## Does Orbital's Reflex Aerospace Deal Signal a 250 kW Standard for Space Data Centers?

**250 kilowatts per satellite.** That's the per-node power target Orbital is now designing toward for its Orbital-1 platform — more than double its previous ~100 kW goal — after selecting German manufacturer Reflex Aerospace as the exclusive bus partner for its planned orbital data center (ODC) constellation. The announcement, made September 10, 2026, positions Orbital as the first US customer for Reflex, and sets a clear engineering stake in the ground: match the peak compute envelope of a full Nvidia Vera Rubin rack, or don't bother building.

The deal is a design-and-development partnership in which Reflex contributes its flight-tested [satellite bus](https://orbital-intel.com/glossary/bus), freeing Orbital's Los Angeles-based engineering team to concentrate on the compute payload. Target launch for Orbital-1 is 2028. The broader ambition is a [mega-constellation](https://orbital-intel.com/glossary/megaconstellation) of up to 100,000 satellites, each pairing large solar arrays and radiators with a GPU stack cooled passively by the thermal sink of space.

The two companies were introduced through a shared investor, Human Element, and the relationship solidified during a joint engineering visit to Germany.

---

## Why 250 kW — and Why Now

The power target jump from ~100 kW to ~250 kW isn't arbitrary. According to Orbital founder and CEO Euwyn Poon, 250 kW is the peak draw cited for a full Nvidia Vera Rubin rack — the same figure [SpaceX](https://orbital-intel.com/companies/spacex) has cited for its AI1 computing satellite's peak payload power. Poon's logic is straightforward: designing significantly below an emerging benchmark just means you'll need to redesign.

> "It didn't make sense to aim well below the watermark that becomes the benchmark in the next couple years," Poon told Payload.

This is a meaningful architectural decision, not a marketing headline. Sustained 250 kW power generation in [low Earth orbit](https://orbital-intel.com/glossary/leo) requires substantial solar array area and thermal management infrastructure. Space's near-zero-temperature vacuum does provide a passive radiative sink that terrestrial data centers cannot access, but getting heat from GPU stacks to radiators efficiently at this power density is a genuine engineering challenge Orbital will need to demonstrate — not just claim.

The decision to align with Vera Rubin rack specifications also bets heavily on Nvidia maintaining its position as the dominant compute substrate for AI workloads through the late 2020s. That's not a guaranteed outcome, but it's a commercially rational one given current market dynamics.

---

## A Transatlantic Structure With Strategic Intent

The Reflex partnership carries implications beyond hardware. Making Orbital the German manufacturer's first US customer is a deliberate signal about how Poon intends to build the company: as a global supply chain rather than a purely domestic one.

> "We're seeing this landscape as truly a global one," Poon said. "We're looking to pursue partnerships with different sovereigns around the world."

Poon indicated Southeast Asian manufacturing partners are also under consideration. From a supply chain resilience standpoint, this hedges against single-nation procurement risk — relevant for a constellation aspiring to 100,000 nodes. From a political economy standpoint, it positions Orbital to access non-US government incentives and potentially non-US anchor customers.

The Human Element connection that brokered the deal is worth noting: shared investors brokering operational partnerships between portfolio companies is increasingly common in the ODC sector, where the component ecosystem is still forming and trust relationships matter for design-phase commitments.

---

## The ODC Competitive Landscape: Well-Funded and Crowding Fast

Orbital is entering a sector that is attracting serious capital. The source material cites [Starcloud](https://orbital-intel.com/companies/starcloud) raising $250 million at a $2.3 billion valuation. [SpaceX](https://orbital-intel.com/companies/spacex) and [Blue Origin](https://orbital-intel.com/companies/blue-origin) have both filed regulatory paperwork to launch their own large computing constellations. That's formidable competition — two of the three largest launch providers in the world moving to own the compute layer as well.

Poon's counter-positioning is architectural: small satellites at massive scale, rather than large monolithic platforms.

> "It's super clear to me that the future of this space is relatively small satellites, and a constellation of a large number of them," he said.

This is a defensible thesis. A distributed constellation provides graceful degradation, easier incremental deployment, and potentially lower per-unit manufacturing costs at scale. The counterargument is that 100,000 satellites at 250 kW each represents an orbital power deployment and spectrum coordination challenge of staggering complexity — one that has never been attempted. Orbital will need to demonstrate that the per-satellite economics justify the constellation management overhead before investors will fund the mid-stages of that buildout.

---

## Launch Dependency: The Starship Elephant in the Room

Like most ODC operators, Orbital is implicitly banking on [SpaceX](https://orbital-intel.com/companies/spacex)'s Starship to make the economics work. Poon acknowledges this but argues the design isn't Starship-dependent, and that a compelling enough satellite design will secure launch slots on competitive terms.

> "If we do that, the lanes should clear for us to secure launch at economical cost."

That's a reasonable operating assumption for planning purposes, but it's worth flagging the circularity: the same company Orbital is competing against in the ODC market controls the launch vehicle most critical to Orbital's cost model. SpaceX has commercial incentives to sell launch broadly, but it also has direct strategic incentives in the compute-in-orbit space. Orbital will need either a credible backup launch path or a sufficiently differentiated product to negotiate from a position of strength.

---

## Key Takeaways

- **250 kW per satellite** is Orbital's new per-node power target, aligned with the peak draw of a full Nvidia Vera Rubin rack — more than double its prior ~100 kW goal.
- **Reflex Aerospace** is the exclusive bus partner for Orbital-1, contributing a flight-tested platform; Orbital becomes Reflex's first US customer.
- **2028** is the stated target launch date for Orbital-1, the first of a planned constellation of up to 100,000 satellites.
- **Human Element** brokered the partnership through a shared investor relationship.
- Orbital's constellation strategy bets on high node count and small satellite economics rather than large monolithic platforms.
- Launch economics remain implicitly tied to Starship despite management's claim of vehicle-agnostic design.
- Starcloud ($250M raised at $2.3B valuation), SpaceX, and Blue Origin are already active in the ODC space.

---

## Frequently Asked Questions

**What is Orbital's Orbital-1 satellite?**
Orbital-1 is the first satellite in Orbital's planned orbital data center constellation, designed to host GPU compute payloads — specifically targeting the power envelope of a full Nvidia Vera Rubin rack at up to 250 kW per satellite. Reflex Aerospace is providing the satellite bus; Orbital focuses on the compute payload. Launch is targeted for 2028.

**Why did Orbital double its power target to 250 kW?**
Orbital CEO Euwyn Poon cited alignment with the peak power requirements of a full Nvidia Vera Rubin rack and noted that SpaceX's AI1 computing satellite uses the same 250 kW figure. Designing significantly below that benchmark would risk obsolescence against competitors building to the same standard.

**Who is Reflex Aerospace?**
Reflex Aerospace is a German satellite manufacturer providing flight-tested bus hardware. Orbital is its first US customer. The partnership was brokered by shared investor Human Element following a joint engineering visit to Germany.

**How does Orbital compete with SpaceX and Starcloud in the ODC market?**
Orbital's stated differentiation is architectural: a constellation of relatively small, standardized satellites at very high node count (up to 100,000), versus larger monolithic platforms. Starcloud has raised $250 million at a $2.3 billion valuation; SpaceX and Blue Origin have both filed to launch competing constellations.

**Is Orbital's design dependent on Starship for launch?**
Poon says the design is not Starship-dependent, but the company expects cost-effective launch economics to follow from building a competitive satellite. In practice, most ODC operators targeting the late 2020s are implicitly reliant on Starship's full commercial cadence to hit their cost-per-kilogram assumptions.