SpaceX surged 19% on Friday in its Nasdaq debut following the world’s largest IPO, closing near $161 after opening at $150 and valuing the company north of $2 trillion.

J.P. Morgan + SpaceX= Largest IPO

Congratulations to the @spaceX team on this milestone, we were proud to serve as a lead bookrunner on the transaction. pic.twitter.com/axxob266QP

— J.P. Morgan (@jpmorgan) June 12, 2026

Investor excitement over the potential commercialization of the Starship mega-rocket is certaintly a major driver, but also markets are beginning to view SpaceX as one of the most pivotal players in the emerging orbital data-center race, where launch dominance, Starlink infrastructure, satellite manufacturing scale, and plunging access-to-orbit costs could position Elon Musk’s rocket company at the center of the next frontier in AI compute.

Nearly six months ago, we read the tea leaves and told readers how to position ahead of the SpaceX IPO and the coming space-and-data center buildout race in low Earth orbit. That thesis is moving from speculative to investable, after SpaceX’s public-market debut yesterday and Starship commercialization story nears (read report).

Starship is a very big rocket https://t.co/0RyGe3CPzS

— Elon Musk (@elonmusk) June 13, 2026

A continuation of the space-based data center theme and how to profit comes from Barclays analyst Brendan Lynch in a new report titled “Ashburn, then Abilene, then space.”

Lynch sees the story of space-based data centers gaining ground as territorial deployment woes materialize amid intensifying constraints on power, land, and grid.

This year alone, hyperscalers plan $800 billion in capex to build out data centers. There is growing resistance to the buildout, which has already derailed nearly half of the nation’s planned 16-gigawatt capacity, with only 5 gigawatts currently under construction.

The good news for terrestrial-based data centers is that Lynch and his team don’t see orbital data centers as a likely threat over the next decade, citing launch costs, radiation-resistant hardware needs, thermal-management limits, bandwidth constraints, and regulatory uncertainty.

The big attraction in space is unlimited solar power and no permitting. Orbital data centers could use near-continuous solar energy without relying on local utilities, grid interconnection waits, land availability, zoning approvals, or water-intensive cooling systems. Lynch noted that solar panels in orbit can generate up to eight times more power than terrestrial solar panels because of constant sunlight and the absence of atmospheric interference.

However, the analyst noted that the economics of orbital data centers remain a major roadblock. He estimated that orbital data centers cost roughly $51 billion per gigawatt to build and operate over five years, compared with about $16 billion per gigawatt for terrestrial data centers.

Lynch said, “However, there is still a long way to go before the economics and engineering make orbital data centers feasible at scale. Currently, orbital capacity is ~3x more expensive per MW than terrestrial, primarily due to high launch costs. Additionally, further progress must be made on engineering challenges, such as radiation-resistant hardware, thermal management, and connectivity.”

Google estimates launch costs would need to fall below $200 per kilogram by 2035 for its orbital-compute vision to work, while SpaceX’s Falcon Heavy is currently around $1,500 per kilogram.

Given these constraints, Lynch does not see orbital data centers as a “threat to our coverage with data center exposure (DLR, EQIX, IRM, AMT) in the next 10 years.”

Now he added, “Beyond 10 years, it is harder to handicap the impact, but if space-based DCs come to fruition, it will likely be complementary to traditional deployments.”

“That said, as technology advances and costs come down, we anticipate orbital capacity will gain momentum,” the analyst noted.

The moment when launch costs plummet will likely hinge on the Starship commercialization timeline, which could see full-scale commercialization around 2027-28 and, really, at the end of the decade.

Starship is still transitioning from test vehicle to commercial platform. The first monetization wave is likely internal SpaceX demand, mainly Starlink deployment, larger satellites, orbital AI-compute demos, and NASA-linked lunar spacecraft.

Reuters reported SpaceX is aiming to begin orbital AI-computing demonstration missions by late 2027, a key validation point for the orbital data center.

Lynch added more color about the orbital data centers:

How data centers in space operate

Power

• Most orbital data center plans involve many satellites in low earth orbit operating collectively to form the “data center” in space, similar to how terrestrial data centers are comprised of many server racks. Clusters of satellites are often called constellations.
• Large solar panels supply near-continuous power. Satellites can be placed in sun-synchronous orbits (e.g., “terminator” orbits) to maximize solar exposure. Batteries are also required to store energy for eclipse periods when satellites pass into earth’s shadow.

Communication network

• Optical laser links connect satellites so that they can share data. They are a high-speed method of transmitting data through laser beams. This is the same technology that some satellite operators use to provide broadband capacity on earth.
• Satellites transmit data to ground stations, which serve as the “middleman” between the data center and users. Constellations will likely require thousands of ground stations because low earth orbit satellites only pass in range of each ground station for a few minutes per orbit. Ground stations have large antennas to communicate with satellites either through radio waves or optical laser links. Radio waves provide reliable, regulated, lower-bandwidth connectivity, while optical links enable high-capacity, high-efficiency data transfer but require precise alignment and are sensitive to atmospheric conditions. Ground stations will also have fiber optic cables to connect with users.

Compute and cooling

• Advanced computing in space requires radiation-tolerant or radiation-hardened chips. Several semiconductor companies, including NVDA (covered by Tom O’Malley), are exploring specialized space-based computing infrastructure.
• Liquid cooling removes heat from chips, and then radiators dissipate heat as infrared radiation into deep space. Traditional air cooling methods don’t work  due to the lack of atmosphere. Compute density per satellite is primarily limited by the rate at which heat can be radiated into space.

Operations Satellites

• Satellites are launched into space via rockets designed for heavy loads, similar to how traditional satellites are launched, but conceivably at much larger scale.
• Physical maintenance will likely be limited, but software updates are possible. Satellites will likely have redundant components and built-in work-arounds in case of hardware failure.
• Most business models assume no servicing or upgrades. Instead, satellites that reach the end of their operating life will be replaced by new ones carrying the latest technology. Most satellites are expected to have a 5-year useful life. At the end of life, satellites are typically de-orbited into the atmosphere to burn up.

Why data centers in space are attractive

Power

• Space provides less constrained access to solar power with fewer bottlenecks to scale vs. terrestrial power grids. Developers are not reliant on utility companies to provide power infrastructure.
• Power is generated and consumed in the same location, avoiding transmission losses and grid interconnection constraints.
• Solar panels in orbit can generate up to 8x higher output due to constant sun exposure and lack of atmospheric interference (molecules in the atmosphere absorb, scatter, and reflect sunlight, reducing the solar energy that reaches terrestrial solar panels). Solar power in space is also more stable than earth because there are no clouds or weather issues.

Land

• Suitable land sites with sufficient power are increasingly scarce in key data center markets globally. Space offers a solution to land constraints.
• Orbital data centers avoid many challenges faced by terrestrial development, including community opposition, environmental remedies, zoning restrictions, etc.

Resilience

• Infrastructure in space is less exposed to disruption from natural disasters, grid failures, and geopolitical events.
• Constellations of satellites offer high resiliency because workloads can be shifted between satellites if one goes down.

Design

• The modular design enables a more efficient capacity build out, where infrastructure is scaled via incremental satellite launches rather than large upfront development projects. Over time, this could reduce capital intensity and development risk.
• Water usage is one of the most common critiques of terrestrial data centers, particularly as AI increases compute density and cooling needs. Orbital data centers do not require evaporative water cooling

Challenges to near-term deployment

Physical

• Satellites will require very large solar panels to generate sufficient power for AI workloads. Satellites that support compute functions (instead of communications) might need to be ~10x larger to achieve attractive economies of scale.
• Space requires specialized IT hardware due to radiation which can corrupt data unpredictably and degrade equipment. Traditional space hardware uses radiation hardened chips that are more than 100x less powerful than chips in terrestrial data centers and very expensive.
• Thermal management limits compute density per satellite. There is no medium for heat transfer in space (i.e. no air), so satellites require a combination of liquid cooling to remove heat from the chips and radiators to remove heat from the satellite. Heat is emitted into deep space via infrared radiation. The radiators requires a lot of surface area in addition to the large solar panels because radiative heat transfer is relatively inefficient vs. air cooling.
• Orbital data centers face networking and bandwidth limitations. Inter-satellite connectivity (generally via optical laser links) requires complex, precise alignment. Space-to-earth communication via radio waves (most common currently) is heavily regulated and has relatively low bandwidth. The International Telecommunication Union (ITU) coordinates global spectrum allocation, and operators require authorization in each jurisdiction where they transmit signals to/from the ground. Optical laser links (emerging technology) are higher bandwidth and higher efficiency but face atmospheric interference due to clouds and weather and require precise alignment. Additionally, space-to-earth connectivity requires sufficient ground stations to receive/transmit data.
• Orbital systems have high failure rates vs. terrestrial infrastructure. When equipment fails in orbital data centers, it can’t be replaced. As a result, orbital data centers must be highly redundant and have failover measures. If the satellite fails, it must be entirely replaced.
• Launch capacity is the primary constraint on scaling infrastructure due to the limited frequency of rockets launches. Size and weight are pertinent considerations for satellite design due to constraints of the rocket. Many orbital data center business plans are dependent on improvements to the launch process. In 2025, there were 330 launches globally. Each rocket can carry about 40-100 traditional satellites. However, orbital data centers could eventually exceed the size of the largest rockets that are available, highlighting the need for improved launch capabilities.

Regulatory

• A primary concern is overcrowding in earth’s orbit, which increases the likelihood of collisions and long-term debris accumulation. The FCC requires that low earth orbit satellites are de-orbitted within five years of end-of-life, and companies must file orbital debris mitigation plans with regulators. There are currently ~16,000 satellites orbiting earth, but several companies have filed plans with the FCC to collectively increase this by 10x with build-outs in the late 2020s and 2030s.
• There will likely be future challenges due to regulatory and jurisdiction uncertainty given the lack of standards for orbital data centers. For example, spectrum allocation and licensing is currently handled by individual countries. Broader AI regulations and data sovereignty requirements will likely also be factors.

Economic

• Orbital data centers are estimated to cost up to ~$50m/MW, more than triple the cost of terrestrial data centers, at present.
• The biggest financial challenge is launch costs. Google estimates that launch costs would need to fall below $200/kg by 2035 for its vision to be economically viable. SpaceX’s current launch vehicle, Falcon Heavy, is the cheapest available at $1,500/kg.
• In addition to the higher build cost, the useful life of orbital data centers is only ~5 years due to limited maintenance and upgrade capabilities and the harsh environment in space (e.g. radiation, extreme temperatures). This compares to decades of useful life for terrestrial data centers which can be maintained and upgraded more easily.’

And now to the part readers care about most: how to profit from the buildout.

Axiom Space (private, not covered)

• The company has been testing cloud computing capabilities on the International Space Station (ISS) since 2022 and launched its first two orbital data center nodes in January 2026. Its nodes are modular units located on the space station.
• Axiom is also building a commercial space station which it plans to launch ahead of the ISS’s retirement in 2030.

Blue Origin (private, not covered)

• The company announced Project Sunrise with a target of deploying up to 51,600 satellites for AI workloads. It filed plans with the FCC in March 2026, but faces an objection from NASA regarding the proposed orbit altitude (which overlaps with critical human spaceflight paths) and risk of space debris.
• The company also has plans to launch a 5,000 satellite constellation for global high-speed communications infrastructure, called TerraWave. It aims to begin deploying TerraWave satellites in late 2027. TerraWave satellites are designed for networking while Project Sunrise satellites are designed to enable high-density compute.

Cowboy Space (private, not covered)

• The company filed plans with the FCC to deploy 20,000 orbital data center units in a constellation called Stampede in May 2026. Each unit would repurpose the the upper stage of the rocket as a high-density compute platform. Cowboy Space aims to launch its first rockets in 2028.
• The company is also working on a separate constellation that would send solar power back to earth.

Planet Labs (public, not covered)

• The company partnered with Google (covered by Ross Sandler) for project Suncatcher which has a demonstration mission planned for early 2027 to test Google’s TPUs (specialized AI chips designed to accelerate machine learning and inferencing workloads) in space.
• Planet Labs already operates 600+ satellites that form an imaging constellation for geospatial intelligence.

SpaceX (public, not covered)

• The company filed plans with the FCC to launch a million data center satellites for ~100GW of compute capacity in January 2026.
• SpaceX currently operates ~10,00 Starlink satellites and controls ~65% of active satellites globally. Starlink satellites primarily enable communication vs. data center satellites which are designed for high-density compute.

Starcloud (private, not covered)

• The company deployed a ~1kW satellite with a single GPU in November 2025 as proof-of-concept. It plans to launch its next-gen satellite which is 10kW in 2027 and then launch a ~200kW satellite in 2028.
• Its ultimate goal is to deploy 88,000 satellites totaling ~20GW of compute primarily for inference workloads, reaching ~5GW by 2035. Starcloud filed plans with the FCC in March 2026.

Professional subscribers can read much more on SpaceX and the space economy at our new Marketdesk.ai portal.