Starting with just six engineers in October 2018, the team behind Amazon’s proposed Project Kuiper satellite constellation has grown to more than 1,400 people as it prepares to launch beta broadband services from space next year.
The company plans to start churning out production satellites soon after deploying its first pair of prototypes in May, with launches for its 3,200-satellite Ka-band low Earth orbit (LEO) constellation slated to kick off in the first half of 2024.
Although far behind LEO broadband frontrunners Starlink and OneWeb, which are providing services in Ku-band, Project Kuiper is looking to leverage Amazon’s mass-manufacturing and logistical prowess to give it an edge over the competition.
David Limp, who is head of devices at Amazon that range from smart speakers to tablet computers, recently unveiled three Project Kuiper user terminals as part of this game plan.
Its standard terminal for residential customers is around the size of an LP record cover and promises speeds up to 400 megabits per second (Mbps). Starlink currently advertises download speeds of up to 100 Mbps under its standard residential service plan.
For the enterprise and government market, which OneWeb specializes in, Amazon has developed a larger terminal promising speeds up to 1 gigabits per second.
And on the smaller end, Amazon unveiled an ultra-compact terminal about the size of its Kindle ebook reader that would enable speeds up to 100 Mbps.
Limp issued a call for partners and customers after showcasing the antennas March 14 during the Satellite 2023 conference in Washington.
“[W]e haven’t been able to say this for the last four years, but [you] can kind of consider us open for business,” he said.
“We’ve got a lot of work to do to get these satellites up into the sky, but the technology is kind of vetted out enough that we’re pretty confident.”
Project Kuiper also benefits from Amazon’s colossal financial firepower. The internet retailing giant has committed $10 billion to get the constellation online, and a sizable chunk of this helped Project Kuiper secure the largest-ever commercial launch deal last year for up to 83 rockets.
The agreement comprises Vulcan rockets from United Launch Alliance (ULA), Ariane 6 from Arianespace, and New Glenn from Blue Origin, which is owned by Amazon’s billionaire founder Jeff Bezos. None of these launch vehicles have debuted yet. Indeed, Project Kuiper’s prototype satellites have hitched a ride on what would be Vulcan’s maiden flight.
The launch package is in addition to the nine Atlas 5 launches that Amazon previously bought from ULA.
SpaceNews sat down with Limp alongside Rajeev Badyal and Naveen Kachroo, respectively head of technology and business development at Project Kuiper, to learn more.
How many satellites do you plan to build before the end of 2023?
Rajeev Badyal: We’ll have several done, and by some point next year we’ll be able to build three to five satellites per day.
Do you know which rocket will be the first to launch these production satellites?
Badyal: We have an idea, but you know these things can dynamically change. If we’re ready even sooner than we anticipate, we might go with a different rocket than what we currently intend to use.
David Limp: It’ll most likely be Atlas 5. We have those first nine because they’re a proven vehicle, they’re flying, and have high reliability. It’s not the perfect long-term vehicle, but it gives us a great backstop of a rocket that’s available.
How many satellites could go on a single Atlas 5, or any other rocket you plan to use?
Limp: It depends. Each rocket has a different situation. There are a lot of variables: What orbit are we going to? What inclination? There’s not a one-size answer for that, but the purpose of going with medium-large and large rockets is you can get a substantial number on each one of them.
It’s not going to be a dozen or two dozen — it has to be significantly more than that, and it’ll vary from rocket to rocket.
You have 77 launches planned with options for 15 additional New Glenn rockets to cover most of the constellation. When do you need to start booking the rest?
Limp: It again depends. What is the final mass of the satellite? What is the actual performance of some of these new rockets that haven’t launched yet? If New Glenn comes out and it’s better then we’ll get more satellites per launch.
I’m not trying to be too cagey with the answer, it’s just there are a lot of variables. We made conservative estimates to buy enough heavy-lift launch capacity to ensure we could get the vast majority of the constellation up. Did we undershoot or overshoot that by a little bit? That’s all within the error bars right now.
You’re building and operating your own broadband satellites, like SpaceX, but unlike them you technically don’t have an in-house launch provider. How important is vertical integration?
Limp: We treat Blue Origin like any other launch provider and I think our ability to have competition in launch is good for us. It gives us the ability to get the best rockets for our needs.
I think [SpaceX’s] Falcon 9 is an amazing vehicle. It just wasn’t the right vehicle for our constellation. Part of me wishes it was.
But as for vertical integration on the satellite itself, we very much believe that makes us more nimble. I think it’s a big advantage.
Badyal: It means we can iterate fast and it gives us the ability to upgrade over time.
Naveen Kachroo: At the end of the day we’re going to be a network. If you think about all the pieces of the puzzle to deliver the network: the customer terminals, satellites, ground stations, the back-end networking, and you think about the service the customer buys, what you really want to do is manage that experience for them.
If you design that in-house with a vision in mind from day one, you don’t have to stitch together a network with something from the industry here and something there to try to create that network experience.
How many satellites do you need in orbit to provide initial services to large enterprise customers in the second half of 2024?
Badyal: Low hundreds.
Limp: It depends on the customer. There are some that might be able to do something with even lower numbers than that, and there are some that’ll have to wait for higher numbers.
In the early days, we’ll probably serve less end-consumers and more enterprise-like things.
When the constellation fills out in various latitudes, there’ll be a great consumer proposition in addition to that.
And adding more satellites would densify the coverage?
Kachroo: It adds a few things. Starting with a few hundred satellites, further deployments would fill in gaps so you can get continuous availability of the service. You need to make sure the customer terminal on the ground can see another satellite and do an instantaneous handoff.
But as well as covering gaps so customers don’t get any intermittent outages, deploying more satellites also increases capacity.
What kind of customer could make do with intermittent services?
Kachroo: There are a number of enterprise customers today that are really desperate for connectivity. They’re even willing to live with a few minutes of outages, just as one example.
So for those customers, something is better than nothing. They’re really excited just to just kick the tires with us even before we have continuous service availability.
These are areas where not even SpaceX’s Starlink is providing services?
Kachroo: I think they look at us as another viable option, and potentially a different type of service. Our terminals have a really competitive advantage from a size, weight, and performance standpoint.
You’ve said your standard, mid-sized terminal will cost less than $400 to produce — how does that compare with Starlink terminals that are marketed to consumers with price tags starting at $599?
Limp: I don’t know how much their build materials are but I know how much they’re charging customers, and it’s significantly higher than what we’re talking about for build materials.
I don’t know if that’s because they can price it at that or if they have costs around that. Starlink is using Ku-band for their spectrum and it’s a more complicated design. It’s bigger and has a motor.
As soon as you put a motor in anything — and by the way Amazon has products that we’ve shipped with motors — it adds a lot of complexity in the design.
I won’t speak for Starlink, but for our team we’re striving for simplicity in the design. Simplicity brings you three things: generally lower cost, generally a better customer experience, and almost always higher reliability.
A lot of companies talk about a minimal viable product. We don’t use that at Amazon. We use minimal lovable product — what would you take away and it’s still lovable?
Should I assume Amazon will subsidize these terminals like we’ve seen SpaceX do for Starlink?
Limp: I don’t think we have to decide that right now. I think there’ll be different business models for different customers and geographies. The only thing I can guarantee, because it’s true of how we price everything at Amazon, is that it will be affordable.
Will all three terminals be ready by the time you launch initial services next year?
Badyal: We’re developing them all but we haven’t decided how to roll them out yet.
In addition to producing 3-5 satellites a day, Amazon is going to need to build millions of these terminals — how many do you expect to get out the door daily?
Limp: Like all things, there’ll be a ramp up to a production level, but I’m not as worried about that.
With satellites, you’re trying to take the manufacturing of consumer electronics and bring it to something that hasn’t been there before, which is aerospace. So you want to vertically integrate — we’re building our own chip, working on our own solar panels and things like that.
Once we have the design right on the terminal, it’s effectively traditional consumer electronics manufacturing. And you know, while we sat here in this meeting, tens of thousands of Amazon devices have been produced. We’re at scale. I’m not saying it’s easy, but we have teams who know how to do this.
Amazon has committed $10 billion to Project Kuiper — how far does that get you and do you think you’ll need to inject more cash at some point?
Limp: We’ll have to see how the launches go and how customers react. Like a lot of things at Amazon we are tactically impatient and strategically patient.
We believe strategically, in the long term, customers are going to love this, and we’re going to provide a great need around the world in terms of broadband. It’s always hard to tell how quickly that happens because we have to get the product right and we have to get it done on time.
On the other side of the fence, we’re running as fast as we can to get the things we need to get done so that we can start hearing from customers.
I often say the easiest product you will ever build is the 1.0 because customers don’t tell you anything. After that, customers are divinely discontent, and you will never catch up to what they want.
When I got my first book from Amazon I think it took seven days to deliver. If I got something in seven days right now, I would be upset about it because the new normal is one or two days. Tomorrow it will be two hours and then it’ll be 30 minutes via drone.
Customers are always pushing us and that’s how it should be, and I think the same thing will be true here. We’ll launch, we’ll hear from customers, there’ll be a moving target and we’ll continue to make this better. We’ll build a minimal lovable thing for the first product, but it’ll get better and better over time.
So it’s possible customer revenues will mean Amazon doesn’t have to invest more money into Project Kuiper?
We are a for-profit business. We wouldn’t be doing this if we didn’t have a [trajectory] that builds a profitable business.
This interview has been edited for clarity and length.
This article originally appeared in the April 2023 issue of SpaceNews magazine.