The simple version: Tech companies sometimes launch products they expect to fail because the launch itself accomplishes something the product never could. The failure is a feature, not a bug.
What It Looks Like From the Outside
Sometime around 2019, Google launched Stadia, a cloud gaming platform. The gaming press was skeptical from day one. The latency problems were obvious to anyone who had played a video game in the last decade. The library was thin. The pricing made no sense. People who covered gaming for a living were openly baffled.
Google killed it in 2023. Roughly $800 million in the ground, give or take, depending on how you count.
The easy narrative is incompetence. Google is famously bad at building consumer products that stick, and there’s a long graveyard to prove it: Google+, Google Glass, Google Allo, Google Wave. You can spend an afternoon on Killed by Google, a site that exists solely to catalog these casualties.
But incompetence doesn’t fully explain the pattern. These companies aren’t staffed by idiots. They have enormous research budgets and sophisticated product teams. So what’s going on?
The Four Things a Doomed Launch Actually Buys You
Market signaling. Launching a product, even a bad one, plants a flag. It tells competitors, investors, and partners that you’re serious about a space. When Google launched Stadia, every major console manufacturer had to take cloud gaming more seriously, at least temporarily. Microsoft accelerated its own cloud gaming investments. Amazon launched Luna. Whether or not Stadia worked, it changed the competitive behavior of better-resourced players in the market. That has real strategic value.
Data you cannot buy. There’s no substitute for live user data, and you cannot fully simulate it. Surveys lie. Focus groups perform. But watching real people try to use a real product, at scale, reveals things that no internal testing process will surface. Google learned more about cloud gaming infrastructure, latency tolerance, and consumer willingness-to-pay from Stadia’s 18 months of operation than from any amount of internal modeling. That data feeds into every subsequent product decision.
Talent acquisition and retention. Ambitious engineers don’t want to maintain legacy systems. They want to build new things. A series of high-profile launches, even ones that don’t survive, signals to the market that your company is a place where people get to work on interesting problems. Google’s reputation as a place that takes swings, regardless of outcome, is part of why it continues to attract the kind of engineers who have options.
Regulatory and IP positioning. This one is underappreciated. Shipping a product, even briefly, establishes prior art. It supports patent filings. In some cases, it satisfies regulators who want to see evidence that a company is competing in a market rather than simply suppressing it. A launch is evidence of participation in a way that internal research is not.
The Accounting Trick That Makes This Rational
Here’s where it gets counterintuitive. For a company like Google, a product that burns $800 million over four years isn’t evaluated the way a startup would evaluate the same loss. It’s evaluated against what that same $800 million would have purchased in terms of information, market positioning, and talent.
The honest question isn’t “did Stadia make money?” It’s “was $800 million a reasonable price for everything Stadia bought us that wasn’t revenue?”
For a company with Google’s margins and strategic interests, the answer might genuinely be yes. The cloud infrastructure learnings from Stadia fed directly into YouTube and Google Meet. The latency research is relevant to every real-time product in Google’s portfolio. This is why Amazon sold the Kindle below cost for years and barely flinched: the Kindle wasn’t a product, it was a data collection mechanism with a screen.
Small companies can’t afford to think this way. For a startup, a failed product launch is often terminal. But large tech companies operate with a kind of portfolio logic that lets them absorb individual failures as long as the underlying strategic gains hold.
When It Tips Into Cynicism
None of this excuses the cases where consumers genuinely get hurt. Stadia users were sold lifetime game licenses that evaporated when Google shut the service down. Google eventually refunded game purchases, but only after significant public pressure. The people who built careers on Google’s developer platforms, or structured their businesses around its APIs, frequently have a rougher landing.
There’s a real difference between “we launched this knowing the odds were against us, and we learned from it” and “we launched this knowing we’d abandon it, and we let customers build on top of something we never intended to sustain.” The first is strategic experimentation. The second is something closer to using deliberate confusion to extract value from people who are trusting you with their time and money.
The tell is usually in how the company handles the shutdown. A company that genuinely ran an experiment and concluded it didn’t work tends to handle exits cleanly, with generous transition periods and honest communication. A company that knew from the start it was using customers as props tends to move fast and get defensive.
What This Means If You’re Not Google
If you’re building a startup, the lesson is not “launch stuff you know won’t work.” You don’t have the balance sheet to absorb those losses, and your early reputation is much more fragile.
The actual lesson is narrower: understand what a product launch is purchasing beyond revenue. If you’re launching something into a new market and the product has a 30% chance of working, ask what the other 70% buys you. If the answer is “nothing except a bad reputation and a depleted runway,” then don’t launch it.
But if launching gets you real user data, establishes relationships with key partners, attracts engineers who want to work on hard problems, or signals to your market that you’re a serious player, those are returns worth pricing into the decision.
Google’s mistake with Stadia wasn’t launching it. It was the way they treated the people who built lives around it when they decided to leave. The strategy was defensible. The execution of the exit was not.
