28 May 2026

^{Starcloud Co-founders: Ezra Feilden, Philip Johnston, Adi Oltean (image: Starcloud)}

 

Artificial intelligence is pushing the world’s computing infrastructure to the limit on power, cooling and land, and startup Starcloud is betting that the next generation of AI data centers may not be built on Earth at all.

On May 26., 2026, the orbital computing startup announced a contract with SpaceX’s Starlink division to install more than 50 Starlink Mini Laser terminals on over 25 satellites. The first hardware is expected to be in orbit in a year.

The deal gives Starcloud access to the same optical communication technology used in the fast-growing SpaceX Starlink satellite network. More importantly, it suggests that orbital computing is beginning to move from futuristic ideas into the infrastructure strategies surrounding AI.

This news follows Starcloud’s Series A funding round in March 2026, where the company raised $170 million at a $1.1 billion valuation. The company said the round made Starcloud the quickest startup in Y Combinator’s history to become a unicorn after its demo day.

Investors aren’t just funding yet another satellite company. They are backing the idea that Earth itself is becoming a limitation to AI infrastructure.

 

AI’s infrastructure problem is becoming physical

 

The discussion about AI has long been about models, software and semiconductors. But infrastructure is increasingly the limiting factor.

Running and training large AI systems requires massive amounts of electricity, cooling and physical space. New AI data centers are already taxing power grids in portions of the United States and Europe, and permitting for energy and construction projects can take years.

Big tech firms like Microsoft, Amazon and Google compete to lock in future energy supplies for AI expansion. The AI economy has also been characterized by nuclear energy projects, dedicated renewable infrastructure and long-term power agreements.

In that context, Starcloud’s proposal is gaining interest for its complete reframing of the problem.

The company doesn't want to fight for more limited terrestrial infrastructure, but rather wants to put computing systems in low Earth orbit where satellites have near-constant access to solar energy and can dump heat directly into space.

“Artificial intelligence is driving unprecedented demand for computing power, and terrestrial infrastructure is struggling to keep up,” Starcloud said in announcing its funding round in March. The startup contends that orbital systems could circumvent some of those constraints.

 

Why the SpaceX deal matters

 

The partnership with SpaceX centres on optical communication technology.

In the case of traditional satellites, the information is sent back to Earth through ground stations before it can travel through wider networks. Optical intersatellite links allow satellites to communicate directly in orbit with lasers instead of radio frequencies.

That is very important for Starcloud.

The commercial value of orbital data centres is only unlocked when satellites can rapidly transfer large amounts of information to each other, with minimal reliance on terrestrial relay infrastructure. The company says the satellites will be equipped with two Starlink Mini Laser terminals each.

The partnership effectively gives Starcloud access to one of the world's largest operating satellite laser networks.

SpaceX has been deploying optical links in the Starlink constellation for years to provide better internet coverage across the globe and reduce latency. Starcloud is now taking these same networking concepts and applying them to distributed orbital computing infrastructure.

The company says potential uses include analyzing Earth observation, climate modeling, monitoring wildfires and other data-intensive AI workloads.

That's an important shift in the way satellites are positioned.

Previously, satellites gathered data and returned it to Earth for processing. Instead, companies like Starcloud are proposing that more and more data could be processed in orbit.

 

From cloud computing to orbital computing

 

The concept of data centres in space has been discussed for decades in aerospace research and science fiction. But launch costs and hardware limitations have kept the idea mostly theoretical until recently.

Several developments have altered the equation.

The first is the economics of launching. SpaceX and others have built reusable rockets that have significantly lowered the cost of getting hardware into orbit. That has opened up whole new types of commercial space businesses.

The second is the AI itself.

The economics of AI infrastructure are increasingly influenced by access to electricity and cooling. Earlier this year, Goldman Sachs estimated that global data centres’ power demand could spike sharply by the end of the decade, driven largely by AI workloads.

Orbital computing startups believe space offers structural advantages that Earth increasingly can’t.

Outside of eclipse periods, solar energy in orbit is essentially constant. Space also lets heat radiate away without the water-intensive cooling systems used by many terrestrial facilities.

Starcloud says its orbital infrastructure could in the long run operate at lower energy costs than similar systems on Earth.

The company has already tried to validate parts of the concept. Starcloud launched a satellite last year with an NVIDIA H100 graphics processing unit, or GPU, one of the most sought-after AI chips being used today for training and inference workloads.

The company later said it had successfully run AI workloads in orbit, including a version of Google’s Gemini model.

Those demonstrations are limited in scale compared to terrestrial AI infrastructure. But they helped push orbital computing discussions from theory to early deployment.

 

The legal and environmental questions are growing

 

There are serious technical and economic challenges for orbital data centres. Governance questions may be even more difficult.

Under the 1967 Outer Space Treaty, satellites remain the property of their launching state. However, orbital data centres may eventually have to handle information associated with multiple legal systems at once, including regulated personal or commercial information originating from Europe, the United States or Asia.

That presents tricky enforcement questions.

Regulators may have practical limits on their ability to audit, investigate or seize orders that would normally apply to terrestrial facilities if sensitive corporate or personal data is processed on board privately operated orbital infrastructure. Legal scholars have become increasingly skeptical about how existing frameworks like the EU’s General Data Protection Regulation, or GDPR, would work in practice with permanently orbiting infrastructure.

Also, there is growing concern about the environment.

Instead of terrestrial power grids, orbital systems depend on solar energy, but large satellite constellations still need regular rocket launches and hardware replacement cycles. Onboard server hardware is also subject to heavy radiation, which can reduce operating lifetimes and increase satellite turnover.

Scientists have warned that rising satellite activity may add to pollution in the upper atmosphere, the buildup of orbital debris and further strain on low Earth orbit, which is already becoming increasingly crowded.

Concerns have also been raised by astronomers and environmental researchers about the effect of mega-constellations on night skies and ground-based observation systems.

Another strategic concern is the concentration of orbital infrastructure in the hands of a few private companies.

The Outer Space Treaty says outer space is the “province of all mankind”, but the economics of orbital infrastructure is heavily biased towards a small number of well-capitalised corporations with access to launch systems, semiconductor supply chains and large-scale financing.

That dynamic increasingly mirrors broader discussions about AI infrastructure sovereignty on Earth, where computing power is becoming concentrated in the hands of a relatively small number of companies and countries.

 

The scepticism remains substantial

 

Getting hardware into space still costs far more than constructing traditional facilities on the ground. Satellite hardware also has to contend with radiation exposure, maintenance limitations and reliability challenges that terrestrial operators largely avoid.

Another engineering challenge yet to be solved is cooling high-performance AI hardware in space.

Thermal management has recently been highlighted by Jensen Huang, CEO of NVIDIA, as one of the biggest technical roadblocks for orbital computing systems. Although space offers advantages in terms of radiating heat, developing reliable cooling structures for next generation processors in space is still a very difficult task.

There’s also a bigger economic question of which AI workloads actually benefit from being deployed in orbit.

Many enterprise AI applications are tied to the proximity of terrestrial users, cloud systems, and fiber networks. The only reason to send payloads to orbit is if the operational or energy benefits outweigh the latency and launch costs required.

That calculus may initially favour certain categories, notably Earth observation, defence, climate analysis and space-based sensing systems that already generate large amounts of orbital data.

 

An infrastructure category is emerging

 

The latest SpaceX deal is no proof that orbital data centres can be commercially viable. But it does show that some of the world’s biggest space and AI infrastructure players are starting to take the idea seriously.

Starcloud is not alone. More aerospace and infrastructure companies are looking at orbital computing ideas focused on addressing issues around demand for AI, energy limits, and autonomy. The idea that space will become part of the world’s computing backbone is sounding less and less like science fiction and more and more realistic.

 

 

 

Liked this article? You can support our independent journalism via our page on Buy Me a Coffee. It helps keep MoveTheNeedle.news focused on depth, not clicks.

👉 https://buymeacoffee.com/movetheneedle.news