service Stadia on 18 January 2023 (see Dead End: Google Stadia), it refunded all the hardware and games bought through its stores, and the day before the shutdown released a browser tool that switched the Stadia controller’s dormant Bluetooth radio on, so the controllers could be used with other devices. The tool was first available until the end of 2023 and was then extended by a year.

Dead Links and Deprecated APIs

Links break when their host goes. Google’s URL shortener goo.gl, opened in 2009, stopped accepting new users in 2018 and new links in 2019. In July 2024 Google announced that all existing goo.gl links, which had been printed in books, papers and on packaging, would stop resolving on 25 August 2025. On 1 August 2025, after protest from archivists and publishers, it reversed itself in part: only links that had shown no activity in late 2024 were switched off.

The same doubt followed Google into businesses that paid. In August 2020 Steve Yegge, a former Google engineer then working elsewhere, published “Dear Google Cloud: Your Deprecation Policy is Killing You,” arguing that Google treated breaking changes to its cloud APIs as a sign of good engineering hygiene while its customers experienced them as unpaid migration work, and that “backwards compatibility keeps systems alive and relevant for decades.” On 27 July 2021 Google Cloud published three “API stability tenets” for its enterprise APIs: no feature would be removed or changed incompatibly “for as long as customers are actively using it,” with exceptions only for security, legal or intellectual-property reasons; customers would get at least a year’s notice; and changes would pass a central review board.

Dead End: The Base Rate

The graveyard changed how people judged new Google products. By the time Stadia launched in November 2019, the question of how long it would last was part of how it was reviewed and bought, and its closure three years later confirmed the forecast. A company that kills a large share of its launches makes each new launch less attractive, because customers who have read the base rate hold back, which lowers usage, which makes the next closure more likely.

Google also kept its main products for decades: Search, Gmail, Maps, YouTube, Android and Chrome were still running in 2026, and the graveyard is in part a record of an unusually large number of experiments. What made it a trust problem was that the experiments had been offered to the public as products, with data and hardware inside them.

📚 Sources