Your browser is not supported. Please update it.

12 January 2021

Amazon ups the ante in satellite race with SpaceX’s Starlink

Heading into the holiday break, Amazon announced that initial testing of its Project Kuiper low earth orbit (LEO) satellite hardware was showing promising results. Since then, the pace of rival announcements seems to have increased, as the rest of the market realizes that between SpaceX’s Starlink and Amazon’s Kuiper, the horse is getting ready to bolt from the stables.

For Kuiper, the business case is pretty simple. Bundling broadband in with an Amazon Prime subscription, and ideally tying it in with a WiFi or mobile plan too, is a way to ensure that Amazon customers are always connected – always able to make that next purchase. Additionally, watching the flow of data should provide Amazon with a lot of valuable marketing and product design data too.

Crucially, as we understand that Kuiper faces the same cell capacity constraints as Starlink, Amazon would be opening the door to rural customers that are not currently as well served as their urban and suburban counterparts. While this is a mostly US-centric point, being able to sling Prime Video, Audible, Amazon Music, as well as the AWS-backed file storage services, is a way to increase the stickiness of the core Prime subscription too.

Providing far better broadband to these prospective Prime subscribers is a way to increase margins, and the technology has progressed sufficiently that it makes commercial sense to launch a satellite constellation to this end. For SpaceX, ostensibly Starlink is there to generate revenue for its missions to Mars, but there’s a nice second-order effect to be enjoyed by all the Tesla cars that will have data connections when out and about.

Starlink’s beta testing is already underway, as it has nearly 900 satellites in low earth orbit already. Amazon is a long way behind in the launch cadence, but has just completed testing of some of its ground equipment – specifically the antenna design that will be integrated into the end user terminals.

From a geostationary orbit (GEO), the test managed to stream 4K video, says Amazon, in “multiple environments” – presumably urban, suburban, and rural. The GEO orbit is around 35,000 kilometers, which is around 50x further than the LEO orbit that the satellite constellation will be deployed in.

In the test, the customer terminal design was able to hit throughputs of 400 Mbps, although it is not clear if that limitation was due to the terminal or the bandwidth available from that GEO satellite. The downlink was using the 17.7-19.3 GHz range (K-band), with the uplink in the 28.5-29.1 GHz range (Ka-band), and the terminal’s single aperture phased array antenna measures 30 centimeters across. Amazon claims this phased design hasn’t been done in the Ka-band before, and we believe this is the case.

For comparison, we do not know what the maximum throughput of the Starlink design is. We do know that the Starlink beta testers saw downloads between 11-60Mbps, uploads between 4.5-17.7Mbps, and latencies of 20-94ms. One user managed to reach 135Mbps downloads, in rural Idaho.

These are immensely impressive compared to incumbent satellite offerings, but it will be very interesting to see how Kuiper shapes up, as it would be a more equivalent comparison. We also wait to see how much the Kuiper terminal is going to cost the customers, as Starlink is asking for $600 upfront – $500 for the terminal, and the first month’s service.

We also know that Kuiper is subject to the same cell capacity limitations that Starlink has. There are only so many satellites that can rotate through the sky above a customer home, and so you have a finite amount of bandwidth to play with. In rural environments, you should be able to plentifully split this throughput with homes, but in urban environments with a dense customer footprint, your bandwidth would slow to a crawl.

So, while the terminal is capable of 400Mbps, we anticipate that the available speeds will be more in line with Starlink’s. The other wrinkle is that ‘4K quality video’ isn’t a concrete value these days. Netflix still recommends a broadband package with more than 25Mbps, while Amazon Prime recommends 15Mbps.

Of course, 4K is just the resolution. If we up the bit-rate of the signal, increase the color gamut, up the frame rates, and add HDR into the mix, suddenly the required bandwidth is far greater. This is the clash between 4K and the actual UHD standards, where the latter is a completely reinvented signal stream, while the former is just the old Full HD specs stretched out to 3840 x 2160p. The choice of codec would also make a big difference too, given the promises of the new VVC, EVC, and LCEVC suite.

Kuiper still has a way to go. It plans to begin operations when it has 578 satellites in orbit, and has publicly stated it plans to invest $10bn into its 3,326-unit constellation. Of course, CEO Jeff Bezos’ Blue Origin rocketry side-gig might afford Kuiper some launch discounts, but SpaceX has a distinct advantage over both Kuiper and the rest of the satellite market currently, on that front.

On that note, the European Commission has recently unveiled a consortium of firms that it is asking to evaluate a European-owned satellite network, to ‘strengthen European digital sovereignty and provide secure connectivity for citizens, commercial enterprises, and public institutions.’ The firms involved are Airbus, Arianespace, Eutelsat, Hispasat, OHB, Orange, SES, Telespazio and Thales Alenia Space.

OneWeb also announced it plans to provide commercial service to India in 2022, with North Europe and North America this year. Two deals with Intellian and Hughes show that work on the ground infrastructure is well underway, but we are still immensely skeptical that the UK government is going to get both a return on its investment and meet its national security and infrastructure goals, via the LEO endeavor.