In June 2026, Derek Huerta did something unusual for a man who had just watched SpaceX list on Nasdaq. He left the company and started a business whose premise is that no other nation should have to depend on SpaceX to reach orbit.
Huerta was the senior manager of satellite payload engineering at SpaceX. Around him when he walked out the door were people who had built the Starlink phased arrays, the telemetry radios, and the supply chain that took constellations from prototypes to dozens of satellites a week. Thirteen had been there since the constellation was called Satellite Development.
Their new company is Eclipse Space, out of stealth on June 26, based in Redmond, Washington, the same city where Starlink satellites are still assembled. Redmond matters. The founding team of roughly thirty sits across the street from the factory that scaled the only megaconstellation so far.
The company targets governments that have been priced out of owning space infrastructure. Today the alternatives are renting from a foreign operator or buying from China.
First customer hardware is due in late 2026, an integrated demonstration satellite is planned for 2027, and the model claims a minimum output of five satellites per day.
2025 ──────── 2026-06-12 ──── 2026-06-26 ─── 2026 H2 ───── 2027
[🔧] [💼] [📣] [⌛] 🔥
Eclipse SpaceX lists Eclipse First customer Integrated
founded in on Nasdaq comes out of hardware demo satellite
Redmond stealth
Source of chronology: SpaceNews, Via Satellite, company materials
The timeline is the pitch. SpaceX went public, the people who built the constellation put the windfall into a less well-known belief: the next ten constellations will not be built in one factory.
Why the founders left the factory
Here is the logic in one pass. A megaconstellation of a thousand satellites was, for a decade, a thing only two or three organisations in the world could build. The capital bill ran to billions, the supply chain took years to stand up, and the manufacturing had to be vertically integrated. Starlink's advantage was never the satellite. It was that a handful of engineers had the full stack, from phased array to radio, power, software and launch cadence, under one roof, iterating weekly.
Huerta's point, put plainly: vertical integration had a specific job, done. The next constellations serve a broader, slower-moving market with a different constraint. Consider what a customer actually needs to shift the pattern. A government that rents bandwidth has no control over pricing, coverage decisions, or whether the service survives a geopolitical shock. A government that owns its satellites holds those decisions.
Eclipse emerged with that logic and a technology transfer in hand. It launched with customer engagement across Europe, Asia and the Middle East, plus a commercial pipeline in telecom and satellite services. First hardware deliveries are due before the end of 2026. The flagship step up: a demonstration spacecraft in 2027 that validates the bus, the radios and the power before the operational "Starlink-class" model scales.
Fabless, applied to orbit
Fabless manufacturing is the semiconductor term for a company that designs the chip, owns the intellectual property, and lets a foundry do the physical fab. Eclipse is applying the same split to satellites: define the design and process, stand up the integration elsewhere, and hand the network to the end customer. The old model made the whole stockpile a single-factory bottleneck; the split removes it.
The economics only work if a constellation is a repeatable build. It claims a minimum production rate of five satellites a day at scale, and for a sovereign client it stands up an in-country production line with a local partner, satisfying local-content requirements as a matter of course. That is the opposite of the single-factory model: the distributed build is the product.
To accelerate design, it took in the engineering team behind Agent Studio, an AI development platform from Rendered.ai, under an exclusive licence with an option to buy the underlying technology. It is unusual for a satellite company to open with an AI acquisition. Huerta's rationale: most teams use AI to help write code, but Eclipse wants tooling for the harder problem of spacecraft engineering, the bespoke designs, and the long tail of systems work.
The sovereign red line
Most of the world is on track to never own its space infrastructure. Until now, the only routes were renting from a foreign operator or building a space industrial base from scratch. For the customer the acceptance problem is real: owning hardware means owning the risk of operating it. It argues that its regional build network is the answer to that, because partners carry the local operating base and the political work.
For the customer, the upside of sovereign ownership is contractual: capacity that cannot be switched off, service shaped and run on their own terms. For a vendor like Eclipse the product is the choice itself: instead of renting, a nation owns the constellation and decides what happens to it.
As we wrote in August around Amazon Leo's waiver, the LEO broadband race is real competition for sovereign paths already; the capacity race is not a separate lane. Its bet is that the sovereign path is an addition to that race, not a subtraction from it.
What has to be proven
Eclipse is honest about the open items. A fabless satellite is a good story, but satellites are one-piece physical systems, and the phased array is the hardest part to transfer. The 2027 demonstrator deliberately foregoes the operational payload, focusing on bus: a total-ionising-dose sensor, a Langmuir probe, plus the base systems. The proof of the model is not the satellite. It is a customer who signs before the hardware is proven.
Which is why the company stresses near-term delivery. The prototype phased array and a telemetry, tracking and command radio reach users this year, before the full constellation; the 2027 demonstration turns that technology into a public milestone.
The market story in the next 24 months will be defined by a simple number: how many countries choose to own rather than rent. That number, not the technology, is the first-order variable for an investor.
Comparison grounded in SpaceNews and TechTimes reporting on Eclipse's sovereign model
What would confirm the sovereign bet
The first sovereign constellation contracts, actual signed deals and not preliminary agreements
Whether the 2027 demo satellite lifts on schedule and withstands the first public milestone
The choice of regional integration partners and which countries open in-country assembly lines
How the Rendered.ai engineering team matures into product velocity in spacecraft design
The machine that turns a lease into property
For the investor the specific reason to study Eclipse Space is not the satellite. It is the act of turning a lease into property. A category that has spent years under two companies now has a credible third path, and the first company to make ownership tangible becomes the default vendor for every country that was waiting for this moment.
As the constraints close (hardware delivered, demo flown, sovereign contracts published), more capital is likely to follow. This is a private-market watchlist item, not a public index. That is exactly where the informational edge lives.
Detail-rich trade coverage used as the anchor for the company history and the market argument.
A second witness for the launch narrative with the AI-tooling angle in detail.
The counterweight piece: what sceptics would add to the prospect of distributed satellite production.
| Parameter | Rent Starlink/LEO service | Own a sovereign constellation |
|---|---|---|
| Capital outlay | Operating spend | Large up-front CAPEX |
| Control of coverage | ✗ provider decides | ✔ operator decides |
| Local content | ✗ usually none | ✔ in-country assembly |
| Geopolitical exposure | ✗ high | ◐ lower (IP still shared) |
