Alphabet announced it is shutting down Loon, the experimental project tasked with providing internet access to underserved regions through aerial connectivity.

The decision brings to a close a high profile effort that spanned nearly a decade, four continents and a panoply of telecom operator partnerships targeting disaster recovery situations through to general internet connectivity. Given its position as an early pioneer in aerial approaches to connectivity and the benefits Alphabet’s Google stands to reap as the remaining 50 per cent of the world (3.6 billion people) come online, the move could be seen as surprising. However, the combination of stubbornly unfavourable deployment economics, a lack of viable commercial partnerships with mobile operators, and a labyrinth of national regulatory barriers made Loon unsustainable. What can we make of this episode?

Quick primer: what is Loon and how does (did) it work?
Project Loon was established in 2011 as a new way of thinking to bring connectivity to people out of the range of terrestrial mobile coverage (and for whom satellite was prohibitively expensive). The method devised was to create a mesh network of balloons that would fly in the stratosphere at an altitude of 20km, about double that of a commercial airliner, and navigate through a complex set of aeronautical analytics that capitalised on the power of the earth’s trade wind patterns. Over the course of multiple iterations and improvements, balloons could fly for over 300 days and provide ground coverage to a radius of 11,000 square km. The delivery model was through a set of trial partnerships that involved Loon providing connectivity to base stations or other internet access points in remote regions. The vast majority of these were controlled by the mobile operators and jointly targeted to specific situations, such as providing temporary coverage in the aftermath of a natural disaster, or as a general access method for communities out of range of terrestrial 3G and 4G signals. Very few partnerships were for live and ongoing commercial deployments.

Stubborn economics in reaching the last 10 per cent
On the face of it, Alphabet’s decision is surprising. Project Loon was one of the first bets in Google’s X R&D factory, lasting almost a decade in an inherently high risk; high reward environment. It was consistently evangelised by Google’s management in mission terms, but with a logical purpose of providing a means of increasing the internet base that could eventually plug into the Android ecosystem and engage with Google services. Advances in ground based mobile network expansion the last five years mean that 3G coverage, the minimum speed for most internet activities, has now reached approximately 85 per cent to 90 per cent of population worldwide. This is, of course, a very good thing. However, the final 10 per cent still equates to some 600 million people (see chart, below, click to enlarge) which do not live within range of a 3G or 4G signal, undermining any prospect of mobile internet access. It is this cohort that Loon was seeking to reach.

To that end, Loon did establish a host of operator partnerships, notably Telefonica (Latin America), Vodacom and, most recently, Telkom Kenya.

The problem is that the costs of extending coverage to the final 10 per cent are disproportionately higher than all incremental expansions up to that point. We show in the below table a broad split of costs involved in the outlay and maintenance of mobile networks. The takeaway is that, on average, a single cell site becomes approximately 35 per cent more expensive to operate in remote regions compared with an urban/metropolitan baseline (see chart, below, click to enlarge), and this can be much higher in certain circumstances. The main reason for that comes down to backhaul links, given that Ethernet and microwave options are often not available and so rely on satellite. On the other side of the coin, the average spend (or ARPU) levels of a prospective consumer (or business) in remote regions is likely to be in line with, or below that, of the wider customer base for a given mobile operator. The risk is therefore of a sustained operating loss.

Alphabet’s blog post announcing Loon’s closure made explicit reference to costs: “While we’ve found a number of willing partners along the way, we haven’t found a way to get the costs low enough to build a long-term, sustainable business.” Commercial terms between Google and its carrier partners were never disclosed so it is impossible to discern the split of deployment costs and resulting revenue, or other form of reimbursement, between either party. However, it can be surmised that because most of Loon’s partnerships were trial-based, Telkom Kenya being the lone commercial agreement, Alphabet would have borne the lion’s share of capital outlay and therefore risk.

While Google does not split out the financial performance of its various commercial ventures, it does provide disclosure for the other bets line that amalgamates these. Total revenue for other bets in 2019 reached $659 million with a substantial operating loss of $4.8 billion, up from a loss of $2.7 billion in 2017. In short, for every $1 Google makes from other bets as a whole, it loses $7. Google tolerates this because of its vast financial reserves and because it regards it as a cost of doing business, which carries many non-financial benefits such as IP transfers. But for Loon, with the exception of Taara, there are few obvious technology transfers to other Google businesses, so the unfavourable RoI and lack of commercial operator partners made the bet unsustainable.

Where next? Read-across to satellite and other HAPS
Loon’s demise should not be interpreted as a death knell for aerial connectivity in general, but rather a reminder of the inherent challenges involved. The fact that 50 per cent of the global population remains unconnected is a problem that everyone has an interest in solving and that is not going away. Low Earth Orbit (or LEO) satellite constellations are among the leading methods currently in trial, with SpaceX, OneWeb, Telesat and host of defence sector participants having established live deployments and substantial funding commitments including, crucially, strategic investments from a range of mobile operators.

High Altitude Platform Stations (or HAPS) are also in play, including from HAPS Mobile and SPL, a partner of Deutsche Telekom.

The baseline cost realities of reaching remote regions are, of course, the same. Yet, both LEO and lower altitude HAPS plays are well positioned to provide an overall better level of network performance in speeds and latencies compared to Loon, with advances in ground station routing and reduced need for consumer CPE. Regulatory issues in air traffic control and mitigation of space debris will persist, but these are surmountable. The issue is simply that this all takes time and money. For all of its prowess, Google was operating in effect as a sole player in balloon connectivity, and therefore had to create and orchestrate a supply chain from scratch. LEO satellite and HAPS have more mature supply chains in place, willing partners and potential scale economies from common spectrum bands and equipment specifications, but deployments for most of the above companies will not reach scale (more than 50 per cent of vehicles in operation) for four to five years. So, the journey will go on, if only in different form but with the same underlying challenges.

– Tim Hatt – head of research, GSMA Intelligence

The editorial views expressed in this article are solely those of the author and will not necessarily reflect the views of the GSMA, its Members or Associate Members.