Interview

Starcloud: Solving AI’s Energy Problem in Orbit

Philip Johnston is the co-founder and CEO of Starcloud, the company building data centers in space. In November 2025, Starcloud launched an Nvidia H100 GPU into orbit and trained the first large language model in space. They've since raised $200 million, hit a billion-dollar valuation just 17 months after YC demo day and filed with the FCC to deploy 88,000 more satellites. In this episode, Philip walks us through their wild origin story, the engineering challenges behind the Starcloud-1, why they booked a SpaceX launch before they even knew what they were building and how data centers in space make sense both economically and politically.

Chapters:
00:00 - First thing every space company should do
01:06 - Why data centers in space
01:35 - The Starbase trip that started it
02:53 - Asteroid mining, space hotels, and the ideas they passed on
03:20 - Why space-based solar doesn't work
04:38 - Launching Starcloud-1
05:35 - Cooling an H100 in an ice bath at 5AM
06:24 - The separation video
07:15 - First contact and the satellite that kept restarting
09:42 - The two hardest problems: heat and radiation
11:27 - Testing GPUs in a particle accelerator
12:49 - Automotive parts instead of space-grade
13:55 - "The dumbest thing I've ever heard"
14:30 - What 100 VCs got wrong
18:38 - The path to 20 gigawatts in space
20:31 - First customers
21:16 - Designing a GPU for space with NVIDIA
23:09 - A Bitcoin miner in orbit
24:00 - Why hard tech suddenly got popular with VCs
25:37 - Inside the Benchmark round
26:57 - Their hiring strategy
28:08 - How to be contrarian but right
30:49 - Recruiting co-founders before having an idea
31:44 - What if data centers get banned on Earth?
32:45 - Data center myth busting
35:09 - Advice for hard tech founders

Source attribution: Original source

Interview details

Publication date: Aug 14, 2026
Duration: 36m

Speakers

Diana HuManaging Partner, Y Combinator Garry TanCEO, Y Combinator Harj TaggarManaging Partner, Y Combinator Jared FriedmanGeneral Partner, Y Combinator Philip JohnstonCEO, Starcloud

Viewpoints in this interview

Space data centers could solve AI energy bottleneck using abundant solar energy.

Philip Johnston argues that building data centers in space solves the AI energy bottleneck by providing access to abundant, low-cost solar energy, as terrestrial energy constraints make new construction increasingly difficult.

Philip Johnston Space Data Centers Source at 1:08

Reusable rockets lower space data center costs

The core idea for space data centers became viable because the breakeven launch cost is around $500 per kilogram, a more achievable target than the $50 per kilogram required for space-based solar, due to the emergence of reusable, high-capacity launch vehicles.

Philip Johnston Launch Costs Source at 3:37

Engineering challenges for space data centers

The two primary engineering challenges for space data centers are managing heat dissipation in a vacuum and making commercial chips operate reliably in a high-radiation environment. Solutions include deployable radiators and extensive ground testing for radiation tolerance.

Philip Johnston Engineering Challenges Source at 12:18

Booking the earliest available launch creates immediate momentum and a hard deadline.

The recommended first step for any space startup is to book the earliest available launch as a forcing function, even before the payload is fully designed. This creates immediate momentum and a hard deadline.

Philip Johnston Startup Strategy Source at 16:53

Investor sentiment shifted after VC rejections due to falling launch costs and terrestrial barriers.

The initial idea of space data centers was met with strong skepticism, leading to over 100 VC rejections. Investor sentiment shifted due to the convergence of falling launch costs and increasing terrestrial barriers to building energy-intensive infrastructure.

Philip Johnston Funding and Investor Reception Source at 14:42

Growing AI energy demands and terrestrial limits make space-based compute an alternative.

AI's growing energy demands are hitting terrestrial limits, with new data center construction being banned in multiple states. This political and regulatory headwind makes space-based compute a necessary and potentially unavoidable alternative.

Philip Johnston AI Energy Bottleneck Source at 32:36

Founders succeed in hard tech by recruiting top engineers and handling commercial operations.

Success in hard tech is fundamentally about assembling the right technical team. The founder's role is to recruit the world's best engineers for the core problem and then handle the commercial and operational aspects around their expertise.

Philip Johnston Team Building Source at 16:17

Starcloud-1 demonstrated an NVIDIA H100 GPU in orbit using immersion cooling.

Starcloud-1 was a proof-of-concept satellite that successfully ran an NVIDIA H100 GPU in orbit using a novel phase-change material immersion cooling system, demonstrating the feasibility of data-center grade hardware in space.

Philip Johnston Satellite Technology Source at 5:06

Space compute faces engineering hurdles in interconnectivity, thermal management, and radiation hardening.

Diana highlights that while space offers abundant energy, it presents distinct engineering hurdles including data interconnectivity, thermal management without convection, and radiation hardening for electronics.

Diana Hu Engineering Challenges Source at 11:23

Path to profitable space data centers

Jared probes the sequencing of Starcloud's development, questioning how the company plans to progress from initial test missions to a profitable, commercial-scale data center network.

Jared Friedman Startup Strategy Source at 19:24