Depends on the level of mnitoring a day. One ping a day, and inbound alerts on "quake detected"? A PIII on ADSL would probably handle that!
Or, of course, you could give the contract to EDS, and pay $38B.
You can't spec the back end system out on the non-earthquake situation.
The "value" in this guys solution is that you have a metric shitload of dumb sensors, and you process the data to determine the epicenter and then send out warnings appropriately. However once you are in an earthquake event, then you are going to get a sizable fraction of that metric shitload of sensors all instantly sending messages to the backend saying "Look at me! Look at me!", and your backend needs to be able to handle those messages without choking, perform the requisite calculations within fractions of a second, and then either get the 40 million warnings out instantly or try and decide what fraction of those warnings need to be sent - and again do that in a timely manner. . And that will take money for network infrastructure at the very least.
And given that it is intended as a life saving system you have to be able to regularly test and validate the reception of data and propagation of alert messages to a high level of certainty. Again, more costs. (If you have ever been in Hawaii then you should know that they test their tsunami sirens once a week)
Throw in some additional administrative costs for oversight and integration into *every* vendors cell network, and now you are talking about a decent chunk of change to run the entire system.