Normally I would agree (The race to the bottom is huge) but in a market where the cheapest Samsung 32" TV in the UK on Amazon is £183 (~$240), The brand carries some weight for "quality", the data the module would generate can be sold on to generate income and you would be in a fine position to sign a custom deal with the carrier where you only pay for data if the customer didn't connect the device to their own connection the extra module itself is gonna cost $5. The extra $20 is only in data charges if they chose to exploit it.
Samsung could throw in the cheapest caps they could find from Shenzhen Markets, but they usually come with a decent brand (at least in the TV's I've taken apart in the past couple of years).
Sneaking in the extra cost of the modulke to the customer (if its as low as $5) will be pretty easy as long as the rev generated from the data collection is more than the data charges they are billed for (omitting the cost of dev'ing the FP'ing software and generating the FP's in your server farm).
EDIT: The 3 years of service was just a random number plucked out of the air. A length of time that could be used to study to see if the costs would be worth it (and a number I picked from experience with updates / new features no longer being pushed to smart TV's). If they found its not worth it, they could drop the service 12 months in and not have to pay for the 2 years left on the experiment (thus reducing the cost again).
The manufacturers have the numbers (number of units sold vs the number of units phoning home) to know if it would be a worthwhile endeavour. Just making the point that we are pretty much at the point where the costs of such connectivity is not a huge sum in the grand scheme of things.