SpaceX’s push to build artificial intelligence data centers in orbit is aiming to sidestep rising costs and local opposition tied to power-hungry terrestrial facilities. But analysts say the economics and operational challenges of maintaining compute hardware in space could keep Earth-based infrastructure in the lead for years—an outcome that would likely support electric utilities.
Key takeaways
-
SpaceX is pursuing orbital AI data centers as a long-term approach to reduce cooling and energy constraints.
-
The concept faces major hurdles, including equipment limits in extreme cold and the high cost of launching and servicing hardware.
-
While orbital sites could be cheaper in theory, one estimate suggests orbital facilities could cost more than three times as much as comparable terrestrial ones near term.
-
If orbital deployments take time, terrestrial power demand for AI data centers could continue to rise.
-
Electric utility infrastructure may be positioned to benefit as grid operators prepare for expanding data center loads.
What SpaceX is proposing
AI data center buildouts are expanding as demand for training and deploying large language models grows. The plan under discussion is to move part of that infrastructure off the ground and into orbit. By locating compute in space, proponents argue the environment could reduce some of the typical operational burden faced by terrestrial sites.
In SpaceX’s outline, extremely cold conditions in orbit—reported as around -455 degrees Fahrenheit—would reduce the need for conventional cooling systems. The company also points to solar power as a potential advantage, arguing that sunlight is more concentrated in space and could make solar panels more efficient. For connectivity, SpaceX would use and expand its Starlink satellite network to handle data transfer between users and the orbital compute site.
Why orbital data centers are still a hard sell
The biggest issue may be that “cold” is not automatically “easy” for electronics. Most electrical equipment generally cannot operate at temperatures below roughly -340 degrees Fahrenheit, according to the report cited in the article. While workarounds exist—such as using costlier materials and designs—they can raise the bill of materials and engineering complexity.
Economics are also a challenge. Wood Mackenzie estimates that a 1-gigawatt orbital data center would cost about $170 billion, described as more than three times the cost of an equivalent terrestrial facility. That gap would matter most during early deployment phases, when launch costs and specialized hardware requirements remain elevated.
Operational reliability presents another problem: in terrestrial facilities, technicians can often repair a failed component on-site. In space, component-level fixes are constrained by the cost and risk of launching replacement parts and performing service missions. The article argues that even if launch costs decline over time, it is unlikely that servicing individual failures would become cheap enough to match the flexibility of land-based operations.
How fast would orbital compute need to become viable?
The report estimates that the cost of orbital data centers would need to drop by about 70% to become competitive. The article suggests that such improvement could occur over a longer horizon—potentially by around 2040—if launch costs continue to fall at an accelerated rate.
Until that cost compression happens, the market would still rely heavily on terrestrial data centers. That implication ties directly to the near-term demand profile for electricity and grid capacity, which investors increasingly track as AI infrastructure spreads across regions.
Terrestrial power demand could be the bridge to the next phase
Power consumption is central to why AI data centers have faced both cost pressures and community pushback. The International Energy Agency has been cited in the article, saying a ChatGPT query uses about 10 times as much electricity as a Google search. Beyond raw energy use, data centers also generate substantial heat and typically require large cooling systems.
The article highlights that cooling and water needs can vary by approach. It notes that air-cooled systems typically require additional electricity to run fans and related equipment, while liquid-cooled systems can require significant water resources. These factors feed into local concerns about environmental impacts and strain on electricity and water supplies.
As a result, even if orbital concepts eventually reduce some terrestrial constraints, the timeline for feasibility may keep grid expansion urgent in the meantime. The article describes how residents in some areas have used zoning and political channels to prevent proposed AI data centers.
Which companies could benefit if terrestrial growth continues
The article argues that electric utility companies may be positioned to gain as AI load expands on the grid. One highlighted beneficiary is American Electric Power, which the article describes as operating the largest electricity transmission network in the U.S. and serving regions historically associated with large-scale energy development, including Texas, Oklahoma, Louisiana, and Appalachia. The report also mentions that the company has near-monopoly exposure to 765-kilovolt transmission infrastructure.
It further points to American Electric Power’s introduction of a “Data Center Tariff” in Ohio, where data center developers reportedly entered binding contracts for 5.6 gigawatts of data center load. The article frames this as a mechanism that can insulate the utility from financial repercussions if a project’s contracted capacity is not fully utilized or if a project is canceled.
For investors, the implication is straightforward: as long as most AI computing remains on land, grid operators with transmission reach and contractual structures that manage demand risk may capture more of the value from the buildout—even if orbital deployments remain a longer-dated option.
Investors watching this theme may focus next on grid expansion plans, utility permitting and rate structures for data center loads, and any further updates on orbital hardware development and launch economics. On the broader macro side, the pace of AI-related electricity demand, utility capital expenditure guidance, and upcoming energy-policy or regulatory decisions could shape how quickly the market re-prices utilities versus alternative infrastructure concepts.
