Planet and Google Plan 2027 Orbital Compute Demo
Key insights
- Planet is flying Nvidia GPUs on Pelican satellites and has begun a Google collaboration framed as risk reduction ahead of a 2027 orbital compute demo.
- Data sovereignty is the consensus near-term use case: 141 countries have data localization laws satellite processing could address without cross-border routing.
- Current optical satellite links deliver about 100 Gbps; GPU training clusters require 7.2 Tbps, a gap of one to two orders of magnitude.
Why this matters
The energy constraints on terrestrial data centers, projected to consume 12% of U.S. electricity by 2028, create real commercial pressure to find alternative infrastructure, and orbital compute is now the first candidate being actively funded and hardware-tested rather than theorized. Data sovereignty laws in 141 countries represent a concrete, billable use case that does not require orbital facilities to match terrestrial performance, which shifts the investment calculus from science project to addressable market. The 5-10 year timeline to terrestrially competitive workloads means technical leaders and founders building AI infrastructure today need to track this space now to avoid being caught unprepared when the first revenue-generating orbital facilities come online in the early 2030s.
Summary
Planet is flying Nvidia GPUs on Pelican satellites. Its Chief Space Officer calls the 2027 Google collaboration risk reduction work, not a commercial launch.
Three bottlenecks drive the push: land, power, water. U.S. data center spending rose nearly 70% in one year; energy could hit 12% of U.S. electricity by 2028.
Essentially: Planet, Lonestar, Edge Aerospace, and Ramon.Space are building toward a market seen as real but not imminent.
- 141 countries have data localization laws: sovereign cloud is the consensus near-term use case.
- Optical links max at 100 Gbps; GPU clusters need 7.2 Tbps, a gap of one to two orders of magnitude.
- Small clusters: 2-3 years out. Terrestrially competitive workloads: 5-10 years out.
Orbital compute solves regulatory geography before raw performance.
Potential risks and opportunities
Risks
- Planet's Nvidia GPU Pelican satellites could face radiation-induced failures before the 2027 Google demo, damaging investor confidence across the orbital compute sector.
- The 100 Gbps optical link ceiling means early orbital data centers from Lonestar, launching Star Vault in October 2026, and Edge Aerospace will be limited to narrow sovereign-cloud tasks, constraining near-term revenue.
- Ramon.Space and Ingrasys/Foxconn's production partnership faces uncertainty if terrestrial demand for space-grade compute hardware stays too thin to justify scaling before the early 2030s.
Opportunities
- Lonestar's Star Vault satellite launching October 2026 and Edge Aerospace's March 2026 demonstrated mission give both firms a head start on sovereign-cloud contracts in the 141 data-localization-law countries.
- Ramon.Space's Foxconn/Ingrasys production partnership positions both firms to capture early hardware contracts as multiple orbital compute players scale toward deployment.
- Hyperscalers facing community resistance to terrestrial data center expansion due to the 12% U.S. electricity projection could become anchor customers for orbital facilities, accelerating the path to early-2030s revenue.
What we don't know yet
- SpaceX's collaboration with Anthropic on orbital compute: the nature, scope, and timeline of the partnership are not disclosed in the article.
- Whether next-generation 400 Gbps optical satellite links are commercially available yet or still developmental, and which orbital missions will carry them first.
- Funding amounts and technical specifications behind Starcloud's planned 88,000-satellite constellation remain undisclosed in the article.
Originally reported by satellitetoday.com
Read the original article →Original headline: Via Satellite: Orbital AI Data Centers Move From Concept to Multi-Company Race — SpaceX, Google, Planet Labs, and Startups Target 2027 Deployments