Rosa Del Mar

Daily Brief

Issue 71 2026-03-12

Orbital Compute: Thermal Physics, Debris/O&M, And Weak Near-Term Economics

Issue 71 Edition 2026-03-12 10 min read
General
Sources: 1 • Confidence: Medium • Updated: 2026-03-14 12:25

Key takeaways

  • Space-based (orbital) data centers are unlikely to be economically cheapest within three to four years and are framed as an 'endgame' pathway unlikely before 2030.
  • Off-grid data centers must self-provide grid 'shock absorber' functions such as inertia, fault response, and blackstart, which is complex and expensive at gigawatt scale.
  • The shares among grid-connected, off-grid, edge, and off-world compute depend heavily on the absolute size of total compute demand in 10 years (e.g., hundreds of gigawatts versus multiple terawatts).
  • Land cost savings from edge siting are unlikely to materially change total data center economics because land is a small portion of fully loaded cost relative to GPUs, buildings, and labor.
  • Large clustered data centers can raise power-quality and broader grid-impact concerns that may affect whether regulators and utilities are willing to serve them at scale.

Sections

Orbital Compute: Thermal Physics, Debris/O&M, And Weak Near-Term Economics

  • Space-based (orbital) data centers are unlikely to be economically cheapest within three to four years and are framed as an 'endgame' pathway unlikely before 2030.
  • Radiative heat rejection scales with the fourth power of temperature, so operating chips hotter and denser can improve heat rejection for space-based systems.
  • In hyperscale data centers, engineers can replace failing CPUs/GPUs in near real time, whereas failed components in space are effectively stuck without future robotic servicing, creating economic drag.
  • It is contested whether space is meaningfully more scalable than terrestrial off-grid buildout given the comparison between Starship capacity constraints and the possibility of expanding Earth-side power manufacturing.
  • Maximizing terrestrial solar-plus-storage, geothermal, and new nuclear is argued to be less extreme than relying on very high-frequency Starship operations (e.g., multiple launches per day) to scale orbital compute.
  • Rejecting heat in space is intrinsically difficult; the ISS rejects under about 100 kW using radiator area on the order of a soccer field.

Off-Grid Terrestrial Compute: Large Theoretical Potential, Reliability And Equipment As Rate Limiters

  • Off-grid data centers must self-provide grid 'shock absorber' functions such as inertia, fault response, and blackstart, which is complex and expensive at gigawatt scale.
  • Achieving five-nines-like reliability off-grid generally requires overbuilding both generation capacity and storage, raising project cost and complicating financing for very large assets.
  • If data centers go off-grid, dominant scaling constraints shift to power-generation and electrical equipment supply chains such as turbines, solar, batteries, transformers, and switchgear.
  • The participants express a view that off-grid compute growth is more plausible than edge computing representing a large share of total compute.
  • A study co-authored by Stripe, Paces, and Scale Microgrids identified over a terawatt of off-grid renewable-plus-storage opportunity in the American Southwest, including approximately 50% solar plus batteries at cost parity to all-gas and up to approximately 80–90% solar without major cost increase.
  • For the next decade, land availability is not the binding constraint for compute growth; power generation and delivery infrastructure is the limiter for off-grid terrestrial data centers.

Constraint-First Framing For Compute Scaling

  • The shares among grid-connected, off-grid, edge, and off-world compute depend heavily on the absolute size of total compute demand in 10 years (e.g., hundreds of gigawatts versus multiple terawatts).
  • Data-center scaling can be analyzed as two separate questions: how much compute demand grows and how the energy supply to serve it is delivered.
  • For analysis, the speakers assume compute demand continues scaling for the next 5–10 years and there is no major energy-efficiency breakthrough that resets the paradigm.
  • Each proposed data-center configuration has one or two core constraints and distinct strengths that determine where and how it can scale.
  • Discussion about data-center siting and configuration is often dominated by maximalist advocacy for a single pathway rather than explicit tradeoff analysis across pathways.
  • The episode’s approach is to start with incumbent grid-connected hyperscale data centers and then compare constraints across proposed alternatives.

Edge Compute: Narrative Challenges, Operational Scaling Uncertainty, And A Distinct On-Device Pathway

  • Land cost savings from edge siting are unlikely to materially change total data center economics because land is a small portion of fully loaded cost relative to GPUs, buildings, and labor.
  • Latency is unlikely to be the primary driver for most AI inference workloads, weakening the case that edge computing must dominate for latency reasons.
  • The participants express a view that off-grid compute growth is more plausible than edge computing representing a large share of total compute.
  • On-device inference (e.g., vehicles deciding locally, phones running smaller models) is a distinct 'edge' pathway that differs from deploying smaller data centers.
  • A potentially viable 'edge data center' form factor is a smaller grid-connected facility in the ~15–30 MW range, rather than very small (kW to a few MW) deployments.
  • A key unresolved question for edge compute is whether it can deliver capacity faster at comparable aggregate scale given the need to secure and develop many more individual sites.

Grid-Connected Hyperscale Bottlenecks And Non-Technical Constraints

  • Large clustered data centers can raise power-quality and broader grid-impact concerns that may affect whether regulators and utilities are willing to serve them at scale.
  • For grid-connected hyperscale data centers, a core scaling constraint is the lead time to add transmission deliverability to serve gigawatt-scale loads, often on the order of five to seven years in many markets.
  • Community and political pushback has emerged as a material siting constraint for data centers, including blanket bans and cancellations of previously announced developments.
  • In the United States, building new transmission lines—especially interregional or cross-state—has become so difficult that project timelines can be effectively unbounded compared with generation or substation upgrades.

Watchlist

  • Large clustered data centers can raise power-quality and broader grid-impact concerns that may affect whether regulators and utilities are willing to serve them at scale.

Unknowns

  • What are the actual time-to-energize distributions for large grid-connected data centers across major markets, and how often are five-to-seven-year timelines observed versus outliers?
  • How frequently do regulators/utilities impose new requirements or deny service for data centers due to power-quality and broader grid-impact concerns, and what technical thresholds trigger these actions?
  • How prevalent and durable is community/political pushback (e.g., bans, moratoria, cancellations), and what project characteristics most correlate with rejection versus acceptance?
  • What portion of behind-the-meter generation associated with data centers is truly off-grid versus grid-connected peak shaving or bridge supply, and how is this reported in filings and contracts?
  • What are independently verifiable uptime metrics for off-grid or islanded data center pilots, and how do reliability and cost change with scale (tens of MW to GW)?

Investor overlay

Read-throughs

  • Near term capacity growth likely concentrates in grid connected hyperscale, with constraints shifting value toward transmission deliverability, interconnection execution, and power quality solutions rather than novel compute locations.
  • Off grid gigawatt scale data centers may face higher cost and complexity due to needing grid stability functions, implying upside sensitivity for vendors enabling inertia, fault response, and blackstart if adoption rises.
  • Orbital data centers appear unlikely to be cost leading before 2030, implying limited near term revenue relevance and a higher bar for any company narrative premised on cheap space power.

What would confirm

  • Observed time to energize for large data centers frequently clusters at multi year timelines across major markets, supporting the view that transmission deliverability is the binding constraint.
  • Utilities or regulators add new power quality and grid impact requirements or deny service for large clustered loads, indicating rising friction for rapid on grid scaling.
  • Independent uptime data from off grid or islanded pilots shows acceptable reliability at tens of megawatts and a clear pathway to scale, implying feasibility beyond niche use.

What would kill

  • Time to energize for large sites is routinely short with few multi year cases, weakening the thesis that interconnection and transmission dominate near term scaling.
  • Regulatory and community pushback proves rare and non durable for large data centers, reducing the importance of non technical constraints in siting outcomes.
  • Off grid pilots show poor uptime or sharply rising costs with scale, undermining the case that off grid can expand materially without expensive overbuild.

Sources