On March 30, 2026, Starcloud, a Redmond, Washington startup building compute in orbit, announced a $170 million Series A at a $1.1 billion valuation — unicorn status for a company founded roughly two years earlier. SpaceNews reports the company is awaiting permission to deploy an 88,000-satellite orbital data center network, while a Payload interview lays out where the money goes: a new manufacturing facility for spacecraft sized to fly on Starship.
The timing matters. The day before, OpenAI closed its record $122 billion round (our coverage); the same week, Mistral arranged $830 million in debt financing for its Paris data center (our coverage). With terrestrial power, land, and interconnection queues now the hard limits on AI expansion, capital is starting to price “put the data center in low Earth orbit” seriously — and Starcloud is the first candidate with flight heritage, a concrete schedule, and a unicorn valuation to its name.
Why Put Data Centers in Orbit
The pitch has three planks: near-continuous solar power, radiator-based thermal rejection into vacuum with no cooling towers to site, and a complete bypass of grid interconnection and land permitting. Starcloud is underwriting three businesses on top of that. First, near-real-time compute for space-generated data — satellite imagery and telemetry have limited downlink bandwidth, so processing in orbit and returning results is often cheaper than shipping raw data. Second, hosting sensitive terrestrial workloads in space, where physical isolation buys security and sovereignty. Third, competing head-on with ground data centers for everyday inference.
CEO Philip Johnston was blunt with Payload: the first business already works at Falcon 9 launch prices; the second and third wait for Starship-class cheap mass to orbit. “The big constraint is getting them up cheaply.”
The $170M Series A Lineup
The round arrived in two tranches: Benchmark led the initial close with EQT Ventures participating, and the two co-led the extension. Other backers include funds administered by Macquarie Capital, plus NFX, Nebular, Y Combinator, Adjacent, Seven Seven Six, Fuse Ventures, Manhattan West, and Monolith Power Systems. The angel list skews aerospace-and-power: Gen. Stephen Wilson, former vice chief of staff of the US Air Force; Dennis Muilenburg, former Boeing CEO; and Kevin Johnson, former Starbucks CEO.
Johnston claimed the round could have been far larger — “at the current price, we could’ve easily taken three times the amount of capital” — and compressed the investment case into one line: “you can’t back Google, you can’t back Amazon, and we’re like a mile ahead of any other startup.” Coverage has also described Starcloud as one of the fastest Y Combinator companies ever to reach a unicorn valuation.
From One H100 to Starship: Three Craft in Three Years
- Starcloud-1 (November 2025): flew on SpaceX’s Falcon 9 Bandwagon-4 rideshare mission carrying the first NVIDIA H100 GPU ever flown to space, demonstrating that AI hardware keeps running through radiation and thermal stress.
- Starcloud-2 (later in 2026): targeting 100 times the power of its predecessor and a deployable radiator billed as the largest ever in orbit. It will carry an NVIDIA Blackwell chip, AWS Outposts hardware, and bitcoin-mining ASICs, running edge compute and cloud workloads for partners including Crusoe, AWS, Google Cloud, and NVIDIA.
- Starcloud-3 (Starship-class): the new funding builds a manufacturing facility for spacecraft designed to fit commercial Starship flights, which Johnston expects to begin by the end of 2028. The team grows from 13 employees to as many as 50 by year-end.
The Business Model and Its Real Constraints
Putting bitcoin mining and AI inference on the same satellite sounds like a stunt, but it exposes the same economics: orbital compute cost is set by price-per-kilogram to orbit, and whoever first integrates mass-produced spacecraft with deployable power and radiators turns workloads nobody else can serve into revenue. Starcloud-2 flying AWS Outposts, Blackwell, and mining ASICs together is effectively one platform validating three customer classes’ willingness to pay.
The constraints are just as clear. Until Starship delivers cheap lift, orbital data centers can only serve workloads that are impossible — not merely cheaper — on the ground. The 88,000-satellite network still awaits regulatory permission, and spectrum, orbital slots, and debris rules all remain open variables.
What It Means for Developers and Infrastructure Teams
Three practical observations. First, launch cadence is becoming a compute supply-chain variable: when planning multi-year inference capacity, the orbital option has moved from science fiction to a supply line worth tracking. Second, in-orbit processing of satellite data is the nearest-term credible use case — summarizing and running inference on imagery and telemetry before downlink saves real bandwidth cost. Third, sovereignty and isolation workloads are the second plausible line: for financial and government data under strict compliance regimes, “physically in orbit” is a strong isolation boundary. Until Starship commercializes around 2028, treat Starcloud as a technology outpost to watch — the ground-based, multi-cloud, multi-supplier strategy remains the mainstream answer.
Sources
- Starcloud achieves unicorn status with $170 million raise for orbital data centers — SpaceNews
- Starcloud raises $170M Series A at $1.1B valuation — Payload
- Orbital AI: Seattle-area startup Starcloud hits $1.1B valuation to build space-based data centers — GeekWire
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.