Alexa Was a Razor Built to Sell Blades. Nobody Ever Bought the Blades.

On November 6, 2014, Amazon announced a black cylinder that listened. The Echo arrived by invitation only to Prime members, went on sale to everyone the following July, and asked for $179.¹ That price was not the point. The point was what happened after the speaker landed on your kitchen counter — the things you would supposedly buy by simply asking. The hardware was the loss leader. The shopping was the profit. Ten years and tens of billions of dollars later, the second half of that sentence never showed up.

The official story is that voice computing was ahead of its time and the market wasn't ready. That's the comfortable version. The truth is that Amazon ran a razor-and-blades business in which customers happily took the razor and never bought a single blade — and an internal accounting metric kept insisting they had.

The razor sold. The blades never did.

The whole Echo thesis was Gillette logic: lose money on the device, win it back on what the device drives. Put a microphone in every home and people would reorder paper towels, restock coffee, summon products with their voice — and Amazon, sitting on the world's largest store, would collect. It is an elegant model when the device pulls customers toward the thing you sell. It collapses when the device pulls them somewhere free. And that is exactly what happened: people use Echo to set alarms, check the weather, and play music — tasks that cost Amazon money to power and return nothing.⁶ Amazon's own line is that more than half of Echo owners have used the device to shop. But former Alexa shopping-team employees say voice shopping never contributed meaningful e-commerce revenue.⁶ The razor went out the door by the hundred million. The blades stayed on the shelf.

How a money-loser kept scoring as a winner

Here is the part that makes Alexa more than just a bad bet. Amazon had a metric, developed internally in 2011, called downstream impact — DSI. Instead of asking what a product earned, DSI asked what spending it set in motion across the rest of Amazon's ecosystem afterward, and credited the product with that value.⁵ For the Kindle, DSI was honest, because the downstream was concrete: you bought a Kindle, then you bought e-books, and the e-books were real, traceable revenue. For Echo, DSI projected an optimistic stream of downstream shopping and subscriptions that user behavior never produced — and the metric sometimes double-counted the same revenue across different product teams.⁵ So a structurally loss-making device kept generating impressive numbers on its own scorecard. The thing being measured was never the thing being bought. The taste test, in effect, was rigged in the product's favor.

The bill nobody could itemize

The losses were enormous, but pinning a clean number on Alexa is its own trap — and almost everyone repeating the figures gets the trap wrong. Amazon's devices business lost more than $25 billion between 2017 and 2021, per internal documents.² But that figure covers Echo, Kindle, Fire TV, Ring and the rest — not Alexa alone. In Q1 2022 the 'Worldwide Digital' division lost more than $3 billion, and an employee said it was on pace to lose $10 billion for the year — but that division also houses Prime Video.³ The reason no one can hand you an Alexa-only loss is structural: Amazon doesn't break out Alexa or Echo as a segment in its SEC filings. Devices revenue and costs sit buried inside North America and International.⁴ A product that couldn't make money also couldn't be made to confess exactly how much it was losing.

When the cuts came, they were not gentle. Two rounds of layoffs in late 2022 and 2023 hit the Alexa and devices division hardest, part of more than 10,000 jobs eliminated in 2022 alone.⁸ By mid-2024, Bank of America analysts were warning that if a paid Alexa tier failed to catch on, Amazon could cut its investment further still.⁸ A decade after the cylinder shipped, the question was no longer how to win — it was how much more to spend before walking away.

But wasn't the data moat the real payoff?

The fair objection is that profit was never the goal — that Amazon happily runs products at a loss for years to win a category, and that putting a listening computer in hundreds of millions of homes was worth every dollar as a beachhead in voice. There's truth in it: there are now some 600 million Alexa devices in the world.⁷ But ubiquity is only an asset if it leads somewhere monetizable, and the entire failure was that it didn't — the homes filled up while the downstream stayed empty. The strongest version of the bull case isn't shopping at all; it's that the install base would one day matter for something else, like AI. And that is precisely the bet Amazon is now making — not by defending the old model, but by abandoning it. In February 2025 it launched Alexa+, a generative-AI assistant priced at $19.99 a month and then made free for every Prime member.⁷ Read that carefully: Amazon no longer trusts Alexa to earn its keep directly. It has folded it into the one subscription people already pay for. The razor-and-blades model wasn't fixed. It was quietly retired, and the device was tucked inside a bundle that was already profitable.

Alexa never made money because Amazon built a toll road and then discovered the drivers only wanted to ask it the time. The device did its job; the customers did theirs; the business model assumed a third thing that never occurred. And for years a clever internal number papered over the gap, crediting a loss-maker with revenue that lived only in a forecast. The most expensive part was never the cheap hardware. It was the decade Amazon spent believing its own metric instead of watching what people actually did with the thing in their kitchen.