Why is Starlink terminal so cheap?

Phased array antenna is the magic sauce which makes Starlink feasible. Electronically steerable antennas are (or will be available) from multiple vendors and the prices are not publicly available. For example, Alcan aims to launch enterprise solution for 10000USD and consumer level antenna for 1000USD. Under a five year plan. Commercially available Kymeta u8 terminal has average price of 30,000USD.

So how come that Starlink offers phased array antenna with complete terminal for just $499? In 2020? One possibility is that SpaceX is heavily subsidizing terminals, hoping that their cost curve will catch up in the future. But that is not likely, as they have huge investments into buildout of gateways, satellite production, rocket launches, network development and future launch vehicle (Starship). And they aim to lower the price of the terminal to make it more appealing to the public.

But even if we assume that information from Business Insider article (2400USD per terminal) is true. That is a cost for a first million terminals. And it takes 19 months to recoup that investment. What will be the cost for the 2nd million order? And that is not all. What if SpaceX negotiated gradual payment of terminals...in several tranches. I would not be surprised it the actual payments are delayed. After all, such deal would represent 25% of STMicroelectronics yearly revenue. It would probably make SpaceX their largest customer. So from cash-flow perspective, even 2400USD per terminal does not seem unachievable goal. But it is still far below anything a competition has.

One possible answer is the target market. Starlink terminal has strictly targeted Ku-band frequency band which has been used for DTH and VSAT applications for decades. It simplified development and integration cost since it has to be vertically integrated inside the same company, with a single type of satellite.

But that cannot answer an order of magnitude cost reduction. The other segment is related to signal strength. Other phased arrays are usually optimized to work with geostationary satellites, which are 60 time further away. Which means that power flux or signal strength is 3600 times weaker. Which means that electronics in typical GEO optimized phased array must be far more precise in beam shaping and have better signal to noise ratio. Which means better electronics. Higher computing resources. More sensitivity to rain fade. 

Starlink might be using their own silicon. This is the first phased array antenna that will see a million plus production numbers. And there comes the learning curve effect. Better known as Moore's law for semiconductors, Swanson's law  for PV panels, or Wright's law of aeronautics. It basically estimates that average cost decreases 15% for every cumulative production doubling. Elon Musk counts on that. It was successfully applied to Merlin engines, with over 700 produced units. For Tesla batteries. For satellites. And now it will be applied for phased array antennas. Lets say that target production rate for user terminals is a million per year. If we estimate the cost for a thousand terminal batch to 30000USD, the same cost curve (using factor of 15% cost decrease) for the first million terminals would give the price of 6000USD. But that is a price for equivalent GEO-capable antenna. LEO is 60 times closer. Which means it needs 60 times less precise pointing accuracy. Or has much higher signal to noise ratio. Using less bits for digital phase shifters. Lower power amplifiers, or filters. Which means....cheaper components. Ordered in much higher quantities than before. Or producing their own ASICs. 

So SpaceX does have some secret sauce for user terminals, but the majority of this low terminal price is really based on a commitment of SpaceX to go big on terminal production. By going immediately to mass production of the terminals, they have achieved 5x cost reduction compared to competition. This is a big bet, that might cause the bankruptcy of SpaceX. But completely consistent with the way how Elon Musk operates. Tesla model S, model 3 had similar risk. Gigafactory too. Falcon 1 also. From financing point of view, even six month payment delay would provide almost half of the financing cost.