The Original Satellite-Telephony company
Firstly, it's worth highlighting the original satellite-based communication platform, Iridium. Conceived in the late 90's, Iridium is based off a constellation of 66 low-earth orbit satellites (originally 77 were planned) at an altitude of 781km. In its original incarnation, Iridium was designed primarily for voice communication (with basic slow-speed data), however Iridium NEXT currently being launched (2017/2018) adds full data capabilities and allowing for a maximum of 8MB/sec for terminals.
In Iridium's vision, it was, and is, always a supplement for earth-based mobile networks (fixed and wireless), with a primary benefit being global coverage in the many out-of-reach parts of the world (it's worth remembering that almost half the world's population is still not connected to the internet).
Elon Musk, founder of SpaceX slipped into a Tweet in November 2014 about a proposal/plan for an satellite-based internet system. At the time, it was very light information and definitely not picked up by many reports or articles (with the exception of tech site, TechCrunch.com). However, since then the proposal has continued to evolve and expand to the point where it would appear that Elon Musk, SpaceX (and Tesla) CEO, is setting their satellite-based system as a key pillar of the SpaceX business model.
SpaceX's proposal is a different proposition entirely to Iridium: instead of 66 satellites, SpaceX's is made up of 11,900+ satellites. This is on an order of magnitude difference to anything currently in place - for context, there is only 2,000 satellites in total in orbit right now. Of the 11,000 satellites, 4,425 satellites would be in low-earth orbit of 1,100 - 1300km: similar in concept to Iridium's only at a massively larger scale. This original 4,425 satellites is what made up the original proposal in 2016. This design also comes with the inherent limitations of satellites at that altitude: very high latency in the order of hundreds of milliseconds making real-time communication a challenge.
Added in 2017 is the other 7,500-odd satellites that would be in very-low-earth-orbit, at around 340km in altitude. The key piece to this is with this number of satellites at this lower altitude, is a targeted 1Gbps and latency of 24-35ms becomes viable and thus very usable from a range of perspectives. The other key piece of this is they can offer these speeds anywhere on the planet.
The design of the two layers is to provide a massive mesh network where traffic is routed from the VLEO satellites to the LEO satellites and globally as required: and for more technical information, I suggest digging into the full proposal documents.
SpaceX and an end-to-end holistic strategy (i.e. Elon's other companies)
As is always the case with Elon Musk projects, there is multiple layers to each business and all are interlinked. His electric car business now owns a solar-roof business to allow home-owners to grab their own power for charging their vehicles, and also owns a full battery business providing both home-storage batteries and large scale utility batteries - a full end-to-end solution for electricity supply and transport, but also a full end-to-end solution from a design, creation perspective.
From this car business, some components are re-used in the space business: the large monitors used in the Tesla's are also being used in the launcher that will be added to the rockets for carrying humans into space bring economy of scale reductions in pricing to these components.
As has been heavily publicised, SpaceX (along with Blue Origin, another company) is currently upending the space-launch business with reusable rockets (having successfully flown and landed several rockets as of July 2017) and bringing the cost of launches down from several hundred million dollars to several tens of millions, even when factoring in profits. However, SpaceX also realises that in time, this reusable rocket business will become a commodity business and therefore is looking at other opportunities for revenue streams for their long-term goal of setting up business on the Moon and Mars. Closing the loop, and tying into Musk's car business, at present all Tesla vehicles currently have a mobile data connection that is heavily used to upload data around use of vehicles so by building its own internet-access method, they are taking away this payment to whatever telecoms provider (currently AT&T in the USA, although I'm not aware of who is used worldwide).
It also appears that up to now, SpaceX has also included applications for space-based ground photography in their applications for the satellite-internet business (an application is required to allow inclusion of cameras on satellites) - a possible additional revenue stream through sale of ground imagery, as well as providing real-time updates to traffic forecasting, and changes to roads, buildings for use with autonomous driving - I suspect many car companies would be jumping at the high-resolution imagery.
What has also been demoed is a sample fixed receiver - in this case, built directly into solar panels and powered by the solar panels. Again, worth reminding that Tesla now sells solar panels and thus we're looking at a possible full integrated solution: mount these antennas on all houses along with all their vehicles and the possibility is there for a local, ground-based mesh network.
The key piece from SpaceX's satellite proposal is their timing when it comes to manufacturing of satellites. Based on the innovations coming via the smartphone revolution of the past 10 years, we are hitting a tipping point with component manufacturing both in size and in cost: Iridium satellites are measured at 700kg per satellite and requiring multiple launches for all 66. SpaceX highest satellites are measured in weight at less than half that, and the VLEO satellites can likely be launch 70 satellites at a time due to their size. As highlighted by the re-use of Tesla's dash monitors on a space rocket, the difficulty of building satellites is now easy for much shorter construction and difficulty time-frames. An analogy? When rumours came out about Apple developing their own mobile phone, most telecoms handset manufacturers refused to believe it based on belief that building a good telephony/network stack was a multi-year challenge. Instead, Apple developed the whole stack in only a few months as the complexity was significantly reduced by the mid-2000’s.
Timelines for SpaceX? The current proposal is for the first launches in 2019 with completion within a year or so: when you're re-launching rockets, by 2019 it's highly likely multiple rocket launches per week could be occurring (SpaceX has already shown in the past month they can launch two rockets in one weekend, and then launch another rocket a week later). On the flipside, all of Musk’s companies have missed target dates (up until the recent Model 3 announcement) so this timeline may be slightly aggressive – but it’s also worth reminding oneself that they have also managed their sky-high targets (and surpassing them), even if taking slightly longer than planned.