Lumen Orbit
Unicorn Β· $2.3BRedmond, US Β· Founded 2023 Β· Delaware corporation Β· 61 employees on LinkedIn Β· 26 known investors
Starcloud builds data center infrastructure deployed in space, leveraging continuous solar energy and radiative cooling to enable large-scale compute clusters for AI training. The company targets hyperscale data center operators and AI companies seeking to avoid terrestrial permitting constraints.
Also known as Lumen Orbit, Inc. Β· Starcloud Β· Starcloud, Inc.
Founders & leadership
Lumen Orbit was founded in 2023 by Ezra Feilden, Paul Adrian Oltean, Philip Johnston, and Adi Oltean.
Board

Investors Β· 26
Also in the syndicate Β· 16
Reported raises Β· per SEC filings
Form D private placements$14.1M disclosed across 1 of 10 rounds Β· 2024β2026
βΆ$14.1MraisedJan 2025 Β· 65 investors Β· Other TechnologyRule 506(b)
- Morgan BellerDirector
- Ezra FeildenExecutive Officer, Director
- Paul Adrian OlteanExecutive Officer, Director
- Philip JohnstonExecutive Officer, Director
- Offering amount
- $14.1M
- Amount sold
- $14.1M
- Minimum investment
- $500
- First sale
- Dec 2024
- Incorporated
- Corporation, Delaware, 2023
- Federal exemptions
- 06b
Source: SEC EDGAR Form D. Amounts as filed; amended filings shown once at their latest values.
Valuation Β· disclosed
Disclosed eventsSource: SEC prospectus filings, and round valuations the company or its investors disclosed β follow each entry's link for the claim.
Company profile
researched Aug 2026Starcloud, Inc., founded as Lumen Orbit, designs, builds and operates data centers deployed in space. The company's thesis is that as launch costs decline, orbital compute can exploit continuous solar illumination and radiative cooling to scale toward gigawatt capacity while avoiding terrestrial permitting and power constraints. Its architecture is based on compute pods that can be launched individually and clustered around large deployable solar arrays, with high-bandwidth optical laser links used to move data to and from Earth. The company has published a white paper ("Why we should train AI in space") setting out the case for training AI models in orbit and the architectures it expects to be used, and has described renders and statements concerning a potential 5GW deployment, which it says would require roughly 4 square kilometres of solar array.
The initial product concept was an in-orbit edge-processing constellation: other satellites would transmit raw data to Starcloud's spacecraft, where onboard GPUs would run customer-selected AI models and only the extracted insights would be downlinked, reducing bandwidth, cost and latency. An early plan described a 300-satellite constellation in very low Earth orbit flying beneath customer spacecraft to act as an "in-orbit ground station" providing near-real-time downlink. Alongside this, the company positions orbital capacity for large-scale AI training clusters, arguing that launch and solar costs compare favourably with terrestrial electricity spend.
Demonstration missions have progressed from a 60 kg (132 lb) Lumen-1/Starcloud-1 satellite built on Astro Digital's Corvus-Micro bus, carrying data-center-grade Nvidia GPUs into a 325 km low Earth orbit with an expected 11-month mission life, through to a proposed constellation of up to 88,000 satellites filed with the FCC in February 2026 and a planned Starcloud-2 mission that the company says will fly bitcoin-mining ASICs.
Founding story
Johnston, a former McKinsey consultant who worked on satellite projects for national space agencies and previously co-founded the e-commerce venture Opontia, and Feilden, a satellite designer from Airbus Defence and Space and Oxford Space Systems, were childhood friends who initially considered space-based solar power but concluded that transmitting energy back to Earth was impractical. They pivoted to orbital data centers, which consume solar energy in place and benefit from the cold of space for cooling. They were then introduced to Adi Oltean, a SpaceX Starlink engineer who had been independently exploring the same idea; he joined as co-founder and chief engineer about a month after the company started.
Business model
The company builds and operates its own satellites carrying GPU compute and sells access to that in-orbit capacity and processing to third parties, including satellite operators wanting onboard analysis of their data and AI companies needing training and inference capacity. A partnership with Crusoe is intended to host Crusoe's cloud platform on a Starcloud satellite planned for late 2026, with GPU capacity from orbit expected to be offered commercially from early 2027. The website also invites suppliers and customers to make contact directly.
Sale of in-orbit compute capacity and data-processing services; under the Crusoe agreement, orbital GPU capacity is expected to be commercially offered from early 2027.
Traction
Graduated from Y Combinator's Summer 2024 batch, where it raised one of the largest seed rounds at demo day; TechCrunch reported more than 200 VCs approached the company. Launched the Starcloud-1 demonstrator on a SpaceX Falcon 9 on 2 November 2025 with an Nvidia H100, ran Google DeepMind's Gemma model and trained nanoGPT in orbit in December 2025, filed an FCC proposal for up to 88,000 satellites in February 2026, and reported a $2.3bn valuation following an August 2026 raise. Employee count was listed as 15 in 2026.
Latest developments
In August 2026 the company announced a $250M raise at a $2.3bn valuation, described by one outlet as an extension of the March 2026 $170M Series A, with Nvidia investing $25M and Cisco and Benchmark participating. Earlier in 2026 it filed an FCC proposal for a constellation of up to 88,000 satellites (February), announced plans to mine bitcoin in space using ASICs on Starcloud-2 (March), and signed a contract with SpaceX to integrate Starlink mini-lasers onto its next 25 satellites (May).
βΈFull profile β market position, technology, go-to-market, geography, history, risks & controversies
Market position
One of a small number of ventures pursuing space-based data centers, alongside Axiom Space (with Kepler Space and Skyloom), the EU-funded ASCEND project led by Thales Alenia Space, and lunar-focused Lonestar Data Holdings. Coverage describes the field as not yet crowded, and Wikipedia states Starcloud became the fastest company to reach unicorn status in Y Combinator history, 17 months after the program.
Emphasis on frequent, incremental launches rather than multi-year development cycles; the company said it moved from founding to first launch in about 18 months. It claims to have flown roughly 100x more powerful GPU compute than had previously been operated in orbit and to be first to run and to train a large language model on high-powered GPU hardware aboard a spacecraft.
Technology
Compute pods carrying terrestrial, data-center-grade Nvidia GPUs (including an H100 on Starcloud-1) attached to large deployable solar arrays, using continuous solar power and radiative cooling instead of batteries and active terrestrial cooling. Data is returned via high-bandwidth optical laser links; a May 2026 contract with SpaceX covers integration of Starlink mini-lasers onto the next 25 satellites for high-bandwidth, low-latency connectivity. Starcloud-1 ran inference on synthetic-aperture radar data from Capella Space on orbit, operated Google DeepMind's Gemma large language model onboard a spacecraft, and performed in-orbit training of nanoGPT.
Go-to-market
Direct outreach to suppliers and customers, demonstration missions on SpaceX rideshare launches to validate the technology, and partnerships and accelerator programs including Nvidia Inception, the Google Cloud Accelerator and the Defense Innovation Unit Accelerator, plus a cloud-platform agreement with Crusoe.
Earth-observation and other satellite constellation operators seeking in-orbit processing and faster downlink of insights rather than raw data, plus AI companies and hyperscale data center operators seeking large-scale training and inference capacity.
Geography
Founded in El Segundo, California in January 2024 and relocated in February 2024 to Redmond, Washington (2517 152nd Ave NE), chosen for proximity to space and data center talent at Starlink, Amazon Leo, AWS and Azure; earlier coverage also references Bellevue and Seattle, Washington. Participated in Y Combinator in San Francisco.
History
Founded in January 2024 as Lumen Orbit in El Segundo, California by Philip Johnston, Adi Oltean and Ezra Feilden, moving to Redmond, Washington in February 2024. It emerged from stealth in March 2024 with $2.4M in pre-seed funding, released a white paper and completed Y Combinator's Summer 2024 batch, then closed an $11M seed round at a $40M valuation in December 2024 led by NFX. In March 2025 the company rebranded to Starcloud following a legal challenge from Lumen Technologies over the "Lumen" name for data centers, and raised a further $10M in SAFE funding, bringing the seed total to $21M (Wikipedia cites approximately $34M in total at that point). Starcloud-1 launched in November 2025, and in March 2026 the company announced a $170M Series A at a $1.1bn valuation led by Benchmark and EQT Ventures, followed in August 2026 by a $250M raise at a $2.3bn valuation with Nvidia, Cisco and Benchmark participating.
Risks & controversies
DatacenterDynamics reported that the company's proposals rest on potentially optimistic assumptions about the cost of building and launching large-scale orbital modules, noting its claim that it could launch a 40MW data center for $8.2M. TechCrunch noted the business depends on launch costs continuing to fall and staying low, and that customer adoption may be difficult for space-based data center startups. The company also faced a legal challenge from Lumen Technologies over its original name, prompting the 2025 rebrand to Starcloud. Investors in the $10M SAFE round were not disclosed.
Compiled by commissioned research from 8 cited public sources β announcements, filings, and press listed under research sources below.
Key figures
latest reportedCompany-reported or press-reported figures, each dated to when it was claimed β not independently audited.
Founder mafia
2 people who came through Lumen Orbit went on to found or lead other companies.
Timeline Β· 14
launches, deals, and filingsStarcloud raised an additional $250M at a $2.3bn valuation, with NVIDIA investing $25M and Cisco and Benchmark participating; Payload described it as an extension of the March 2026 Series A.
$250M source β
Starcloud signed a contract with SpaceX to integrate Starlink mini-lasers onto its next 25 satellites for high-bandwidth, low-latency connectivity.
Series A led by Benchmark and EQT Ventures made Starcloud the fastest company to reach unicorn status in Y Combinator history, 17 months after completing the program.
$170M source β
The company said it intended to be the first to mine Bitcoin in space, flying bitcoin mining ASICs on its second satellite, Starcloud-2.
Starcloud submitted a proposal to the FCC for a constellation of as many as 88,000 satellites for orbital data centers.
Starcloud reported being the first company to operate a large language model (Google DeepMind's Gemma) on a high-powered GPU aboard a spacecraft, and the first to perform in-orbit training of a large language model (nanoGPT).
The 60 kg Starcloud-1 (Lumen-1) satellite, built on Astro Digital's Corvus-Micro bus, was deployed from a Falcon 9 into a 325 km low Earth orbit carrying an Nvidia H100 GPU to test training, inference and edge compute workloads; expected mission life was 11 months. Iridium worked with the company on the project.
Crusoe will deploy its Crusoe Cloud platform on a Starcloud satellite planned for late 2026, with orbital GPU capacity expected to be offered from early 2027.
The company changed its name to Starcloud following a legal challenge from Lumen Technologies, which holds rights to 'Lumen' for data centers, and announced $10M of additional seed funding taking the seed total to $21M.
Closed an oversubscribed $11M seed round at a $40M valuation; NFX general partner Morgan Beller joined the board. More than 200 VCs reached out to the company, prompting an additional SAFE at a higher valuation.
$11M source β
Co-founders Feilden, Oltean and Johnston released a white paper arguing for training AI in orbit and outlining candidate architectures.
The company came out of stealth with $2.4M in funding from eight firms plus more than a dozen angel investors, to build in-orbit edge processing that speeds up downlink for satellite operators.
$2.4M source β
Lumen Orbit went through Y Combinator's Summer 2024 program in San Francisco, where it was selected by Tom Blomfield and raised one of the largest seed rounds at demo day.
The company was selected as a member of the Nvidia Inception program, the Google Cloud Accelerator and the Defense Innovation Unit Accelerator; its demonstrator satellite plan was pursued in partnership with Nvidia's Inception program.
Dated company events from announcements, filings, and press; legal rows summarize public dockets and regulator releases.
Legal entities Β· 1
corporate structureIn the news
VCs wanted to get into Lumen Orbit's $11M seed roundtechcrunch.com Β· Dec 2024βΈResearch sources Β· 8
primary sources listed
- Lumen Orbitlumenorbit.com Β· web
8 public sources were cited for this profile; the first-party ones are listed here.
Frequently asked questions
- What does Lumen Orbit do?
- Starcloud (formerly Lumen Orbit) designs and deploys orbital data centers that run GPU compute for AI workloads in space.
- Who founded Lumen Orbit?
- Lumen Orbit was founded by Ezra Feilden, Paul Adrian Oltean, Philip Johnston, Adi Oltean in 2023.
- Who are Lumen Orbit's investors?
- Lumen Orbit's investors include Ambush Capital, Taurus Ventures, Andreessen Horowitz, Benchmark, EQT Ventures, Everywhere Ventures, NFX, Sequoia Capital and 2 more.
- How much funding has Lumen Orbit raised?
- Lumen Orbit has disclosed $14.1M raised across 1 of its 10 known rounds.
- Where is Lumen Orbit headquartered?
- Lumen Orbit is headquartered in Redmond, US.




