$5B is around 1% of their (current) trailing 4 quarter revenue, which I would say is "not outrageous" if voice/home assistants is an area of growth and investment for Amazon.

Amazon doesn’t release info about the Alexa category in their earnings, so any reporting can only rely on whatever figure has been leaked.

That seems like reporting on Meta's 2022 losses in virtual reality in late 2026, which would seem absurd, even if that's the best/only information they have at that time. (At some point, you have to admit "we don't have any relevant information" rather than report on "the least irrelevant information we can find".)

The story does note how old the figure is, so readers can take it with a grain of salt as appropriate.

You could think of reporting on selective leaks like this as kind of a needling for Amazon to release more up-to-date info.

Amazon's hallmark in the past has been investing big into businesses without immediate profits. Amazon gets massive revenue but has nowhere near the margins of an Apple, or Google. As you know, a net loss and revenue are not apples to apples. For full year 2018, Amazon had net income of $10B on $232B of revenue.

All of which is to say that a $5B loss is a very material number. They obviously thought it would be a big profit driver in the future, which it has not turned out to be yet.

I just don't see the path to profitability for Alexa. Like, sure, it's neat, but how does it make any money?

Do they sell the devices at a profit?

Does usage of the devices lead to increased sales on Amazon.com? Are those sales that wouldn't have taken place on the app/website otherwise?

It's an ads vector

Eventually people will say "Alexa buy my groceries"

Alexa will buy from the highest bidder

The only kind of people I imagine would be okay with a 3rd party blind shopping for them would be those who haven't seen the inside of a store in years because they have help that does all of their shopping and domestic work already.

it may also have to do with price insensitive customers.... if you doordash every meal, even groceries from the highest bidder remain cheap in comparison.

It's one banana, Alexa. What could it cost? $10?

This has been promised for a long time, but from my perspective, I don't really see this getting any closer to reality.

But for what it's worth, I think there's a conflict of interest here. No one is going to believe that Amazon won't prioritize Fresh or Whole Foods for these orders - even if they don't.

How is this any different from a physical grocery store deciding on what to position on the ends of the rows due to incentive? People love to act like these problems brought on by the internet are new but all to often it's just an evolution.

At least my grocery still is reasonably organized and well laid out. If I need ketchup, I know exactly where to go to find all of my options side by side.

If I tell Alexa to order ketchup, who knows what brand I'll get or what size? It's possible I might end up with banana ketchup or even mayonnaise given how poorly Amazon's search functions work.

I use Ocado for my groceries. By default they will deliver based on a combination of things I've marked as "every X unit of time" and a predicted/suggested set based on past purchases. Then I can log in before a deadline to amend it. 90% of the basket typically remains unchanged, and quite regularly I can't be bothered to even log in to check what it's selected for me because I know it'll be close enough unless there's something specific I absolutely need this week.

No reason why Amazon can't do something similar. They only need it to be "good enough" that people start getting used to not always needing to ask for it to list what is in the upcoming order.

We’ll need Reg NMS for groceries.

> Does usage of the devices lead to increased sales on Amazon.com? Are those sales that wouldn't have taken place on the app/website otherwise?

That seems reasonable. Customers can order directly through the device ("Alexa, order some toilet paper"), that's low friction and lets Amazon rank options to their benefit. Plus regularly using Alexa keeps Amazon primed (no pun intended) in the customer's mind so next time they go online shopping they default to Amazon. Once you're using Alexa, there's an Apple effect encouraging adoption of other Amazon products through ease of use (Amazon Music, FireTV, Audible, etc). Amazon is also monetizing their products with ads, I'm sure that's part of the Alexa strategy as well.

Remains to be seen whether there is a "pull back" from smart devices that impacts the play. I was an early adopter of Alexa... at this point no one in my technical circle has proprietary "smart" home automation devices anymore, and the non-tech folks at best use Alexa to search in the Amazon app (mainly because they struggle with phone keyboards)

> at this point no one in my technical circle has proprietary "smart" home automation devices anymore

What are they using now, if you don't mind me asking?

Home Assistant almost goes without saying at this point.

Does it? I walk in tech circles, and actively work for a startup, and I don't think a single person in my circle of friends/coworkers has a Home Assistant setup, and _most_ have either an Alexa or a Google Assistant of some form.

Sure, but you didn't tell me that "no one in my technical circle has proprietary "smart" home automation devices". ;)

Home Assistant for home automation, Mycroft for voice control.

Probably nothing.

As others said - nothing (my preference), Mycroft, or Apple's Home Assistant (yes, propietary)

Thanks. I do however think that people are referring to home-assistant.io, not Apple Home Assistant.

> lets Amazon rank options to their benefit

This might get them in trouble eventually.

But is any of that profitable enough to justify having literally thousands of developers, probably costing a quarter million each on average? That kind of overhead isn't sustainable.

> Does usage of the devices lead to increased sales on Amazon.com?

Yes. We have a few Alexa's and it's helpful to reorder commodities with your voice as you run out of that item (batteries, tape, printer paper, etc). I just add them to my cart and check the price later, but I like it because I'll likely forget to reorder some of those things if I don't do it right away.

Apple and Samsung own our pockets with their devices (and associated platform), Amazon wants to own your counter top.

Alexa doesn't need to make money. It just needs to be another way to make it easier for people to buy from Amazon than from other websites.

Doubtful they sell at a profit. But do they allow developers for their apps (or whatever they're called, skills I think) to charge money and then take a cut just like any other app store?

I'd imagine like any other new product, they're taking a risk and hoping the product takes off with higher adoption so they can earn more. The bet on Alexa always seemed to be that people wanted a hands-off, voice powered way to interact with the internet, and perhaps they could charge for skills (creating their own 'app store') at some point, or earn more on selling products on Amazon because people would enjoy the convenience of just asking their device to buy more detergent or whatever. Not sure I ever really got the premise here, because it was always quite easy to order stuff on my computer or phone.

Anyway, I don't work at Amazon so I'm not sure if I'm mistaken about what the idea was. I imagine only some people close to that division would really know.

The most obvious path to me would be other businesses paying to integrate with Alexa. This would have required a level of platform buy in from consumers that never really materialized though, I think.

There was some belief, I think, that there was an app space for voice apps, similar to mobile. I don't think that's necessarily true, and despite Alexa focusing on making it relatively easy for developers to create their own VUI apps, there hasn't been (as far as I know), any real killer app developed in this space.

Apps on voice are so handicapped - it's really difficult to handle any one-time setup. I wanted to see if I could make a voice app that looked for emails on my third-party mail server but it's just not practical with the stateless function requirement.

The Echo devices are way way too dumb. The user's Cloud account needs app-accessible storage.

When Amazon Music Unlimited was an attempted thing, that was a direct path to monetization that basically only happens if you have Alexa/Echo widely deployed. Otherwise, it seems a tenuous link to increasing sales in retail.

(I say "was" because they recently rolled part of it out as a Prime benefit a couple of weeks ago, but I can't figure out the exact difference between unlimited and what's included in Prime [and don't care enough to chase it down], but they are still offering a $9/mo unlimited offering, so there must be some difference.)

The difference is that on Prime it's radio-style only. If you ask for a specific song you get a playlist of songs like the one you requested which may or may not include the song you asked for.

A massive limitation for some people, but it fits my usage.

It was perfect timing for me. Google just massively raised their prices for Youtube family so I was looking for alternatives.

The one that is included in prime is basically like spotify free without the ads.

You can only play in shuffle mode.

Having a microphone/speaker in people's houses seems like a great way to collect a ton of data & guide their behavior.

No security researcher has ever found proof that they are "always on" and recording everything people say. If they were, it would be front-page news and Amazon would have government investigations.

The devices listen for their wake word before ever transmitting data to the cloud. (This is also because the sheer amount of bandwidth needed for always-on would be un-economical even for Amazon).

This is not strictly true. I examined a master's degree thesis last semester where the student proved conclusively that Alexa devices are periodically spamming bursts of data back to Amazon HQ even when not woken. Beyond a heartbeat, too - MBs of data, not KBs.

No clear idea what that data is, though. There's at least a small chance it isn't benign.

So, mitmproxy that traffic and see what it is?

Scope of his work didn't go that far, he looked at several devices and Alexa was notably chatty. He was measuring for abnormal data bursts from IoT home devices - included things like smart bulbs etc.

isn't the point of SSL that you can't do that (assuming the client is verifying certs, etc)?

link?

Unpublished. It's still going through board approval.

As/when it's put on my institution's open repository I'll post a link. Might be a few weeks though.

Interested to hear more, once it's through! :)

Be sure to put as a submit :)

Amazon is surely aware of the aftermath of Google's accidental Wireless network war-driving.

While absence is hard to prove, an intentional or unintentional data collection beyond what is stated in their policies is highly unlikely.

I also don't think any insiders at any of these companies have leaked that they passively listen to what people say.

Do sample recordings, reduce bitrate (you can have good voice recording and playback at less than 3kb/sec) and upload only if the device detects something worthy of notification.

Hell, don't even send the sound, just send a weekly report of events.

What behavior do they hope to collect that is not a huge privacy liability with the effort? They already know where millions of people spend their money (not say they will buy, but actually follow through).

Smartphones and laptops are already that.

Is it collecting that data when not spoken to?

No.

>but how does it make any money?

"Alexa, put that on my shopping list" directs spending to Amazon in lieu of whatever the local alternative would have been.

They are localizers.

Their actual purpose(s) will be emergent and are still being discovered.

How is that much loss possible? Not a hardware engineer, but from the exterior it feels like there is nothing exotic about the design which would warrant extravagant development costs.

Have they been funneling billions of RD into the software side, or are the units comically underpriced and have a BoM in the hundreds?

My wager is that the devices are sold as a loss leader.

Then on top the software development is fairly novel - so they're probably spending quite a bit on development efforts related to the devices.

They're eating the costs for running the backend processing for all of these devices (they're sold as a one time fee, despite requiring services provided by Amazon servers).

Finally - they were promoting the ever loving shit out of the development tooling to app developers for a long time with a lot of seminars, development guides, free services on aws, etc.

---

My strong guess is they wanted them work out the same way Kindles have - loss leaders that more than make up for it by securing digital goods purchases. But personally - I spend lots of money on amazon for books for my Kindle, and I spend basically nothing related to Alexa (there's literally just nothing I want to buy through it).

Lately my Alexa has started inserting promotional content into the standard voice responses (bs like "Alexa what time is it" returns the time, and then a "and it's a good time to watch the new Rings of Power tv show? should I tell you about it?").

They're going to go into the trash shortly at this rate. I'll switch to a self hosted replacement like mycroft or something.

They’re also not like mobile phones where you buy a new one every year.

I still have my Gen1 Echo and it works as well as the new ones. I mean, it’s just a speaker and a microphone connected to WiFi. There are no “features” they added worth spending $100 for.

My intuition is that it's a) very high R&D costs coupled with b) being partly a hardware play, where unlike software, marginal cost of distribution does not tend to zero and where it's therefore much easier to lose money per unit sold.

In addition to the other replies, I'll add that they have a new frontier for Alexa in the works and while I don't think the hardware has to be all that exotic, the development effort will be. I know they're at least a year into development on that, and those costs will be high with nothing public to show for it (yet).

They’re probably selling every unit at a loss. A big part of the Amazon smart home stuff also has to be the data it collects and the way it integrates with ordering off the retail site which might not have been counted here.

They likely sell them at a loss.

Long-term, the hope is probably that the halo effect - "oh, I have Alexa, so I'll buy a Ring security system" and "Alexa, buy cat litter" sort of stuff - balances it out nicely.

Definitely. They have bundles with their Blink security cameras, which can be viewed on the Alex Show models.

I never got the value of voice devices.

In my experience you learn how to make them do 1-2 things. You learn a few patterns. But beyond that, even if they work 95% of the time, that 1 in 20 is annoying enough to make me not want to use it and just push a few buttons.

Maybe case in point is using voice in the car. It should be the best, most obvious application. Yet aside from maybe texting, people want things like Apple CarPlay - a big touchscreen - not voice commands.

I used to feel that way and I only had Siri devices since I had privacy concerns. Somehow I ended up with a free echo dot and it was significantly better. I got some smart lights since my kids could not remember to turn the lights off and now I have that on a privacy network associated with alexa and I've been impressed with the voice recognition. If you have a huge house being able to say 'alexa turn off the upstairs lights' or 'turn on the garage light' when you have a trash bag in your arms is pretty great.

Voice devices are actually very handy for those with mobility impairments. That being said, I have yet to meet anybody who actually uses these voice assistants daily (other than the aforementioned people with mobility issues).

Definitely use ours on a daily basis. Pause TV, play music, set timers, broadcast to other devices, set the thermostat, call people, and doorbell notifications. Nothing earth shattering or life changing but more useful than not.

My family uses ours daily and they're fully integrated into our routines. They're our alarm clocks that wake us, and are how we check the weather to decide how to get dressed in the morning. I ask them to turn on the lights in the morning, and to turn them off at night. Alexa locks my front door and closes my garage before bed. In the kitchen, every time we cook, we ask Alexa to preheat the oven or air fryer, and ask her to set the timers for whatever we're cooking.

It's not good practice to connect your security systems to your voice assistant, especially door locks. Maybe this isn't the case in your home, but in smaller apartments it's certainly possible for a malicious actor to say "Hey {voice_assistant}, unlock front door", and they gain access. It's the modern day "open sesame".

> it's certainly possible for a malicious actor to say "Hey {voice_assistant}, unlock front door", and they gain access. It's the modern day "open sesame".

I use Google Home and it has a voice match feature which will only accept on commands for allowed voices. Some people might not like this for the privacy aspect though.

Unless they've improved it over the last couple of years, Google's voice match is pretty easy to fool: When I played a recording of my friend's voice saying "hey Google" and then completed the sentence myself, the Home Mini thought I was him and let me access his calendar even though we have completely different voices.

Of course this is all academic. Anyone can get into a home via the windows. It's just that half of all Americans live in a home that has a gun, so getting out alive might not be as easy as getting in.

They thought of that many years ago, before adding lock support to Alexa. You can lock the door with a simple command, but unlocking or opening requires a pin code. Nobody can get into my house by shouting through the windows.

Oh wow, did not know that! Clearly I am only a recipient of this advice, and not a practitioner. I stand corrected.

Can you say your PIN? And if so, can a recording of you or a simulation of you saying it be played?

We use ours daily. Primarily it's a kitchen timer and a way to add items to our grocery shopping list. We used to use it for music but now tend to stream to a tabletop system with better sound. If it dies or becomes unsupported, it won't be a major loss but it does add a bit of convenience to our lives.

We might be a outlier but our family of four uses it at least 50 times a day if I look at our voice history.

Alarms, timers for the kids, playing music all day, asking for knowledge questions, etc. For example yesterday my kids asked "which is after a tortoise or a snail". The kids use it a lot because they don't have phones to google stuff but they always have Alexa to ask.

> people want things like Apple CarPlay - a big touchscreen - not voice commands

I don't know the answer, but i wonder why you say that's what people want, since the high end BMW and Audi and Mercedes cars etc now all have their own voice commands to _avoid_ using the touchscreen (presumably for driving).

I wonder how they even get a survey that's representative.

Because customers are demanding CarPlay

https://www.cnbc.com/2021/05/29/apple-carplay-massive-succes...

thanks (seriously cool article)

CarPlay and Siri is a pretty compelling pairing though. Looking/touching is dangerous while driving, so that is minimised by also being able to talk to the robot.

It would be nice were that actually the case. However, in practice, I tend to find that doing most things just using voice with my hands on the wheel and eyes on the road ends up being a really frustrating experience. This isn't a particular knock on Siri specifically. I'm pretty sure I'd find Google or Alexa similarly annoying in the same context. As someone says up-thread, Alexa at home is fine for a few formulaic things that I know the incantation for and pretty useless for everything else.

I'm not a big user of Siri, but I've had better luck than you describe. Siri will reliably play songs and podcasts I request; read out text messages; transcribe my response to text messages; and find directions to places. This is about as much as I really want.

I find podcasts and playlists pretty frustrating generally. But then I tend to want to listen to something fairly specific as opposed to more or less whatever as background; I've never been a big radio listener. I have used it for texts but that's very uncommon and directions can be hit and miss to make changes on the fly. It's definitely better than nothing but it's a far cru from a passenger doing that sort of thing for you.

I have the same use case but with very different results. “hotwords play my playlist foo” works for many values of foo with both google assistant and Siri. Ditto mapping, ditto sending texts.

Do you have an uncommon accent?

I don't get the value either, but i sure do miss it when I'm away.

Combine this with the privacy concerns and that's why I've never brought one personally. I know a lot of people who have them but rarely ever use them.

The only good use case I've seen is if you're having a conversation / debate with someone you can use them as a real-time fact checker.

I really only use my echo for a few things, but those things I do nearly daily and it provides enough value that I'd replace it if it broke. I don't need it to be an "everything device".

Kitchen use cases for voice for me:

Setting multiple named alarms while cooking and often my hands are not clean to handle a phone or they're holding something going in an oven or they're stirring something.

Adding things to Todo lists or shopping lists as I notice them.

Setting alarms when I set a cup of tea to brew so I don't forget about it. I have to walk back to the room to stop it ringing, which means I can add milk and pick up my tea.

Changing what I'm listening to while washing dishes.

I'm already distractible enough, not having to pick up my phone to do them also means I'm less likely to get distracted.

> and just push a few buttons.

There are just so many buttons to push, it's not only about lights. Remember, they have 'routines' that can easily equate to dozens of button pushes. Even without routines, I can tell it to set all lights in a room at some percentage. And no, I can't just wire a dimmer, because they are spread across light fixtures. Sometimes your hands are busy and it's good to be able to control stuff.

If anything, they should make it easier to add commands (I want to be able to change 3D printer settings without have to fiddle with the interface – it's connected to Homeassistant so it has access to controls).

It's not just for that either. Alarms, timers, unit conversions, translations, calls from echo to echo in different rooms, weather at some location, purchases (and asking where are they)... you can even ask if some food can be given to your dog.

I will agree that it's infuriating when they don't work. And it doesn't seem to difficult to fix. There are commands I use frequently, chances are I am using them again. Asking to turn of the lights and having it play some random song is ridiculous.

To me the key question is the value of voice devices. There may be good use cases for voice assistants – but many of those interactions can take place via your phone, which is always close by.

What do dedicated hardware devices like Alexa add above and beyond that? They have a nicer speaker for playing music, but there's a whole ecosystem of bluetooth speakers you can use your phone with. They can control smart lightbulbs which I think phones don't have the right antenna for... but it feels like that could be solved by a $20 Chromecast-like dongle that your phone talks to over wifi. Is there anything else left to justify separate hardware, especially hardware that's likely sold break-even or at a loss?

I worked for Alexa before. If I remember correctly there are only 4 use cases of it like weather, smart devices, music and timers/alarms. There used to be 10k employees at that time of which 3k were working in music alone. There was definitely a bloat at that time.

The third party ecosystem never caught up and there is no killer app for Alexa other than those 4 use cases. The API (slots/intents) is limited and hard to create any useful interaction with it. My gut feeling is many divisions will be affected in Alexa.

That describes me to a T. It's a convenient alarm in my bedroom and sometimes timer in the kitchen. Good for a cursory weather report if I don't need all the details. Sometimes play some music before going to sleep. Use it to turn my bedroom light on and off. That's about it.