Alexa is in millions of households and Amazon is losing billions
Amazon’s Alexa has reached hundreds of millions of devices yet is reportedly losing billions, prompting debate over whether the voice assistant ever had a viable business model beyond being a “smart timer.” Commenters argue that Amazon misjudged how people shop (they want to see prices and products, not order by voice), eroded trust with a chaotic marketplace and upselling behavior, and failed to evolve Alexa beyond basic home-control and media tasks. Many suggest that without a pivot to genuinely useful, privacy‑respecting assistant and smart‑home features—possibly powered by modern LLMs but not driven by shopping or ads—the platform risks stagnation or paywalled irrelevance.
Business model & strategy
- Many argue Alexa never had a clear, honest path to profit: it was justified with fuzzy “downstream impact” metrics (brand lift, extra Tide Pod sales) that were easy to game and hard to prove.
- Commenters see a classic “build scale then monetize” failure: huge org (thousands of people, multiple services per feature) built on the assumption of voice shopping that never materialized.
- Some view Alexa as a loss‑leader / marketing channel that helped lock in users to Amazon Music, Prime, etc., but not enough to justify ongoing losses.
- There’s broad skepticism that a new “AI Alexa” can be profitably run, given LLM compute costs versus what users will pay (few would pay $20/month just for voice control).
How people actually use Alexa
- Reported real‑world usage is narrow:
- Kitchen timers, alarms, clocks.
- Music/radio/podcasts (often via Spotify, sometimes Amazon Music).
- Simple factual queries, conversions, weather.
- Smart‑home control: lights, thermostats, TVs, intercom/announcements.
- Shared/shopping lists; package delivery alerts.
- Many note Alexa has gotten worse: more misrecognition, timers failing, wrong music, and intrusive “by the way” upsells.
- The Alexa app and “Skills” ecosystem are widely described as confusing, buggy, and poorly designed; feature discoverability is low.
Why voice shopping failed
- Strong consensus that almost nobody wants to buy “sight unseen” via voice:
- Users want to see product details, prices, quantities, and reviews.
- Amazon’s marketplace is seen as chaotic: volatile pricing, many near‑identical listings, 3P sellers, counterfeits, misleading unit pricing.
- This eroded trust makes “Alexa, order X” feel risky; at best people use it to re‑order exact past purchases or add vague items to a list, then buy manually later.
- Dash buttons and Subscribe & Save are cited as examples where price swings and substitutions broke blind‑ordering trust.
Subscriptions, openness, and alternatives
- Many reject paying a recurring fee to control lights or timers on hardware they already bought; others would pay modestly for a privacy‑respecting, non‑salesy assistant.
- Some want devices to be unlockable or flashable with open‑source firmware once Amazon abandons or monetizes them differently.
- A chunk of users are moving to Home Assistant, Hubitat, local storage cameras, or even simple “dumb” devices to avoid cloud dependence and surveillance concerns.
Organizational and technical critiques
- Internal culture is described as metric‑driven, political, and prone to “empire building,” making it hard to pivot or integrate modern LLM capabilities cleanly.
- Several see Alexa as constrained by Amazon’s retail incentives: hard to build a genuinely user‑centric assistant when the primary goal is to sell more Amazon stuff.