Hello friends, and welcome to The Unsophisticated Investor! Brought to you by Scott & Rob, the founders of Shuttle!

If you want to invest alongside the VC funds who've backed breakout companies like Revolut, Asana, JustEat, Bolt, Lets Get Checked, Loom, Runna, Charlotte Tilbury, Deel, Aircall, AngelList, Carta, TransferWise and many more, regardless of your knowledge, network or net worth, join our limited waitlist now.

Now, let’s get into it 👇

Two weeks ago we wrote about SpaceX repricing the space economy. Last week, geopolitics and capital. This week, Nvidia put a chip in orbit and declared space computing "has arrived." The next chapter is already underway.

Before we get to the physics of cooling a data centre in a vacuum, a quick stop in Redmond, Washington.

Starcloud was founded 21 months ago by three engineers. CEO Philip Johnston holds degrees from Columbia, Wharton and Harvard and spent time at McKinsey working on satellite projects for national space agencies. CTO Ezra Feilden and chief engineer Adi Oltean between them shipped features in satellite networks, operating systems and machine learning infrastructure at SpaceX and Microsoft. The company has twelve employees and has raised $34 million to date.

In November 2025, Starcloud launched its first satellite aboard a SpaceX Falcon 9. The satellite, roughly the size of a small refrigerator and weighing 60 kilograms, carries a single Nvidia H100 GPU, 100 times more powerful than any computer previously operated in space. Within weeks of launch, it had trained a language model on the complete works of Shakespeare, then ran inference on Gemma, Google's open-source model, in orbit. The first message it returned: "Greetings, Earthlings."

Among the early backers is Finn Murphy, the Irish solo GP behind New York-based Nebular. A former partner at Dublin's Frontline Ventures, Murphy left in 2023 to build his own fund on a curiosity-first thesis: the founder is the wedge, category comes after. His debut portfolio, $30 million deployed across 24 positions, includes one of the most-watched companies in the emerging orbital compute category. The early conviction on Starcloud didn't come from a specialist space fund. It came from a solo GP operating out of New York on a thesis most institutional LPs would have found difficult to underwrite. That's not a coincidence, but a pattern. The most interesting early-stage bets in frontier categories are rarely made by the obvious actors.

Why this is an AI story, not a space story

The context for orbital data centres is not aerospace ambition. It is a very terrestrial infrastructure crisis.

Global data centre power consumption is projected to more than double by 2030. The average capex cost for a greenfield project now exceeds $3 billion. Grid connection timelines in many markets stretch into years. The people building the largest AI systems in the world are running into limits that are not software problems, they are land, power, water and planning problems. And some of the most serious technologists in the industry have started asking whether the answer is to stop building down here on Earth.

Elon Musk has suggested orbital solar-powered compute could be the lowest-cost way to run AI infrastructure within four or five years. Jeff Bezos has said gigawatt data centres in space will exist within a decade. At Nvidia's annual developer conference this week, Jensen Huang announced the Vera Rubin Space-1 chip module, purpose-built for orbital AI data centres, and declared that space computing "has arrived." Google's Project Suncatcher, announced late 2025, includes a planned launch date for TPU-equipped satellites in 2027 and a long-term vision of 81-satellite clusters forming orbital AI compute grids.

This is not fringe speculation. It is where the leading edge of the industry is pointing.

The case for space

The physics are genuinely compelling. Solar arrays in orbit receive sunlight far more consistently than any ground installation — no cloud cover, no atmosphere, no night at certain orbits. Starcloud projects 10 times lower energy costs than terrestrial data centres over the life of the asset. Cooling works differently too. On Earth, enormous volumes of water and electricity are consumed moving heat away from servers. In orbit, the vacuum of deep space acts as an infinite heat sink, with infrared radiation dissipating waste heat continuously and no water required.

The radiation concern is also being revised. Google tested its Trillium TPU chips under proton beam exposure simulating five years in low-Earth orbit. The chips showed no hard failures up to 15 krad of total ionising dose. Using standard data centre chips in space, with appropriate error correction rather than full hardening, is starting to look feasible. And launch costs continue to fall: Falcon 9 brought the cost per kilogram to orbit down to roughly $2,500, Starship could push it toward $200.

The case for Earth

The engineering sceptics are not wrong. Space is not cold, it's a vacuum, and a vacuum is an insulator. The only mechanism for rejecting heat in orbit is thermal radiation. A large data centre running tens of megawatts requires enormous radiator arrays that must be launched and deployed. Starcloud's own 5-gigawatt vision involves a solar and radiator structure approximately four kilometres across, more than 2,000 football pitches of hardware in orbit.

Latency is a structural constraint. Even at the speed of light, low-Earth orbit introduces 20 to 50 milliseconds of round-trip delay. Maintenance is impossible: when hardware degrades, the satellite de-orbits. Starcloud-1 has an expected mission lifetime of eleven months, after which it will re-enter and burn up. And the gap between a single H100 in a 60-kilogram satellite and a 5-gigawatt orbital data centre is not an engineering roadmap, it is several orders of magnitude of unsolved problems.

How VCs are approaching this

The smart money is not betting on orbital hyperscale data centres in the 2020s. It is betting on something more specific and near-term.

The use case that works today is on-orbit processing of data generated in space. Earth observation satellites produce enormous volumes of imagery. Today, most is downlinked to ground stations over slow radio frequency links, processed terrestrially, then acted upon, a delay that can stretch to days. Starcloud's partnership with Capella Space is designed to run AI inference on synthetic aperture radar imagery in orbit before downlinking. Only relevant outputs, fires detected, vessels identified, changes flagged, come down. The bandwidth efficiency is significant, the latency reduction is real, and the customer is willing to pay.

That is a fundable near-term business. It doesn't require a four-kilometre radiator array. Starcloud-2, launching in October 2026, will have 100 times the power-generating capability of its predecessor and, the company says, will generate more cash than it costs to build and launch.

The broader VC picture reflects this calibration. Global venture investment in space technology reached $4.5 billion across 48 companies in the first half of 2025, more than four times what the sector attracted in all of 2024. The in-orbit data centre market is projected to reach $39 billion by 2035. But the thesis experienced investors articulate is not that orbital compute replaces terrestrial data centres. The enabling infrastructure for whatever comes next is being built now, and the window to back it at early-stage valuations is open.

The pattern is familiar. When broadband was being built out, the long-term winners were often the picks-and-shovels businesses, fibre manufacturers, routing hardware companies, billing software providers. The orbital compute equivalent might be thermal management specialists, inter-satellite laser link providers, radiation-tolerant memory designers. These are not household names yet. They are being funded in private markets right now.

Finn Murphy's position in Starcloud illustrates where that value gets created. He didn't back a category. He backed a founder, earlier than the market, when the bet looked eccentric rather than obvious. By the time Nvidia announces a purpose-built space chip and Jensen Huang declares the category has arrived, the entry point has moved. That is how private markets work.

What this means for private market investors

Two things can be true simultaneously. Orbital data centres at gigawatt scale are probably a decade away from being a real infrastructure category. And the foundational work that makes them possible is being funded and validated right now, by serious capital, with serious engineers, at early-stage valuations that will look different once public market benchmarks exist.

Nvidia's announcement at GTC 2026 matters for the same reason the SpaceX IPO matters, it doesn't create the opportunity, but it reprices it. When the world's dominant AI chip company announces a purpose-built space compute module with a cohort of companies already deploying it, the category moves from speculative to validated. That is the moment when generalist capital starts paying attention. It is rarely the best moment to enter.

The structural forces here are the same ones driving everything else we have been writing about. AI compute demand is growing faster than terrestrial infrastructure can respond. Launch costs are falling on a predictable trajectory. The convergence of those two curves creates a window. The question for investors is not whether orbital compute will eventually matter — the evidence is accumulating quickly. The question is which layer of the stack captures the returns, and whether the companies building that layer are accessible before the public market puts a number on them.

Starcloud launched its first satellite 21 months after it was founded. Its second launch is seven months away. The pace at which this category is moving is not the pace of traditional aerospace. It is the pace of software-era startups operating in a domain where the physics have finally become permissive enough to let them run. 

^{_{What we’ve bee working on at Shuttle}}

• Continuing to prep for our funding round, exciting things on the horizon 💰

• Deep in new feature development and product roadmap planning 👀

The Unsophisticated Investor is brought to you by Scott & Rob, the founders of Shuttle. We’re both sick of private markets being a playground exclusive to the ultra-wealthy so we started a company to challenge the status-quo. Shuttle’s singular focus is to unlock private markets for Millennial and Gen Z tech professionals and help them build wealth through the highest performing private market opportunities.

Scott & Rob
Shuttle Co-Founders