M2M and the Internet of Things

Hardware costs for M2M systems can vary significantly, depending on the complexity of the action(s) being performed, communications methods, and the desired remote control capabilities. However, once you've figured out your core needs, there's still a considerable amount of cost variability based on whether you choose to use off-the-shelf components, or custom design something explicitly for your needs, as we learn from this Quora question:

"I want a simple device that reads low-rate data from USB input, and periodically sends it over a mobile network. Not much data, GPRS is good enough. What kind of price point could I expect for the hardware, in volumes of say tens to hundreds of thousands"

Unless you plan to manufacture at least high tens of thousands of devices, I would strongly recommend looking for an off-the-shelf solution, as certifying a new cellular module is not just expensive (Stephen's estimate of $100-300k is reasonable, if not a little low), but it's also very time consuming and fraught with other problems. Additionally, if you plan to deploy your devices on multiple networks (e.g. AT&T and Verizon), you'll need to certify both variants since they'll use different modems.

How "simple" a device you can get will also depend on what other functionality you may need. For example, will the devices need to be programmed with some data (like a GUID, checkin address or other configuration variables)? Will this be done during manufacturing/assembly, or by the end-user? Will there need to be a webserver or other user-interface method, or will you be relying on a physical interface like GPIO, RS232 or even a touchscreen/programming buttons? Will the device need to be field upgradeable, or take software/firmware updates over the air? Will you need 2-way remote device management, or are you only ever planning to "pull" data from the device?

If you do think you'll require multiple carrier support using different 3G technologies, it may be considerably less expensive and logistically easier to manufacture a generic (and simple) mainboard and plug an already-certified modem into it -- especially since it sounds like you'll already need USB support. This would also let you prototype inexpensively by using an off-the-shelf embedded computer during development, and you'd simply cost-reduce it before your manufacturing run.