Orbital data-center cooling is an extension of the existing liquid-cooling ecosystem, not a near-term disruption

Overview

This report examines the potential impact on data-center cooling-equipment suppliers as AI infrastructure extends into space. The core conclusion is that orbital data centers are not entirely new computing infrastructure, but more like modified satellites carrying GPU racks; the first half of their cooling systems resembles terrestrial direct-to-chip liquid cooling, while the primary difference lies in the final heat-rejection method, which shifts from air- or water-side facilities to radiation into deep space.

Core views

Bernstein believes the orbital data-center cooling architecture is technically credible, but represents an evolutionary combination of existing satellite thermal-management and terrestrial liquid-cooling technologies. Facility-level equipment used in terrestrial data centers, such as chillers, cooling towers, dry coolers, and CRAHs, is generally unsuitable for orbital environments; cold plates, manifolds, simplified CDUs, DCIM software, and auxiliary cooling components remain important, while radiator wings and related aerospace thermal-control hardware represent new categories. Because SPCX favors vertical integration, the actual market share available to traditional suppliers may be below the theoretical TAM.

Analysis framework

The report first compares the cooling architectures of terrestrial liquid-cooled data centers and orbital data centers, then maps the required components individually and estimates equipment content on a $/MW basis. The technical analysis draws on thermodynamics, the Stefan-Boltzmann law, satellite-radiator experience, direct-liquid-cooling loops, and the characteristics of waterless ammonia and other aerospace thermal-management materials; the investment analysis combines the projected orbital data-center share of installed capacity by 2031, equipment-supplier product exposure, and potential M&A opportunities.

Methodology notes

• thermal_physicsStefan-Boltzmann law

  radiative heat rejection

  Because air convection cannot be relied upon in a vacuum, orbital data centers must use radiator wings to reject heat into deep space through infrared radiation; the report estimates heat-rejection capacity per unit area based on surface temperature, emissivity, and area.

• industry_mappingcomponent content per MW

  equipment content per MW

  The report breaks down terrestrial and orbital liquid-cooling architectures by component, including chillers, CRAHs, RDHx, CDUs, cold plates, manifolds, DCIM software, and redundancy, to compare supplier revenue opportunities across different designs.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Vertiv (VRT)

  Supplier of equipment related to data-center liquid cooling, CDUs, cold plates, manifolds, and DCIM software

  Strengths

  Has relatively high equipment-content exposure in both terrestrial and orbital liquid-cooling component mappings; estimated orbital content is approximately $0.5–0.8M/MW.

  Weaknesses

  Traditional facility-level heat-rejection equipment is reduced in orbital designs, and SPCX’s vertical integration could compress the share available to external suppliers.

  Comparison

  Terrestrial liquid-cooling content is approximately $1.4–1.8M/MW, higher than in orbital designs.

  Risks

  Slow commercialization of the orbital market, unclear radiator-wing value, and potentially minimal tail-end service revenue.

• nVent (NVT)

  Supplier of liquid-cooling manifolds, CDUs, and other components

  Strengths

  CDUs and manifolds remain necessary in orbital architectures.

  Weaknesses

  Estimated orbital content is below terrestrial content, at approximately $0.2–0.4M/MW.

  Comparison

  Terrestrial designs are approximately $0.5–0.7M/MW, while orbital designs are more streamlined.

  Risks

  Equipment simplification and vertical integration could limit the addressable market.

• Trane Technologies (TT), Carrier (CARR), Johnson Controls (JCI)

  Companies related to traditional HVAC and heat-rejection equipment

  Strengths

  Terrestrial AI data-center cooling demand remains the larger medium-term market.

  Weaknesses

  Traditional equipment such as chillers, cooling towers, dry coolers, and CRAHs is largely unsuitable for orbital architectures.

  Comparison

  Orbital designs have substantially lower content for these traditional facility-level equipment categories than terrestrial liquid-cooled designs.

  Risks

  If orbital data centers outperform expectations over the long term, exposure to traditional facility-level equipment could be insufficient.

• Eaton (ETN), Schneider, Legrand

  Companies related to electrical equipment, power distribution, and data-center infrastructure

  Strengths

  AI data-center power and infrastructure demand remains a long-term support; some liquid-cooling or data-center systems offer potential exposure.

  Weaknesses

  Direct component opportunities in orbital cooling architectures are limited and depend on the systems-integration pathway.

  Comparison

  In the report’s mapping, Schneider has relatively high exposure across CDUs, cold plates, manifolds, and DCIM; Legrand’s orbital cooling content is lower or unclear.

  Risks

  If space infrastructure is led by vertically integrated operators, opportunities for external equipment suppliers may fall below expectations.

• SpaceX (SPCX)

  Potential driver and vertically integrated operator of orbital data centers

  Strengths

  At the core of the development of the space economy and orbital AI infrastructure.

  Weaknesses

  If it chooses to develop thermal-management systems internally, the share available to external suppliers will be reduced.

  Comparison

  The report identifies SPCX as a key operator for orbital data-center deployment and assumes a one-year delay in orbital data-center launches in its valuation.

  Risks

  Launch cadence, technical transparency, service and maintenance feasibility, and commercial demand all remain uncertain.

Key data

• Orbital data-center cooling contentapproximately $0.6–0.8M/MWExcluding radiator wings.
• Terrestrial liquid-cooled data-center cooling contentapproximately $1.3–1.8M/MWIncluding chillers, dry coolers/cooling towers, CDUs, cold plates, manifolds, and other components.
• Estimated orbital data-center capacity in 2031approximately 27GWApproximately 11% of global installed data-center capacity.
• Radiator area for a 100kW orbital computing platformapproximately 158 square metersEstimated based on a two-sided radiator, an operating temperature of approximately 20°C, and net heat-rejection capacity.
• Solar-array area for a 100kW orbital computing platformapproximately 300–500 square metersThe report notes that more than 95% of spacecraft surface area may be used for heat rejection and power generation rather than the computing hardware itself.

Impact & implications

For traditional cooling-equipment suppliers, orbital data centers are not a major negative factor in the short to medium term, because terrestrial data centers will continue to account for the vast majority of installed capacity and equipment demand. Over the long term, if orbital computing scales, radiator wings, aerospace thermal control, lightweight liquid-cooling loops, and deployable structures could create new supply chains, while established equipment suppliers may gain exposure through partnerships or acquisitions of emerging specialists.

Risks

• Commercialization and launch-deployment timelines for orbital data centers may be slower than expected.
• The actual costs of radiator wings, aerospace thermal-control hardware, and solar arrays remain unclear, which could change the $/MW estimates.
• Operators such as SPCX favor vertical integration, potentially reducing external revenue opportunities for traditional equipment suppliers.
• Orbital data centers are difficult to maintain, and service and repair revenue may be close to zero.
• If the long-term share of orbital capacity is materially higher than expected, terrestrial cooling-infrastructure demand could face greater substitution risk.

What to watch

• Launch schedules and capacity targets for orbital data centers disclosed by SPCX and other operators.
• The supplier landscape for radiator wings, deployable structures, waterless ammonia loops, and aerospace thermal-control systems.
• Acquisitions or partnerships by traditional cooling-equipment suppliers involving orbital thermal-management startups.
• Liquid-cooling penetration and changes in equipment content per MW at terrestrial AI data centers.
• Whether orbital data centers can achieve stable thermal management for multi-megawatt AI computing workloads.