Skip to main content

Starcloud

Category: AI Infrastructure

Building megawatt-scale orbital data centers that leverage continuous solar power and radiative cooling in space to solve energy and cooling constraints for AI compute workloads. Starcloud was founded in 2024. The company is led by Philip Johnston. Based in Redmond, United States. Team size: 12-27. Total funding raised: $200.0M. Latest round: Series A ($170.0M, Mar 2026). Key investors include Benchmark, EQT Ventures, Macquarie Capital, NFX, Y Combinator, Andreessen Horowitz, Sequoia Capital, In-Q-Tel (CIA's venture arm), NVIDIA.

AMW Analysis

Starcloud builds megawatt-scale orbital data centers that use continuous solar power and radiative cooling to address energy and cooling constraints for AI workloads. The company’s value proposition centers on avoiding Earth’s energy bottlenecks, with projected energy costs of $0.05/kWh once launch costs reach ~$500/kg. Since its 2024 founding, Starcloud has raised $200M total, including a $170M Series A in March 2026 at a $1.1B valuation, making it the fastest Y Combinator unicorn. The startup has already deployed an Nvidia H100 GPU in orbit and trained the first LLM in space, and plans to launch a more powerful Starcloud-2 this October featuring Nvidia Blackwell and AWS infrastructure. A partnership with AWS will put AWS Outposts into orbit this October. Starcloud has also filed an FCC proposal for 88,000 satellites, indicating a scaling strategy toward a large constellation. The recent news flow shows rapid fundraising, early orbital GPU deployment, major cloud partnerships, and ambitious satellite plans, positioning the company to move from prototype to large-scale orbital compute.

AMW analysis, generated from 2 tracked news signals.

Founded
2024
Headquarters
Redmond, United States
Team size
12-27
Total funding
$200.0M

Value proposition

Unlocks unlimited solar power and passive radiative cooling in orbit, enabling scalable AI compute to gigawatt scale while avoiding Earth's energy bottlenecks, land constraints, and permitting delays. Projects 10x lower energy costs and cost-competitiveness with terrestrial data centers at $0.05/kWh once launch costs reach ~$500/kg.

Products and solutions

Starcloud-1: 60kg proof-of-concept satellite with first NVIDIA H100 GPU in orbit (launched Nov 2025), Starcloud-2: Multi-GPU cluster satellite with NVIDIA Blackwell, AWS server blade, Bitcoin mining ASIC (launching Oct 2026), Starcloud-3: 200kW, 3-ton commercial spacecraft designed for SpaceX Starship deployment, Orbital Data Center Constellation: Planned network of up to 88,000 compute satellites

Unique value

First company to deploy data-center-class GPU (NVIDIA H100) in orbit and successfully train an AI model (nano-GPT LLM) in space (100x more powerful than any prior space-based compute). Pioneering the 'Space-for-AI' paradigm by relocating power-hungry AI training workloads to orbit where unlimited solar energy and the vacuum of space provide natural cooling without water consumption.

Target customer

AI/ML companies, cloud hyperscalers (AWS, Google Cloud), AI infrastructure providers (Crusoe), satellite operators, enterprises requiring large-scale compute capacity, organizations facing terrestrial energy constraints

Industries served

Artificial Intelligence & Machine Learning, Cloud Computing & Data Centers, Aerospace & Defense, Satellite Communications, Energy Infrastructure, Cryptocurrency/Blockchain

Technology advantage

Combines proprietary thermal management systems (largest commercial deployable radiator) adapted from ISS technology with continuous solar power generation (no batteries needed) and passive radiative cooling using deep space as an infinite heat sink. Achieves 10x lower energy costs by leveraging falling launch costs (SpaceX Starship at ~$500/kg target) while avoiding terrestrial permitting, land acquisition, and grid connection delays. Designed for rapid scalability from 60kg satellites to 3-ton, 200kW spacecraft using SpaceX's PEZ dispenser deployment system.

How they differentiate

First-mover advantage with operational proof-of-concept satellite (Starcloud-1 with NVIDIA H100 GPU in orbit since Nov 2025), successfully trained first AI model (nano-GPT LLM) in space. Proprietary thermal management systems adapted from ISS technology combined with continuous solar power generation and passive radiative cooling. Achieved unicorn status ($1.1B valuation) in just 17 months post-Y Combinator, making it the fastest YC startup to reach unicorn status.

Main competitors

SpaceX (with xAI merger), Google Project Suncatcher, Aetherflux

Key partnerships

NVIDIA (Inception program partner; first H100 GPU deployment in orbit), SpaceX (launch provider for Starcloud-1 rideshare mission; Starship integration planned), Amazon Web Services (Starcloud-2 will feature AWS Outposts server blade), Google Cloud (partnership for edge and cloud workloads), Crusoe (early commercial customer for AI infrastructure), EQT Ventures (strategic investor operating 70+ terrestrial data centers), In-Q-Tel (CIA's venture capital arm - strategic investor), Y Combinator (accelerator program, fastest company to reach unicorn status at 17 months), Benchmark (Series A lead investor)

Notable customers

Amazon Web Services, Google Cloud, Crusoe, AI/ML companies, Cloud hyperscalers

Major milestones

Founded as Lumen Orbit in January 2024, rebranded to Starcloud in March 2025, Y Combinator Summer 2024 cohort - raised largest seed round at demo day, First NVIDIA H100 GPU deployed in orbit (November 2025), First company to successfully train AI model (nano-GPT) in space, Filed FCC proposal for constellation of up to 88,000 satellites, Series A led by Benchmark and EQT Ventures (March 2026), Achieved $1.1B unicorn valuation - fastest YC startup to reach unicorn status

Growth metrics

Achieved unicorn status ($1.1B valuation) just 17 months after Y Combinator demo day, making it the fastest YC startup to reach unicorn status. Deployed first NVIDIA H100 GPU in orbit (Nov 2025) and successfully trained first AI model in space. Planning constellation of up to 88,000 satellites.

Market positioning

Early leader in orbital AI infrastructure with proven technology and strongest funding ($200M total). Positioned as pure-play orbital data center company focused on megawatt-scale compute infrastructure, targeting cost-competitiveness with terrestrial data centers at $0.05/kWh once launch costs reach ~$500/kg via SpaceX Starship.

Geographic focus

Headquartered in Redmond, Washington (USA); filed FCC proposal for constellation of up to 88,000 satellites serving global AI/ML companies, cloud hyperscalers, and enterprises facing terrestrial energy constraints. Primary market focus on North America, Europe, and Asia-Pacific AI infrastructure demand.

Patents and IP

Co-founder Adi Oltean holds 25+ unique patents from Microsoft and SpaceX (including Starlink tracking beam technology). No specific Starcloud patents publicly disclosed yet, but company has developed proprietary thermal and power systems for orbital compute.

About Philip Johnston

Serial entrepreneur and former McKinsey & Company consultant specializing in aerospace and defense, leading satellite projects for national space agencies. Previously co-founded Opontia, an e-commerce brand aggregator that raised over $50M before being acquired. Educational background: MPA in National Security & Technology from Harvard University, MBA from Wharton School, MA in Applied Mathematics & Theoretical Physics from Columbia University, and CFA Charterholder.

Latest news about Starcloud

More AI Infrastructure companies

Official website: