It's a really interesting question, and not something we had done enough research to speak authoritatively on. But I just did some more digging and looking into this, and here's my current thoughts:
Even with Starlink-style LEO constellations, fiber will remain the core backbone for fundamental physics and economics reasons:
1. Path loss: Signals in free space spread out, so by the time they reach Earth the received power is extremely low. That limits throughput per satellite to tens of gigabits. A single subsea fiber pair routinely carries 10–20 terabits per second, three orders of magnitude more.
2. Spectrum vs. optics: Satellites are limited to tens of GHz of allocated Ku/Ka spectrum. Fiber operates in the optical domain with essentially limitless bandwidth, and you can keep upgrading terminal equipment to increase capacity without touching the cable.
3. Upgrade cycles: Fiber capacity scales with new transponders; satellites are frozen hardware once launched. To scale, you have to build and launch thousands more satellites, which is far more expensive per bit.
The likely future is division of labor: fiber remains the bulk intercontinental backbone, satellites provide resilience and reach where fiber can’t go. Data centers in space may exist for specialized uses (defense, finance, satellite-native apps), but latency, hardware refresh cycles, and launch costs make them impractical for general-purpose compute. For almost all workloads, it will remain cheaper and faster to keep servers on Earth next to abundant power and fiber.
