Challenges for Radio Communication in Space

Elon Musk is targeting space-based data centers starting in 2028, highlighting a long-term push in next-generation compute infrastructure. The article focuses on a technical challenge—communication between satellites and the ground—and on Observable Space, the startup founded by former SpaceX software engineering head Dan Roelker to address it. The piece is informational and early-stage, with limited near-term market impact.

Analysis

The investable signal is not “space compute” itself, but the infrastructure bottleneck it reveals: distributed orbit-to-ground networking, thermal management, launch cadence, and fault-tolerant compute orchestration. In the near term, the likely winners are not the headline space-compute aspirants but the picks-and-shovels layer that monetizes every iteration cycle—optical interconnects, radiation-hardened components, edge networking, and ground-station software. That favors incumbents with defense/space procurement exposure and software-defined network stacks, while pure-play “space data center” concepts remain capital-intensive, execution-heavy, and years from commercial scale.

Second-order, this is a forcing function for domestic compute strategy. If orbital compute is framed as a latency- and energy-relief valve for AI workloads, it increases optionality for companies already constrained by power availability, data center permitting, or grid interconnect queues. The practical implication is that space is more likely to start with narrow, high-value workloads—defense, remote sensing, resilient backup compute, and model inference bursts—than training frontier models, which remain overwhelmingly terrestrial because bandwidth and maintenance economics are still unfavorable.

The biggest risk is timeline slippage: the market tends to extrapolate sci-fi announcements into 2028 revenue, but the actual gating item is not launch capability, it is sustained network reliability and cost per usable FLOP after accounting for comms latency, radiation failure rates, and replacement cadence. A credible reversal would come from faster-than-expected progress in terrestrial power buildout and inference efficiency, which would reduce the urgency of off-world compute, or from a failed early demo that resets the narrative for 12-18 months. Near-term enthusiasm is likely to overshoot the actual addressable market; the underappreciated angle is that this is a software-and-orchestration story first, spacecraft story second.

Key Decisions for Investors

• Long NBIS / short speculative space-compute beta via private-market proxies if accessible: express the view that near-term AI demand still accrues to terrestrial infrastructure while orbital compute remains years away; target 6-12 month horizon, 2:1 downside/upside skew in favor of the short leg if enthusiasm gets ahead of fundamentals.
• Overweight defense/space infrastructure suppliers with optical comms and ruggedized electronics exposure (e.g., LHX, TDY, BAE) on a 6-18 month view; these are the more probable monetizers of iterative orbital networking spend before any true data-center-in-space economics work.
• Buy optionality on ground-station/networking software beneficiaries such as IRDM or VSAT on pullbacks; the trade is that each additional satellite compute initiative increases demand for relay, routing, and resilient backhaul, with limited downside if the theme stalls.
• Avoid chasing pure-play venture names tied to space data centers until there is a repeatable in-orbit demo and disclosed unit economics; current risk/reward is poor because financing risk rises faster than technical de-risking.
• If looking for a contrarian hedge, short high-multiple AI infrastructure names vulnerable to delayed capex conversion and concept-driven hype spikes, using 3-6 month call spreads rather than outright shorts to cap squeeze risk.