As I pointed out in a tweet a couple of months ago, Iridium’s SBD service is being used for command and control of Google’s Project Loon. So it was interesting to see just how much Google has been spending on Iridium airtime, when Iridium’s CFO mentioned in their July 30 results call that:
“…our network provides the connectivity to remotely command and control the assets of the large and unique project by a major company who doesn’t let us reference their involvement in the program. We saw significant airtime usage in last year’s third quarter during the testing phase for this project. We now understand from our customer that this high level of activity will decline in the second half of 2015 as the service moves into another, more mature development phase, which will culminate in commercialization in 2016. We expect a full-year decline of $500,000 in M2M service revenue from this customer as a result of this evolution, with much of that coming in the third quarter.”
Its been reported that the Loon balloons have flown for “more than three million kilometers” at speeds of up to 300km/hour, though an average speed of say 40-50km/hour seems more plausible (which would mean it takes 50-60 minutes for the 40km diameter coverage area to traverse a given location if directly overhead, or somewhat less if the balloon path is more distant).
So that would suggest Project Loon has achieved something like 60,000-80,000 flight hours in total over the three years of the project, with a significant fraction of that during the 2014 testing phase. Much of the spending on command and control was likely incurred in 2014, because Google reportedly moved to sending new orders to the balloons “as frequently as every 15 minutes” (and presumably receiving data from them even more often).
But if Google spent something over $500K on wholesale Iridium airtime (and even more with retail markups included) in 2014, then that would suggest the cost of airtime command and control is something like $8-$10 per hour (before retail markup). As a benchmark, the spending level of about $140K per month in Q3 of last year suggested by Iridium would then equate to an average of 20-25 balloons operating continuously during the quarter (which is consistent with Google’s suggestion that it would step up to “more than 100″ balloons in the next phase of testing).
Google has indicated that the operating costs of each balloon are “just hundreds of dollars per day” but it is still surprising to consider that the company would be spending $200+ per balloon per day just on satellite connectivity. Moreover, it seems that Google’s “hundred of dollars per day” quoted cost could potentially exclude all the other costs involved in manufacturing and deploying the balloons and backhauling the traffic carried by them. That seems pretty expensive compared to the costs of a new fixed cellsite and highlights the perhaps questionable economics of the Loon architecture.
Now that Google has announced an MOU to potentially bring internet to remote areas of Sri Lanka next year, it is also interesting to contemplate just what that might mean in terms of Iridium airtime if the deal comes to fruition. Google has said it needs “more than 100 Loon balloons circling the globe” just to provide “‘quasi-continuous’ service along a thin ribbon around the Southern Hemisphere”. So it seems implausible to think that all of the rural areas of Sri Lanka would be served with less than say 300 balloons operating continuously. Assuming Google could get a somewhat better deal for high volume usage of say $5 per flight hour (of wholesale revenue to Iridium), then that would equate to annual wholesale airtime revenues of perhaps $13M for Iridium. And revenues could be even higher if more balloons are used to ensure continuous reliable coverage.
Perhaps Google can afford to spend a few tens of millions of dollars a year for a demonstration project in Sri Lanka (although the funding sources for this project remain uncertain). However, the scalability of Loon to a global deployment must be in much greater question. For continuous global coverage there would need to be as many as 100,000+ balloons in operation simultaneously. Even ignoring capital costs, if the operating costs of the network (for all aspects, not just satellite connectivity) are of order $300 per balloon per day, then that would amount to $11B per year in operating costs (for comparison US wireless carriers are projected to spend $56B in opex between them in 2017 to serve well over 300M customers). Its therefore unsurprising that Google intends to rely on wireless operators (and perhaps governments) to support these costs, rather than taking on the burden of commercial deployment itself.