· investment-strategies · 6 min read
What Is Starcloud’s Valuation in 2026? $2.3B After $250M
Starcloud added a $250M Series A extension on August 21, 2026, at $2.3B. Manhattan West led. Google still ranks the March $1.1B unicorn print. Nvidia’s $25M check is people-familiar, not a company figure.
Starcloud is worth $2.3 billion after a $250 million Series A extension on August 21, 2026. Manhattan West led. Benchmark and Y Combinator were already on the cap table from March.
Unexpected truth: Google still ranks the March $1.1 billion unicorn print and The Information’s ~$2.2 billion talks. The live mark is $2.3 billion. Nvidia’s $25 million is a person-familiar figure in TechCrunch, not a company number. Revenue was not disclosed.
This page is for people deciding whether orbital inference (or this YC space-compute lane) is worth another hour. Last verified August 25, 2026. Fact vs our read is labeled below. We do not have a company extension post; facts below are TechCrunch plus the March Business Wire / YC page.
Five-minute decision
| If you need… | Verdict |
|---|---|
| What happened | Funded. $250M Series A extension @ $2.3B, Aug 21, 2026. Manhattan West lead. |
| What Starcloud is | Orbital inference satellites, not a terrestrial neocloud. First named GPU in orbit: Nvidia H100 on Starcloud-1. |
| Whether it is “shipping” | One satellite flown (Nov 2025). Next: Starcloud-2 rideshares in 2027. Starcloud-3 needs Starship. |
| Whether to diligence | Yes, if you underwrite launch capacity + space GPUs. No, if you need a revenue multiple or a proven Starship cadence first. |
Investigate further when: you believe the scarce asset is booked launch + a space-qualified H100-class GPU, and $2.3B is a factory-and-manifest print.
Wait or pass when: Falcon 9 winding down in 2028 without Starship cadence is a hard stop, or you think Groq already occupies the inference allocation.
What happened
| Field | Detail |
|---|---|
| Company | Starcloud (starcloud.com); CEO Philip Johnston; HQ Redmond, WA |
| Round | $250M Series A extension · $2.3B valuation · Aug 21, 2026 (TechCrunch; company told TC) |
| Lead | Manhattan West Ventures (no /fund/ page) |
| Directory funds | Benchmark (March co-lead, returning), Y Combinator |
| Other names in TC | Nvidia, Cisco, EQT, Soma, NFX, 776, Cedar Capital, Goanna Capital, Standard Capital |
| Nvidia dollars | $25M — person familiar (TechCrunch). Not a company figure. |
| Use of proceeds | Larger manufacturing facility; Starcloud-3 (Starship-class orbital data center); booking launch |
| Traction disclosed | Starcloud-1 H100 in orbit (Nov 2025); FCC filing for 88,000 spacecraft; ~25 employees; 100,000 sq ft Woodinville, WA line |
| Prior close | $170M Series A @ $1.1B, Mar 30, 2026, Benchmark + EQT (Business Wire) |
Do not treat this as a Series B. The company (via TechCrunch) calls it an extension of the March A.
What Starcloud actually does
Product (company / YC): data-center satellites. First job is GPU compute for other satellites; later job is AI energy that does not wait on terrestrial interconnect.
Three named vehicles:
- Starcloud-1 — flown November 2025 with an Nvidia H100. Company/YC: first to train an LLM in space and run a Gemini version in space.
- Starcloud-2 — 8 kW compute satellites on 2027 rideshares; inference for customers including U.S. government agencies (TechCrunch).
- Starcloud-3 — larger craft intended for SpaceX Starship. March coverage described a ~200 kW, three-ton PEZ-dispenser-class bird. August coverage: still the Starship bet.
Launch is the constraint, not the model. Johnston told TechCrunch Falcon 9 is scheduled to end in 2028, competing vehicles are not flying regularly, and “if we can’t book any SpaceX launch capacity in 2029, that will be challenging.” Musk said the same week that a Starship catch attempt slips a few months, with a re-fly aimed at end of 2026 or early 2027 (via TechCrunch).
How big is Starcloud actually?
| Question | Disclosed? | What we have |
|---|---|---|
| Valuation | Yes (to TC) | $2.3B (Aug 21) |
| New cash | Yes (to TC) | $250M extension |
| Prior A | Yes | $170M @ $1.1B (Mar 30). Business Wire: ~$200M total raised then. Arithmetic now is ~$420M if you add $250M to that $170M — do not treat that as a company “total raised” disclosure. |
| Nvidia check | People-familiar | $25M (TechCrunch). |
| Revenue / contracted MW | No | — |
| Headcount | Yes (TC) | ~25 |
| Manifest | Plan | Two Starcloud-2 rideshares in 2027; dedicated Falcon 9 under consideration; Starship for Starcloud-3 |
| FCC | Filing | Permission requested for 88,000 spacecraft — a filing, not a license we verified |
| Older SERP | Stale | $1.1B unicorn pages; The Information ~$2.2B talks. Live mark is $2.3B. |
One flown GPU is not a data-center business. That is the most useful sentence on this page besides $2.3 billion.
Worked numbers (indicative)
- Setup: Mar 30, 2026 — $170M at $1.1B, Benchmark + EQT; YC’s fastest unicorn (17 months after demo day).
- Move: Aug 21, 2026 — $250M more at $2.3B, Manhattan West, with Nvidia/Cisco as strategics.
- Punch: If $250M is new primary at $2.3B post, implied dilution is about 10.9% ($250M / $2.3B). That is headline arithmetic, not a disclosed option pool. The Information’s ~$2.2B target was the rumor; $2.3B is the print.
Same-month inference tape (not Starcloud’s P&L): Groq printed $3.5B on August 17. Etched printed $21B on August 18. Starcloud sits under Groq on price and next to SpaceX on launch.
Starcloud vs Groq — what differs
| Starcloud | Groq | |
|---|---|---|
| Latest mark | $2.3B, Aug 21, 2026 | $3.5B, Aug 17, 2026 |
| Last raise | $250M A extension | $350M Series A (reset / new entity) |
| Product | Orbital inference satellites | Terrestrial inference neocloud |
| What “scale” means | Launch slots + space GPUs + radiators | Megawatts + racks |
| Directory names | Benchmark, YC | Disruptive (no /fund/ page); planned NVIDIA |
| Main risk | Starship cadence + 2028 Falcon 9 sunset | Utilization and cluster ops |
Choose Starcloud when you are underwriting space compute. Choose Groq when you are underwriting Earth inference. Do not mash the two $B marks.
Why investors likely wrote $250M
Known evidence (TechCrunch + March primaries)
- Only known operator of a terrestrial-class H100 in orbit; first to train a model on it (company via TC / YC).
- Nvidia investing now because of Starcloud-1 data, per Johnston; Nvidia is developing Vera Rubin Space-1, a purpose-built space GPU (not built yet; Starcloud hopes to fly it late 2028).
- Woodinville plant next to SpaceX and Amazon satellite lines.
- March A already had Benchmark + EQT (EQT owns terrestrial data centers — March PR). This extension is launch insurance as much as chip R&D.
Our interpretation (not an IC memo)
| Bet | Why it could justify $2.3B |
|---|---|
| Launch scarcity | CEO is raising to book Starship/Falcon before the 2028 cliff |
| Nvidia technical diligence | Chipmaker as LP is a signal, even if the $25M is anonymous |
| Category FOMO | AI power on Earth is the 2026 bottleneck story |
| YC speed | Fastest YC unicorn (March claim) keeps the name in every space-compute roundup |
What we cannot claim: a $/kWh that beats terrestrial once Starship prices are real, or that 88,000 spacecraft is an approved constellation.
What has to be true for Starcloud to win
- Starship flies often enough to make orbital kWh competitive — Johnston’s March figure was on the order of $0.05/kWh if launch is about $500/kg (TechCrunch, March). That is a conditional, not a 2026 P&L.
- Starcloud-2 2027 rideshares actually fly and serve a government inference customer.
- The Woodinville line produces more than one-off birds.
- Nvidia’s space GPU exists by late 2028 and Starcloud is on it.
- Someone else does not lock the Starship manifest first.
What could break the thesis
- Starship slips past the window Johnston called out for 2029.
- Radiation, thermal, or launch loads kill H100-class economics in orbit.
- The $2.3B mark was a scarcity print that later launch comps will not pay.
- Mixing $25M Nvidia (anonymous) with the $250M extension as if both were company 8-Ks.
When not to use this page
- Do not treat $2.3B as the March figure. March was $1.1B.
- Do not treat The Information’s ~$2.2B target as the live mark.
- Do not invent a
/fund/page for Manhattan West, EQT, NFX, 776, or Nvidia. - Do not use the FCC 88,000 request as a licensed constellation.
- Do not confuse this with SpaceX buying Cursor — same week’s orbital story, different product.
Where to go next
Month tape: August 2026 VC news. Same-week roundup: August 21–25 VC news. Prior mark: Starcloud $170M Series A. Inference comps: Groq $3.5B, Etched $21B. Launch owner: SpaceX. Entity: Starcloud. YC applications: Fall 2026 guide.
