Show HN: I did a 4 hour conversational audiobook on the history of data centers(stepchange.show)
stepchange.show
Show HN: I did a 4 hour conversational audiobook on the history of data centers
https://www.stepchange.show/p/data-centers-the-hidden-backbone
2 comments
This is a neat topic and love the format so far.
I’m curious to get your thoughts on the long-term future of the intercontinental submarine cables given advances in cheap high-bandwidth satellites, as proven out by SpaceX. In 100 years, are we still maintaining physical fiber across such distances? And is there a role and/or benefit of data centers in space?
I have a long drive coming up, will have to finish this and come back with more specific thoughts.
I’m curious to get your thoughts on the long-term future of the intercontinental submarine cables given advances in cheap high-bandwidth satellites, as proven out by SpaceX. In 100 years, are we still maintaining physical fiber across such distances? And is there a role and/or benefit of data centers in space?
I have a long drive coming up, will have to finish this and come back with more specific thoughts.
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.
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.
I put together a 4-hour conversational audiobook tracing their history—from IBM punch cards to the internet, the cloud, hyperscalers, and today’s AI factories.
It’s part of a new long-form series on the history of infrastructure (first two episodes were on coal).
Would love your thoughts—both on the story and on whether this “conversational audiobook” format works.