Amazon is about to go head to head with SpaceX in a battle for satellite internet dominance 

Elon Musk and Jeff Bezos are about to lock horns once again. Last month, the US Federal Communications Commission approved the final aspects of Project Kuiper, Amazon’s effort to deliver high-speed internet access from space. In May, the company will launch test versions of the Kuiper communications satellites in an attempt to take on SpaceX’s own venture, Starlink, and tap into a market of perhaps hundreds of millions of prospective internet users.

Other companies are hoping to do the same, and a few are already doing so, but Starlink and Amazon are the major players. “It is really a head-to-head rivalry,” says Tim Farrar, a satellite expert from the firm TMF Associates in the US. 

The rocket that will launch Amazon’s first two Kuiper satellites—the United Launch Alliance’s new Vulcan Centaur rocket—has been assembled at Cape Canaveral in Florida. Its inaugural launch is set to fly two prototype Kuiper satellites, called KuiperSat-1 and KuiperSat-2, as early as May 4. Ultimately, Amazon plans to launch a total of 3,236 full Kuiper satellites by 2029. The first of that fleet could launch in early 2024.

“They have ambitions to be disruptive across the technology sector,” says Farrar. “It’s hardly surprising that they’ve jumped in here.”

In the past few years, companies have been trying to expand access to the internet via satellite, both as commercial ventures and to supply internet to those in remote locations without otherwise easy access. Starlink, the mega-constellation of more than 3,500 satellites built by Musk’s SpaceX, is the biggest of these ventures. 

Amazon announced Project Kuiper in 2019, the same year Starlink began launching, leading Musk to tweet that Bezos, then the company’s CEO, was a “copycat.” Others are in development too, such as the UK-based OneWeb, which currently has more than 500 satellites. But Farrar says the key competition is between SpaceX and Amazon.

To take on SpaceX, last year Amazon revealed it had essentially bought all the spare rocket launch capacity in the world (although with little effect on its rival, because SpaceX launches satellites on its own rockets). Thanks to Amazon’s multibillion-dollar deals with United Launch Alliance, Bezos’s Blue Origin in the US, and Arianespace in Europe, Project Kuiper satellites are expected to fly on 92 different launches over the next five years.

The rapid launch cadence is important. Under its license with the FCC, Amazon has until July 2026 to launch half its constellation. “We are on track to meet that deadline,” an Amazon spokesperson said. Last month, the FCC gave Amazon the full green light to begin launching its satellites after the company finalized details of its plan to address concerns about its potential to increase space junk.

But there is a catch: none of the rockets Amazon has bought a ride on has yet made it to space (in fact, one launch vehicle Amazon had initially planned to use exploded in January). “Those rockets are largely behind schedule,” says Farrar.

The satellites are meant to orbit at an altitude of about 600 kilometers and cover latitudes from Canada to Argentina, reaching “95% of the world’s population,” the Amazon spokesperson said. “Our constellation will serve individual households, as well as businesses, schools, hospitals, government agencies, and other organizations operating in locations without reliable broadband.” 

Amazon has applied to the FCC to increase its constellation to 7,774 satellites, which would allow it to cover regions further north and south, including Alaska, as Starlink does.

There are riches to be had: SpaceX currently charges $110 a month to access Starlink, with an up-front cost of $599 for an antenna to connect to the satellites. According to a letter to shareholders last year, Amazon is spending “over $10 billion” to develop Kuiper, with more than 1,000 employees working on the project. Andy Jassy, Amazon’s current CEO, has said that Kuiper has a chance of becoming a “fourth pillar” for the company, alongside its retail marketplace, Amazon Prime, and its widely used cloud computing service, Amazon Web Services

“Amazon’s business model relies on people having internet connectivity,” says Shagun Sachdeva, an industry expert at the space investment firm Kosmic Apple in France. “It makes a lot of sense for them to have this constellation to provide connectivity.”

Amazon is not yet disclosing the pricing of its service but has previously said a goal is to “bridge the digital divide” by bringing fast and affordable broadband to “underserved communities,” an ambition Starlink has also professed. But whether costs will ever get low enough for that to be achievable remains to be seen. “Costs will come down, but to what extent is really the question,”  says Sachdeva. On March 14, the company revealed it was producing its own antennas at a cost of $400 for a standard antenna, although a retail cost has not yet been revealed.

Amazon has said it can offer speeds of up to one gigabit per second, and bandwidth of one terabit per second. Those are similar to Starlink’s numbers, and the two services seem fairly similar overall. The key difference is that Starlink is operational, and has been for years, whereas Amazon does not plan to start offering Kuiper as a service until the latter half of 2024, giving SpaceX a considerable head start to attract users and secure contracts.

The astronomy problem

There remain concerns, too, about space junk and the impact on ground-based astronomy. Before 2019 there were only about 3,000 active satellites in space. SpaceX and Amazon by themselves could increase that number to 20,000 by the end of this decade. Tracking large numbers of moving objects in orbit—and making sure they don’t collide with one another—is a headache.

“I’m not satisfied that we can safely sustain [even] one of these systems in orbit,” says Hugh Lewis, a space debris expert at the University of Southampton in the UK, who has tracked thousands of close calls between Starlink, OneWeb, and other satellites. “They’re continually rolling the dice. At some point, in spite of all their best efforts, I think there will be a collision.”

Amazon’s spokesperson said the company had “designed our system and operational parameters with space safety in mind.” When satellites finish their mission, the spokesperson added, they will be removed from orbit within one year using onboard thrusters, and in the case of satellite failure, atmospheric drag will “help ensure any remaining satellites will deorbit naturally.”

Amazon has not revealed the size of its satellites, but—like Starlink’s—they might reflect enough sunlight to pose a problem to astronomers and even change the appearance of the night sky. Attempts to lessen the impact satellites have on astronomy have been moderately successful at best, with the satellites appearing particularly bright at twilight. Telescope observations of the universe are already affected by bright satellite streaks, and the problem is likely to worsen in the future.  

Amazon has said it is working with astronomers on the issue. “Reflectivity is a key consideration in our design and development process,” the company spokesperson said. “We’ve already made a number of design and operational decisions that will help reduce our impact on astronomical observations.”

If the problem cannot fully be solved, however, some aspects of astronomy will become much more difficult or even impossible. “Starlink has not managed to make their satellites nearly as faint as they promised,” says Samantha Lawler, an astronomer at the University of Regina in Canada. “I’m quite worried what the sky will look like with yet another company launching thousands of potentially bright satellites.”

With plans to build up to four satellites per day, Amazon plans to progress rapidly. After its first two test satellites have launched, the rest could come thick and fast. Can the company take on Musk? “That’s the big question,” says Farrar. “They have to move quickly.”

This story was updated on 23 March to clarify the figure of $400 is the cost to build a standard Kuiper antenna and to correct a typo regarding Project Kuiper’s bandwidth.

Starlink signals can be reverse-engineered to work like GPS—whether SpaceX likes it or not

Todd Humphreys’s offer to SpaceX was simple. With a few software tweaks, its rapidly growing Starlink constellation could also offer ultra-precise position, navigation, and timing. The US Army, which funds Humphreys’s work at the University of Texas at Austin, wanted a backup to its venerable, and vulnerable, GPS system. Could Starlink fill that role?

When the idea was first proposed in 2020, executives at SpaceX were open to the idea, says Humphreys. Then word came from on high. “Elon told the leaders we spoke to: every other LEO [low Earth orbit] communications network has gone into bankruptcy,” Humphreys told MIT Technology Review. “And so we [SpaceX] have to focus completely on staying out of bankruptcy. We cannot afford any distractions.”

But Humphreys wouldn’t take no for an answer. For the past two years, his team at UT Austin’s Radionavigation Lab has been reverse-engineering signals sent from thousands of Starlink internet satellites in low Earth orbit to ground-based receivers. Now Humphreys says his team has cracked the problem, and he believes that regular beacon signals from the constellation, designed to help receivers connect with the satellites, could form the basis of a useful navigation system. Crucially, this could be done without any help from SpaceX at all. 

In a non peer-reviewed paper that he has posted on his lab’s website, Humphreys claims to have provided the most complete characterization of Starlink’s signals to date. This information, he says, is the first step toward developing a new global navigation technology that would operate independently of GPS or its European, Russian, and Chinese equivalents. 

“The Starlink system signal is a closely guarded secret,” says Humphreys. “Even in our early discussions, when SpaceX was being more cooperative, they didn’t reveal any of the signal structure to us. We had to start from scratch, building basically a little radio telescope to eavesdrop on their signals.”

To get the project started, UT Austin acquired a Starlink terminal and used it to stream high-definition tennis videos of Rafael Nadal from YouTube. This provided a constant source of Starlink signals that a separate nearby antenna could listen in on.

Humphreys quickly realized that Starlink relies on a technology called orthogonal frequency-division multiplexing (OFDM). OFDM is an efficient method of encoding digital transmissions, originally developed at Bell Labs in the 1960s and now used in Wi-Fi and 5G. “OFDM is all the rage,” says Mark Psiaki, a GPS expert and aerospace professor at Virginia Tech. “It’s a way to pack the most bits per second into a given bandwidth.” 

The UT Austin researchers did not try to break Starlink’s encryption or access any user data coming down from satellites. Instead, they sought out synchronization sequences—predictable, repeating signals beamed down by the satellites in orbit to help receivers coordinate with them. Not only did Humphreys find such sequences, but “we were pleasantly surprised to find that they [had] more synchronization sequences than is strictly required,” he says.

Each sequence also contains clues to the satellite’s distance and velocity. With the Starlink satellites transmitting about four sequences every millisecond “that’s just wonderful for dual use of their system for positioning,” says Humphreys. 

If the terrestrial receiver has a good idea of the satellites’ movements—which SpaceX shares online to reduce the risk of orbital collisions—it can use the sequences’ regularity to work out which satellite they came from, and then calculate the distance to that satellite. By repeating this process for multiple satellites, a receiver can locate itself to within about 30 meters, says Humphreys.

If SpaceX later decided to cooperate by including additional data on each satellite’s exact position in its downlinks, that accuracy could theoretically improve to less than a meter—making it competitive with GPS. SpaceX did not respond to requests for comment.

Other researchers have been treading a similar path. Zak Kassas is a professor in the department of Electrical and Computer Engineering at Ohio State University and the director of a US Department of Transportation center focusing on navigation resiliency. Last year, his team was the first to demonstrate that Starlink signals could be used for positioning, in part using machine learning. 

Kassas’s approach, which he calls cognitive opportunistic navigation, analyzes the period and changing frequencies of signals from a satellite as it travels overhead. The receiver also uses the synchronization sequences, learns the satellite’s orbit, and tracks it. With multiple satellite passes, the receiver ultimately calculates its own location. At a recent conference, Kassas claimed his system had now achieved accuracies of less than 10 meters with Starlink. “It’s a framework that is so general we can apply it to any terrestrial or extraterrestrial signal,” he says. “It will learn on the fly, tell you what is being transmitted, and tell you where you are.”

A fuller understanding of Starlink’s signals has implications beyond navigation. For instance, the Starlink satellites currently don’t seem to be using two of the eight channels that SpaceX is licensed for. Humphreys speculates that this could be because Musk is keen not to interfere with radio telescopes operating at neighboring frequencies. The bright streaks of orbiting Starlink satellites have already been accused of disrupting optical astronomy.

UT Austin’s findings also highlight the possibility of deliberate interference with Starlink itself. Humphreys notes that while the synchronization sequences hold promise for navigation, the fact that they are utterly predictable and are used across the whole constellation is a security vulnerability. “Humphreys has done a big service to the navigation community identifying these sequences,” says Psiaki. “But any navigation system working on open-source sequences could definitely be spoofed, because everyone will know how to spot those signals and create fake ones.”

Starlink reportedly suffered a catastrophic loss of communications in late September in Ukraine, where it is being widely used for voice and electronic communications, to help fly drones, and even to correct artillery fire. Although it is unclear whether the outages were due to jamming by Russian forces, Musk tweeted last week: “Russia is actively trying to kill Starlink. To safeguard, SpaceX has diverted massive resources towards defense.”

Starlink has unquestionably been a lifesaver for Ukraine. However, reports of the outages and continued confusion about who will be paying for Starlink services there raise concerns over its future. 

“As time goes on and their dependence on Starlink deepens, Ukraine and its allies in the West are coming to appreciate that they have little control over Starlink and know little about it,” says Humphreys. “But now many millions have a vested interest in Starlink security, including its resilience to jamming.  Assessing that security starts with a clear understanding of the signal structure.”