What is a space-based data center? A space-based data center is a computing facility deployed in Earth orbit that uses solar power, vacuum cooling, and satellite networking to process AI workloads outside the constraints of terrestrial real estate, power grids, and water supplies. In March and April 2026, two major developments pushed this concept from science fiction to funded reality: Starcloud closed a $170 million Series A at a $1.1 billion valuation, and Google's Project Suncatcher advanced plans for an 81-satellite compute constellation. For CRE investors tracking AI infrastructure demand, these developments introduce a new variable into data center investment models that could reshape long-term supply dynamics. For a broader view of how AI is transforming commercial real estate, see our complete guide on AI commercial real estate.
Key Takeaways
- Starcloud became the fastest Y Combinator unicorn in history after raising $170 million at a $1.1 billion valuation for orbital data centers.
- Google's Project Suncatcher will launch two demonstration satellites by early 2027 to test TPU-equipped compute clusters in orbit.
- Space data centers could reach cost parity with terrestrial facilities by 2035 if SpaceX Starship reduces launch costs to $200 per kilogram.
- Near-term CRE impact is minimal, but long-term portfolio models should account for orbital compute as a demand offset in the 2030s.
- Terrestrial data center demand remains robust through 2030, with hyperscalers committing $650 billion in capex for 2026 alone.
The News: Two Billion-Dollar Bets on Orbital Compute
On March 30, 2026, Starcloud announced a $170 million Series A led by Benchmark and EQT Ventures, valuing the company at $1.1 billion. Founded in January 2024, Starcloud became the fastest unicorn in Y Combinator history just 17 months after completing the program. The company launched its first satellite, Starcloud-1, in November 2025 with an NVIDIA H100 GPU on board and became the first company to train a large language model in orbit, running Google DeepMind's Gemma and Andrej Karpathy's nanoGPT from space.
Starcloud-2 is scheduled to launch in October 2026 with NVIDIA's Blackwell B200 chip, 100 times the power generation of its predecessor, and a Crusoe Cloud server blade. The company has also filed with the FCC for a constellation of up to 88,000 satellites, signaling ambitions far beyond a single test vehicle.
Meanwhile, Google's Project Suncatcher is advancing in partnership with satellite imaging company Planet Labs. As detailed by Data Center Dynamics, the project envisions clusters of 81 satellites equipped with Google TPU accelerator chips, flying in tight formation in sun-synchronous low Earth orbit to maximize solar exposure. Two demonstration spacecraft are targeted for launch by early 2027. Google CEO Sundar Pichai has framed the effort as a long-term solution to AI's energy problem, noting that the sun provides 100 trillion times more energy than what humanity produces on Earth today.
How Space Data Centers Solve Terrestrial Constraints
The logic behind orbital compute directly addresses the three biggest bottlenecks facing CRE data center development today: power, cooling, and land. As our coverage of US data center build delays documented, half of current projects face delays from Chinese equipment shortages and grid interconnection queues of 3 to 5 years.
- Power: In the right orbit, a solar panel produces up to 8 times more energy than on Earth and operates nearly continuously, eliminating dependence on grid power and fossil fuels.
- Cooling: The vacuum of space serves as an infinite heat sink. Satellites radiate waste heat via infrared, eliminating the water-intensive cooling systems that consume millions of gallons per year at terrestrial facilities.
- Land: Orbit offers unlimited expansion without zoning restrictions, community opposition, or competing land uses. Space eliminates the real estate bottleneck entirely.
Google's radiation testing showed its Trillium-generation TPUs survived simulated five-year exposure to low Earth orbit proton and cosmic ray bombardment, though High Bandwidth Memory subsystems showed some sensitivity. These results suggest that current AI chips can function in space with appropriate shielding.
What This Means for CRE Data Center Investors
CRE investors should understand the timeline before adjusting portfolio strategy. Space-based data centers are not an immediate threat to terrestrial demand. Current launch costs of $1,500 to $2,900 per kilogram make orbital compute roughly three times more expensive per watt than ground-based facilities. Google's own analysis estimates cost competitiveness could arrive around 2035, contingent on SpaceX's Starship achieving commercial launch costs near $200 per kilogram.
In the near term through 2030, the picture for terrestrial data centers remains bullish. Hyperscalers including Microsoft, AWS, Meta, Alphabet, Oracle, and CoreWeave have committed over $650 billion in combined data center capex for 2026. Regional grid operators like PJM and ERCOT report that data centers account for over 70% of new large-load interconnection requests. As our analysis of Digital Realty's $3.25 billion fund highlighted, institutional capital is flooding into terrestrial data center development.
However, CRE investors with 10 to 15 year hold periods should begin modeling scenarios where orbital compute absorbs a portion of future AI workload growth. If Starcloud's planned 5-gigawatt orbital constellation and Google's Suncatcher concept mature on schedule, they could serve as supplemental capacity that slows the pace of new terrestrial campus starts in the 2030s rather than displacing existing facilities.
The Competitive Landscape in Space
Starcloud and Google are not alone. Multiple companies are pursuing orbital data center strategies, creating a competitive landscape that mirrors the early days of terrestrial cloud computing:
- Starcloud: $1.1 billion valuation backed by Benchmark and EQT Ventures. FCC filing for 88,000 satellites. Partnership with Crusoe for GPU cloud from orbit by early 2027.
- Google (Project Suncatcher): Partnership with Planet Labs. Two demonstration satellites launching 2027. Research-stage project exploring TPU clusters in sun-synchronous orbit.
- Axiom Space: Exploring compute infrastructure integration with its commercial space station modules.
- NTT and Ramon.Space: Developing radiation-hardened computing platforms for space-based enterprise workloads.
- Chinese entrants: ADA Space and Beijing Astro-Future are building space compute architectures aligned with China's domestic AI infrastructure strategy.
For personalized guidance on evaluating how emerging data center technologies affect your CRE portfolio, connect with The AI Consulting Network.
How CRE Investors Should Respond
The prudent approach for CRE investors is to monitor the space data center sector without overreacting. Consider these practical steps:
- Continue investing in terrestrial data centers: Demand fundamentals through 2030 remain strong. Cap rates of 4.5% to 6.0% on stabilized hyperscale assets reflect real, sustained demand from hyperscaler AI buildouts. Cap rate is calculated as NOI divided by purchase price and does not include debt service.
- Stress test long-hold assumptions: For acquisitions or developments with hold periods extending past 2032, model a scenario where 10% to 20% of incremental AI compute demand shifts to orbital infrastructure.
- Watch launch cost benchmarks: SpaceX Starship's commercial pricing trajectory is the single most important variable. If costs drop below $500 per kilogram, space data centers become economically viable at scale.
- Track tenant diversification: Data center assets leased to tenants actively investing in space compute, such as Google and AWS, may see future demand shifts. Diversified tenant rosters reduce this exposure.
CRE investors looking for hands-on AI implementation support can reach out to Avi Hacker, J.D. at The AI Consulting Network to discuss how these emerging trends fit into your investment strategy.
Frequently Asked Questions
Q: Will space data centers replace terrestrial data centers?
A: Not in the near term. Space-based data centers are currently about 3 times more expensive per watt than terrestrial facilities. Industry analysts project cost parity could arrive around 2035, and even then, orbital compute is more likely to supplement terrestrial capacity than replace it. Existing data centers with long-term leases and established power infrastructure will retain their value.
Q: What is Google's Project Suncatcher?
A: Project Suncatcher is a Google research initiative exploring AI computing in space using clusters of 81 satellites equipped with TPU chips. The satellites would fly in sun-synchronous low Earth orbit to capture near-continuous solar power. Google is partnering with Planet Labs to launch two demonstration spacecraft by early 2027.
Q: How does Starcloud's valuation compare to terrestrial data center companies?
A: Starcloud reached a $1.1 billion valuation on $200 million in total funding, making it the fastest Y Combinator unicorn in history. For context, Digital Realty has a market cap exceeding $40 billion and Equinix exceeds $70 billion. Starcloud represents early-stage venture capital betting on a disruptive technology, not a comparable operating business today.
Q: When should CRE investors start adjusting portfolios for space data center risk?
A: Investors with hold periods under 7 years face minimal exposure to space data center disruption. For longer holds extending past 2032, it is prudent to begin stress testing assumptions about continued terrestrial demand growth. The key trigger point is SpaceX Starship achieving commercial launch costs below $500 per kilogram, which would make orbital compute economically viable at scale.