It sounds cheap, but the info I'm finding makes it seem feasible. They supposedly had 62 channels, with a final revenue of $14 million and net losses of $170 million. If each channel were 128 kbps, even with just a single continent, that could earn them $300 million a year at $10/MB.
Only until the rollout of data lines into Africa were to take that work away, but that still could take decades. Still, I wonder if this is even necessary. My understanding is that the rollout of cell phone carriers is rather substantial in Africa, despite the difficulties in getting Internet to those carriers. So the most likely scenario is that cell carriers would sponsor their own version of a BitcoinSpinner server on their local network, with a vendor locked app for their customers. Less than ideal, but probably still less than $10/mb.
Yet, this datacasting idea still has legs. I'd wager that it's no longer $10/mb, and that there is a per-broadcast fee involved, so that daily digests are still less expensive than hourly. However, hourly would significantly limit the possibility of a local double-spend attack against a subscriber. Doing the same thing with a more local DRM shortwave station would likley cost more relative to the subscriber base, but might offer the ability to send new blocks closer to their creation. Either way, a bitcoin datacasting company could stand to make money by using a trick out of the BitcoinSpinner playbook to reduce broadcasted data while also preventing 'free riders' by pre-emptively pruning the broadcasted blocks to include the headers, the merkle tree (perhaps complete, perhaps pruned) and only those transactions that have an address of a subscriber as either an input or output. This way, subscribers could present the datacasting company with a list of the addresses that they wish to include, and have a standard limit, say 100 addresses. This would also permit a local 'business association' to get several vendors who only use a couple of addresses apiece (i.e., one for the cash register, one for personal spending, one for a savings account and one for the wife?) and share on downlink equipment & subscriber fees. The datacasting company could, instead, charge subscriber fees based upon how many addresses it's scanning for; thus an entire town could have one downlink (local Internet service company, perhaps; trying to consolidate their own middle level Internet costs by blocking bitcoin & other p2p ports, but still provide the minimum level of service for customers?) and all the locally relevant data while excluding addresses and transactions that only matter to North America, Japan or Europe. In this sense, the downlink company becomes a trusted provider, but if any bad things happen the damage would still be limited and identifiable. Also, even though the digests would be encrypted & a hacker could extract that code from downlink equipment, that digest could still be 'signed' by the downlink company's own address private keys, which no hacker is going to have; so even the malicious things that an attacker could do locally would be limited even with a trusted blockchain provider.
In order to use this method effectively, portable light-client devices that have the ability to present a disconnected vendor's bitcoin client a copy of their own transaction plus their inputs & merkle tree positions would be necessary; but I think that is where we shall end up anyway.
