Project Loon

» see loon.com as it lived, on the wayback machine

~ the obit ~

project loon started inside google x around 2011 as a thought experiment about whether you could use stratospheric balloons as flying cell towers. the public announcement came on june 14 2013, and the first thirty test balloons floated over canterbury, new zealand two days later with fifty pilot users on the ground. the technology was unusual and the math was hard. balloons rode jet-stream wind layers at altitudes between eighteen and twenty-five kilometers, and the system steered them by raising or lowering each balloon into a wind layer that pointed the right way.

for about seven years loon kept flying. one balloon stayed up 312 days. the fleet accumulated more than a million hours of stratospheric flight by july 2019. loon graduated from x to become its own alphabet subsidiary in july 2018, then launched commercial service with telkom kenya in july 2020 covering about 50,000 square kilometers of rural highland. it also flew emergency-response missions: lte coverage over puerto rico after hurricane maria in 2017, then over peru after a 2019 earthquake.

on january 21 2021 alphabet announced the wind-down. the reasoning given was that the road to commercial viability had proven much longer and riskier than expected. internet availability in target regions had already reached 93 percent, and the remaining seven percent could not afford a 4g phone in volumes that would pay for fleets of self-piloting balloons.

~ the rap sheet ~

~ the rise ~

loon was one of the more legible google-x projects from an engineering standpoint. there was a real technical thesis. the autonomous-soaring software they built, which decided which wind layer to enter and when, is still cited in stratospheric-platform research, and softbank's later high-altitude-platform-station efforts hired some of the loon team. by 2020 the project had multiple national-spectrum agreements signed and a working tier of commercial pricing. it was not a science demo.

~ the actual end ~

loon failed at the last unit-economics step. flying a balloon for a year is one thing. paying for ground stations and spectrum and local-partner integration and a spare-fleet rotation across enough subscribers to break even at any reasonable price is something else. for the last-mile populations loon was aimed at, the willingness-to-pay number was below what the system could deliver. the smartphone-and-4g rollout in those same regions kept improving, and terrestrial coverage closed the gap from the ground up faster than loon could close it from the sky down. alphabet wrote off the project rather than chase a permanent subsidy model.

~ what survived ~

loon left a few things behind. its software stack and patents went to softbank's high-altitude-platform-station unit, which kept the autonomous-stratosphere work alive. its disaster-response model became the proof case for follow-on programs at att flycoms and at aalto haps. the consumer story it never produced was the more interesting one to me as an engineer. i work on edge ai and uav vision in ankara, and the loon obituary reads less like a moonshot failure and more like an architecture decision that did not survive contact with the real per-user revenue curve in the markets it was designed for.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »