If each button is seen as ad and paid for by the brand shown on it, the costs likely look good from both Amazon's perspective and the brand's perspective.
A button that costs somewhere between $5 and $10 might result in something like $240 worth of sales over its lifetime. (eg. 1 sale per month x $10 sale price x estimated 24 month lifetime)
I think the concern is that I have, for example, around sixty different spices on my spice rack. Do I stick a button behind each one? My cluttered cabinet with AA batteries, AAA batteries, CFLs, compressed air, WD40, several dozen other items? How about each item in my cluttered fridge?
I think the Amazon Fresh Direct concept mentioned elsewhere by mason55 makes a lot more sense. Stuck to the front of my fridge I want a little scanner--one on my fridge, one by my cabinet, one in my laundry room. Running something across the barcode scanner when it's low/empty is as easy as pressing a button, and doesn't mean I have to put a button on/beside every single item, especially ones that move around my disorganized shelves.
How frequently do you replenish your spices/WD40? If you are a restuarant, then a button would make sense. The dash, in its current form, seems suited to high-turnover products where the sales generated over the devices lifetime can cover the initial cost.
As someone else mentioned though, think of it as "Dash, the physical button" + "Dash, the API". The API is probably the long term goal, the physical button is just a (temporary) drop-dead simple API client that people can use today. Amazon is likely betting on smart-fridges/-pantries/-homes catching up in the future
Edit: in the near-future, I imagine Amazon adding a touchscreen (so it's not product specific) without adding much to BoM. They can also charge manufactures for on-device screen-time (who doesn't like ads that convert to an immediate sale?)
I think the next iteration is (or should be) a Dash button with an e-ink screen built in that can order any item and the screen changes to display the brand and name of the item it's setup to order.
Trying to add a changing ad screen to this would kill the battery life because of the check-ins to fetch the next ad. I also don't see that generating many sales because the customer would still need to change their set order, I think few people would accept Amazon outright changing what the button ordered based on the current ad.
A button that costs somewhere between $5 and $10 might result in something like $240 worth of sales over its lifetime. (eg. 1 sale per month x $10 sale price x estimated 24 month lifetime)
Seems scalable to me.