Starcloud Hits $2.3B Valuation, Triggers Redmond EE Hiring Boom

The Fastest Unicorn

A startup founded in a Los Angeles suburb in January 2024 reached a $2.3 billion valuation before its second birthday, Business Wire reported. In between, it put an NVIDIA H100 into orbit, trained the first AI model in space, and filed for a constellation larger than every active satellite combined. The speed is the story — and the question is whether the physics can keep up with the capital.

Starcloud, formerly Lumen Orbit, designs and operates data centers in space. The pitch: terrestrial AI infrastructure is hitting hard limits on power, water, and permitting. Orbit offers continuous solar energy, passive radiative cooling, and no zoning boards. CEO Philip Johnston, a former McKinsey consultant, founded the company with Adi Oltean, who built hardware at SpaceX and Azure, and Ezra Feilden, an Airbus Defence and Space veteran. They incorporated as Lumen Orbit in El Segundo, then moved to Redmond, Washington within a month to sit beside the talent pools at Starlink, Amazon Kuiper, AWS, and Azure, a relocation that seeded the hiring surge now rippling through the region's satellite cluster.

Summer 2024 brought a white paper, a Y Combinator batch, and one of the largest seed rounds in the accelerator's history. By March 2025, a trademark dispute with Lumen Technologies forced a rebrand to Starcloud and a top-up that brought total seed funding to roughly $34 million. The team was still under 20 people.

The technical milestone arrived November 2025. A 60-kilogram satellite, Starcloud-1, rode a SpaceX rideshare to low Earth orbit carrying an H100 GPU (roughly 100 times the compute of any prior spaceborne GPU), CNBC reported. In December, the company announced two firsts: it trained nanoGPT, Andrej Karpathy's lightweight LLM, on the complete works of Shakespeare, producing the first model trained in space; and it ran Google DeepMind's Gemma, an open LLM derived from Gemini, marking the first time a high-powered GPU served LLM inference in orbit. "From one small data center, we've taken a giant leap toward a future where orbital computing harnesses the infinite power of the sun," said Dion Harris, NVIDIA's senior director of AI infrastructure. Tris Warkentin, product director at Google DeepMind, called Gemma's operation "a testament to the flexibility and robustness of open models."

Capital followed the proof points. March 30, 2026: a $170 million Series A led by Benchmark and EQT Ventures at a $1.1 billion post-money valuation, SpaceNews found — the fastest Y Combinator graduate to reach unicorn status, 17 months after demo day. NVIDIA participated. August 21, 2026: a $250 million Series A extension pushed the valuation to $2.3 billion and total capital to $450 million, TechCrunch reported; Business Wire's data shows. The company now employs roughly 25 people and is building production lines in a 100,000-square-foot facility in Woodinville, Washington, down the road from SpaceX and Kuiper satellite factories.

On February 3, 2026, Starcloud filed with the FCC for up to 88,000 satellites, according to TechCrunch, a constellation that would dwarf Starlink's current fleet. The long-term vision: 20 gigawatts of orbital compute, a 5-gigawatt array spanning four kilometers per side. The next satellite, Starcloud-2, targets October 2026 with 100 times the power generation of the first, multiple GPUs including a Blackwell chip, an AWS server blade, and a bitcoin mining ASIC. Starcloud-3, a 200-kilowatt, three-ton spacecraft, is designed for SpaceX's Starship "PEZ dispenser" deployer.

The bet is explicit: Starship must fly routinely and cheaply, or the economics collapse. Johnston has said as much. "We're not going to be competitive on energy costs until Starship is flying frequently." The first orbital data center that matches terrestrial pricing, he projects, arrives when commercial launch hits roughly $500 per kilogram — a threshold SpaceX has not yet demonstrated. For now, the company has proven it can run the chips. Proving it can scale the constellation is the next test.

Orbit's Cooling Problem and the Launch-Cost Trap

Starcloud's orbital data center proposition rests on two structural advantages terrestrial operators cannot replicate: near-continuous solar energy and a vacuum that functions as an infinite heat sink. The company's first satellite, Starcloud-1, proved the concept in November 2025. The 60-kilogram spacecraft (roughly the size of a dorm-room refrigerator) carried an NVIDIA H100 GPU on a Corvus-Micro bus and delivered 100 times more compute than any previous space-based system.

Power in orbit is not a constraint. It is an abundance. "In space, you get almost unlimited, low-cost renewable energy," said Philip Johnston, cofounder and CEO. "The only cost on the environment will be on the launch, then there will be 10x carbon-dioxide savings over the life of the data center compared with powering the data center terrestrially on Earth." Starcloud projects orbital energy costs at one-tenth of land-based equivalents even after accounting for launch expense. A satellite in the right orbit almost never sees night — no clouds, no winter. Google's own research indicates solar panels in orbit can be up to eight times more productive than ground installations. The company's roadmap calls for a 5-gigawatt platform backed by a solar array spanning roughly four square kilometers. Starcloud-2 satellites, slated for rideshare launches in 2027, will each deliver 8 kilowatts of compute power. The next generation, Starcloud-3, is designed for it.

The free power comes with a catch. The real challenge isn't making energy. It's getting rid of it.

Cooling is where the physics turns hostile. Space is not cold like a refrigerator; it is cold like a thermos. Vacuum is the best insulator known — there is no air, no water, nothing to conduct or convect heat away. A chip in orbit has exactly one path to shed waste heat: infrared radiation. The International Space Station's entire cooling system rejects about 70 kilowatts. A single rack of NVIDIA's latest AI servers draws roughly 120 kilowatts. The station's radiators could not keep one rack alive. Physics dictates that every gigawatt of orbital compute requires approximately one square kilometer of radiator area. Radiators are mass, and mass costs launch dollars. Google's competing prototype, Project Suncatcher, runs in 15-minute bursts before it must cool down. Starcloud's engineers are tracking the relationship between chip operating temperature and radiator sizing as a primary design variable.

Radiation adds a third axis of difficulty. High-energy particles moving at near light speed penetrate electronics and flip bits — a zero becomes a one. On Earth, a technician can swap a failed board. In orbit, nobody comes. The NVIDIA Space-1 Vera Rubin module, which Starcloud plans to integrate, is built for this environment: radiation-hardened, cooled by large deployable radiators. Johnston's team is also optimizing radiation shielding placement and the ruggedizing required to survive the vibration and acoustic loads of launch.

The economics only close if launch costs fall from today's roughly $3,000 per kilogram on Falcon 9 to around $200 per kilogram — a fifteen-fold reduction that depends entirely on Starship becoming operational. Boston Consulting Group calculated that launching one gigawatt of compute at current prices would cost $30 billion before purchasing a single chip. SpaceX plans to retire Falcon 9 by 2028. Starship has yet to demonstrate routine commercial flight. Starcloud's FCC filing for 88,000 satellites assumes a launch cadence that does not yet exist.

The next milestone is Starcloud-2. Two 8-kilowatt satellites on rideshare missions in 2027 will test GPU clusters, persistent storage, and integrated power systems in a compact bus. Blackwell GPUs, expected to deliver up to 10× the performance of the current Hopper architecture, are slated for later flights. A 100,000-square-foot production facility in Woodinville, Washington, is already building the line. The technology works in prototype. The question is whether the launch market can deliver the tonnage at a price that makes orbital compute a business instead of a stunt.

Capital Bets on a Rocket That Hasn't Flown

Starcloud's capital trajectory has compressed what typically takes a decade into two years. The company, founded in 2024 under the name Lumen Orbit, closed a $34 million seed round backed by Andreessen Horowitz and In-Q-Tel, the CIA's strategic investment arm. By March 2026, the $170M round led by them (with NVIDIA participating) valued the startup at $1.1 billion, making it the fastest unicorn in Y Combinator history at 17 months. Five months later, that extension led by Manhattan West Ventures pushed it to $2.3 billion. NVIDIA contributed $25 million to that extension; Cisco also participated. Total capital raised since founding: $450 million.

The money buys a bet on physics. Starcloud has asked the FCC for permission to operate 88,000 spacecraft delivering 20 gigawatts of orbital compute capacity. That constellation would dwarf every existing satellite network. The first production-class vehicle, Starcloud-3, is such a spacecraft designed to ride SpaceX's "PEZ dispenser" deployment system on Starship. Starcloud-2, an 8-kilowatt pathfinder, targets rideshare launches in 2027; commercial operations would begin roughly 30 days after reaching orbit, with paying workloads from Crusoe, AWS, and Google Cloud scheduled in the first 90 days.

The economics only close if they collapse. CEO Philip Johnston has said the first orbital data center becomes cost-competitive with terrestrial facilities at roughly $0.05 per kilowatt-hour — but only if commercial launch reaches $500 per kilogram. That price point is a Starship promise, not a Falcon 9 reality. SpaceX's own timeline has slipped: Musk said in August 2026 that a Starship catch attempt would be delayed months, with re-flight not expected until late 2026 or early 2027. The space-ready Blackwell GPU Starcloud plans to fly hasn't been built yet; the target is late 2028. Johnston frames the 88,000-satellite constellation as a 2035–2040 milestone, not a near-term promise.

Capital intensity is brutal. A Boston Consulting Group report from April 2026 pegs orbital data centers at 2.5 to 3 times the fully loaded cost of terrestrial equivalents: $660–750 million per megawatt in orbit versus $230–300 million on the ground. Introl forecasts the orbital data center market growing from $1.77 billion in 2029 to $39.09 billion by 2035, a 67.4 percent compound annual growth rate. Big Tech's 2026 AI infrastructure spend is on track to exceed $650 billion. Seven gigawatts of planned U.S. AI capacity has stalled on interconnection queues; utilities are preparing a $1.4 trillion grid overhaul. The terrestrial bottleneck is real. Whether orbital compute clears it depends on a rocket that hasn't yet flown commercially.

Industry sources expect Starcloud to raise another $300–500 million between mid-2027 and early 2028 at a $3–5 billion valuation, contingent on Starcloud-2 hitting performance targets. That round would fund the Starcloud-3 manufacturing ramp and the initial sub-constellation. An IPO is not on the current roadmap before 2029 at the earliest, and only if SpaceX's own public debut lands well. The entire architecture (satellites, factory, constellation, business model) rides on Starship. If Starship slips, the $450 million buys a very expensive ground test program.

Redmond's Satellite Cluster and the Talent Crunch

Starcloud's headcount sits at 25 today, up from a founding team of three in early 2024, and the company is actively recruiting. A live Electrical Engineer posting lists a Redmond base, a $100,000–$200,000 salary band, and a hard requirement for U.S. citizenship or export-authorization eligibility, standard for ITAR-controlled satellite work. The role asks for ownership of high-reliability electronics and PCBs from concept through orbital operation. That single listing is the visible tip of a hiring wave the company's $450 million war chest is designed to fund.

The Redmond area already hosts the densest satellite-manufacturing cluster on the planet. SpaceX and Amazon's Project Kuiper operate sprawling factories within miles of each other, and together they account for the majority of the nearly half of active satellites built in Washington state. The state's space sector employs thousands of workers and generates $8.1 billion in annual economic output, while the broader Greater Seattle aerospace complex supports roughly 114,000 jobs — more than triple the national average for a metro this size.

Mayor Angela Birney stood at the April 2023 State of the City and announced the Redmond Space District, a formal partnership between the city and OneRedmond to turn a half-century of aerospace history into a coordinated economic engine. The district's mandate is explicit: support the existing cluster, recruit new companies, coordinate with regional partners, and strengthen the workforce pipeline from K‑12 STEM through senior engineering hires. It was not a ceremonial gesture. Redmond already produces more satellites than any other city on the planet. Alliance Velocity research cited by the city shows Redmond‑based companies built nearly half of every active satellite in orbit, and current FCC filings put the region on track for more than 75 percent of future manufacturing.

The roots run to 1967, when Aerojet Rocketdyne broke ground on its first Redmond site (then horse pastures and barns) to build thrusters for Apollo and, decades later, Artemis II. That propulsion heritage attracted the two satellite manufacturing giants that now dominate low‑Earth orbit: SpaceX Starlink and Amazon Project Kuiper. Around them sits a dense supply chain: Honeywell Aerospace, Kymeta, Microsoft Azure Orbital, RBC Signals, Spectralux, Triumph, and a crop of startups like Xplore and Starcloud that lease space in the same industrial corridors. Starcloud's headquarters at 2517 152nd Ave NE places it a few miles from SpaceX's Starlink campus and Amazon's new 219,000‑square‑foot Kuiper facility, which combines offices, R&D labs, and prototype manufacturing under one roof. The company's 100,000‑square‑foot production line in nearby Woodinville is explicitly designed to feed the "PEZ dispenser" deployment system SpaceX built for Starship.

"Washington has deep roots in the space industry," Governor Bob Ferguson said when he signed the executive order creating the Washington Space Council in September 2026. The 23‑member council (announced during Seattle Space Week at the Museum of Flight) is charged with assessing the state's capabilities, identifying investments and policy levers, and exploring homegrown commercial launch capacity. That last item matters: every orbital data center plan, Starcloud's included, runs into the same bottleneck — affordable, high‑cadence launch. The council's work dovetails with the Redmond Space District's talent‑pipeline goal, linking state‑level policy to the shop‑floor hiring needs of companies that collectively employ thousands of space workers and generate billions in economic activity across Washington.

What emerges is a feedback loop: public investment de‑risks private capital, private capital draws talent, talent density attracts the next startup, and the cycle repeats. The Redmond Space District gave that loop a name and a budget. The Washington Space Council gave it a policy voice in Olympia. For an electrical engineer weighing an offer in Redmond versus Hawthorne or Kent, the distinction is concrete: you are not joining a single company's program; you are entering a cluster where the person who designed your satellite's thruster, the team that wrote its ground‑station software, and the founder launching the first orbital H100 all eat lunch in the same business park.

The Market for Flight Hardware

The talent crunch in Redmond isn't a headline — it's a daily constraint. SpaceX, Blue Origin, Amazon Kuiper, and now Starcloud are all fishing from the same specialized pool: engineers who can design hardware that survives launch vibration, rejects radiation in orbit, and dissipates kilowatts of heat into vacuum. The shortage isn't about headcount. It's about the intersection of power electronics, thermal-mechanical integration, and radiation-hardened compute — a Venn diagram with a tiny center.

Skills That Get You Past the Screen

Starcloud's own job descriptions make the priority explicit: electrical engineers who take "full ownership of complex hardware projects from concept to production and orbital operation." That means board-level power conversion, FPGA-based control loops, and the kind of thermal architecture where a radiator's view factor to deep space determines your compute ceiling. Since Starcloud-1 flew the first H100 in November 2025, the company has trained an AI model in orbit, run Gemini, and demonstrated high-powered inference on flight hardware. Every one of those milestones required engineers who understand GPU cluster bring-up and the single-event upset rate of commercial silicon at 325 km.

The same skill set appears across the Redmond cluster. Blue Origin's TeraWave optical communications team in Kent is hiring principal modem engineers and RFIC/mmWave leads, roles that demand photonics packaging experience alongside space-qualification discipline. Amazon Kuiper's hundreds of open positions span RF payload, antenna arrays, and ground-segment software. SpaceX's Redmond operation, still the largest employer in the district, lists principal roles in silicon DFT, CDN architecture, and supercomputing for special programs — the last explicitly requiring Top Secret clearance.

Security clearance is a force multiplier. The SpaceX supercomputing role in Redmond specifies TS/SCI. Starcloud's government customer pipeline (U.S. agencies for Starcloud-2 inference tasks in 2027) implies similar requirements. Candidates who already hold clearance, or can obtain it, skip a 12- to 18-month queue.

What the Money Looks Like

First-party board data from Zero G Talent shows the current bands for the two largest Redmond employers. SpaceX carries 1,413 salaried roles with a median of $150k and a band of $52k–$255k. Blue Origin shows 1,143 salaried roles, median $183k, band $81k–$276k. Both added over 100 roles in the past week alone.

Starcloud, at roughly 25 employees as of August 2026, doesn't yet appear in aggregate board data. But a $450M war chest and a 100,000-square-foot Woodinville production facility signal compensation competitive with the upper quartile of the table above, especially for engineers with flight-hardware tape-outs on their résumé.

The Interview Gauntlet

SpaceX's Redmond process bifurcates sharply. Starlink software-engineer candidates describe a selective loop that fails a large portion of applicants. New-grad software engineers, by contrast, report an easy process where the vast majority receive offers. The split reflects two different hiring motions: experienced hires for the operational constellation, and volume recruiting for the next cohort.

Blue Origin publishes its Fair Chance Act compliance explicitly: criminal histories are considered consistent with Washington, California, and local ordinances. That openness broadens the aperture for candidates who self-select out of traditional defense contractors.

Starcloud's process is less documented, but the company's "concept to orbit" ownership model suggests a practical, hardware-in-the-loop evaluation. The team is sharing thermal and radiation data with NVIDIA for the Vera Rubin Space-1 chip; engineers who can speak to that feedback loop (radiator sizing versus chip junction temperature, shielding mass versus single-event functional interrupt rate) will move fast.

Leverage in a Cluster

Redmond's density is the candidate's advantage. Within a 15-mile radius sit SpaceX's Starlink factory, Blue Origin's Kent campus, Amazon Kuiper's R&D, Aerojet Rocketdyne's heritage propulsion, and now Starcloud's Woodinville build-out. The Redmond Space District, launched in 2023, and the Washington Space Council, stood up in September 2026, exist to keep that cluster expanding. Engineers with flight heritage can interview at three majors in one week — and use each offer to move the others.

Starcloud's next milestone is Starcloud-2, a pair of 8 kW compute satellites targeting rideshare flights in 2027. The team building that bus (power, thermal, compute, radiation) is hiring now. If you've qualified a board for orbit, the door is open.

Working in space? Zero G Talent tracks the openings: see every open SpaceX role, browse space jobs, openings at Blue Origin, and the people building the field.