It’s happening 400 miles above your head.

Starcloud trained the first AI model in orbit last month. Google is launching test satellites in 2027. SpaceX is deploying 30,000 satellites equipped with built-in computing power. China just launched 12 satellites for what it calls the “Three-Body Computing Constellation,” with plans to scale to 2,800.

This isn’t sci-fi. This is infrastructure.

And here’s the part that should make every supply chain strategist and investor sit up a little straighter: the entire global capacity to move this infrastructure to space is controlled by exactly two entities, SpaceX and China. Why Data Centers Are Moving to Space

The math is brutal and non-negotiable.

AI data centers are projected to consume 13% of U.S. electricity by 2030, up from 5% currently. A single large hyperscale facility burns 100–200 megawatts, and roughly half of that goes to cooling alone. On top of that, you’re competing for grid capacity with residential users, you need land, you need water, and you need all of it in the right place at the right time, while “the right place” keeps shifting as power prices and regulations move faster than you can build.

On Earth, you’re stuck with a zero-sum optimization problem: trade-offs between land, water, energy, and cost.

In space, you solve most of that in one move:

• 24/7 solar power: No sun at night? Not an issue in orbit. You get continuous power generation with no grid dependency.

• Natural cooling: The vacuum of space is a near-perfect heat sink. No pumps, no water, no cooling towers; thermal radiation does the work.

• Zero land footprint: You deploy infrastructure where it doesn’t compete with housing, agriculture, or industrial use.

• Lower energy cost: Starcloud projects 10x lower energy costs than terrestrial data centers, with carbon breakeven within five years even after launch emissions.

This isn’t innovation theater. It’s energy arbitrage with physics on your side.

Who’s Actually Building This

Starcloud (formerly Lumen Orbit)

Starcloud has raised $24M from Nvidia, In-Q-Tel, and Y Combinator and launched Starcloud-1 on November 2, 2025, with an Nvidia H100 GPU—the first time this level of compute has gone to orbit. They trained an AI model in space in December 2025, and Starcloud-2 is scheduled for October 2026 with Nvidia Blackwell chips.

They’re not experimenting for press releases. They’re deploying.

Google Project Suncatcher

Google’s first test satellites launch in early 2027 to validate TPU performance in sun-synchronous orbit. CEO Sundar Pichai has already said this will be “the normal way to build data centers within 10 years.” When a Google CEO calls something “normal,” the market eventually treats it that way.SpaceX Starlink V3

SpaceX

Elon Musk has said that “simply scaling up Starlink V3 satellites would work” as orbital data centers. Each V3 satellite delivers around 1 Tbps download and 160–200 Gbps upload, and SpaceX plans to deploy roughly 30,000 of them starting Q4 2026 with high-speed laser links between satellites. That’s effectively a 30,000-node distributed computing constellation.

Axiom Space

Axiom is deploying orbital data center nodes on the ISS in 2025, with Kepler Space and Skyloom providing optical links. Think of this as the MVP: the first live customer validation is happening now.

China’s Three-Body Computing Constellation

China deployed 12 satellites in May 2025 with a target of 2,800 total and an expected capacity of 1 exaFLOPS once complete. Each satellite delivers 744 TOPS and 100 Gb/s optical links, backed by a state-directed program with effectively unconstrained capital and loose timeline pressure.

The Real Constraint: Launch Capacity

Here’s the part nobody in traditional logistics circles is really modeling yet:

The entire global capacity to lift 20+ tons to low Earth orbit—the minimum for meaningful orbital infrastructure—sits on six operational rocket platforms.

Not six companies. Six vehicles.

• SpaceX Falcon Heavy: 64 tons to LEO at roughly $1,500/kg. Proven and dominant.

• ULA Vulcan Centaur: 24.6 tons to LEO, still in its early testing phase.

• Blue Origin New Glenn: Advertised at 45 tons, but early flights will be closer to ~25 tons.

• China Long March 5: 25 tons to LEO, state-controlled and geopolitically restricted for many Western customers.

• Russia Angara A5: 24 tons to LEO, sanctioned and unreliable for commercial access.

• NASA SLS Block 1: 95 tons to LEO, at roughly $2 billion per launch and not commercially available.

That’s the entire heavy-lift universe.

SpaceX already accounts for roughly 88.5% of satellite launches as of Q2 2025, and they dominate by default because no one else has the capacity. Starship Changes the Economics

Starship is still in flight testing, but once it becomes operational sometime around 2026–2027, the entire architecture changes.

Starship targets 100–150 tons to LEO for roughly $10–20M per launch. Falcon Heavy currently sits around $1,500/kg; Starship is targeting sub-$100/kg. That’s a 15x step-change in cost per kilogram, and it completely rewrites what’s economically viable to send to orbit.

Today, moving a 10-ton data center module to orbit on Falcon Heavy costs about $15M. Starship drops that closer to $1M. The core question shifts from “can we afford this?” to “why wouldn’t we do this?”Supply Chain Implications

Concentration Risk

Orbital data center deployment is effectively gated by SpaceX’s execution roadmap and China’s state capacity. This is concentration risk on par with port consolidation or Class I rail monopolies—except this time you can’t negotiate with physics or geopolitics.

If Starship’s development slips, the whole timeline pushes out by 18–24 months. If China ends up dominating orbital computing capacity, Western firms get a rerun of today’s semiconductor constraints—only this time at the infrastructure layer.

For Supply Chain Strategists:

1. Start modeling orbital compute procurement now. If Google and the hyperscalers deploy gigawatt-scale orbital infrastructure in 2027–2028, your cloud RFP playbook changes. Providers with 10x lower energy costs either cut your bill or pad their margins, but either way, your cost structure moves.

2. Watch launch manifests like spot rates. Once Starship is operational, launch slots start to look a lot like tight truckload capacity on a peak lane. That’s a brokerage and optimization opportunity. 3PLs that understand orbital logistics early will own a new segment.

3. Plan explicitly for geopolitical exposure. If your business depends on compute-heavy workloads, you now have exposure to launch vehicle access controlled by two entities: SpaceX and China. Your resilience planning needs to treat U.S.–China tech competition as a direct operational variable, not background noise

4. Track the energy ripple effects. Cheaper compute energy means more compute deployed, which cascades upstream into demand for power infrastructure, satellite manufacturing, launch services, orbital servicing and refueling, and debris remediation.

For Investors:

1. The play isn’t just “pick the right space startup.” The real bottleneck is launch capacity and cadence. Starship’s ability to reliably lift 100+ tons determines the addressable market for orbital data centers. Watch test-flight cadence—each successful flight can effectively pull the market forward by a year or more.

2. The infrastructure layer has the leverage. Axiom Space, Kepler Space, and Skyloom are assembling the backbone: connectivity, power, thermal management, and debris tracking. Axiom’s ISS data center nodes in 2025 are the MVP signal that customers will pay for orbital “landlord” services.

3. The Chinese state’s backing changes the rules. The Three-Body Computing Constellation is a $10B-plus bet over 5–7 years and functions as geopolitical infrastructure more than a commercial product. Western investors should assume Chinese dominance in orbital compute by 2030 unless Starship economics and U.S. policy move aggressively.

4. Software and orchestration are the big upside. Starcloud’s early lead with Nvidia backing positions it to own a large share of orbital compute. But the real upside is orchestration software: provisioning, failover, and optimization across orbital and terrestrial infrastructure is easily a $50B-plus opportunity.

5. Cadence beats specs. Starship’s payload capacity makes headlines, but the real unlock is high launch frequency. If Starship hits 100+ launches per year, the capacity constraint evaporates—that’s the actual inflection point to watch.

The Bottom Line

Space data centers are a genuine supply chain inflection point on the level of containerization, rail electrification, or port automation. The new modal effectively comes online around 2026, the capacity constraint is locked in for the next 3–5 years, and the concentration risk is real.

What changes:

• Energy economics for compute infrastructure

• Geopolitical exposure tied directly to launch vehicle access

• Procurement strategies as orbital compute becomes a standard cloud option

• Demand patterns for launch services, orbital infrastructure, and space logistics

What doesn’t:

• The physics of heat dissipation, orbital mechanics, and radiation exposure

• The geopolitical reality that only two entities control meaningful heavy-lift capacity

• The fact that first movers win by operationalizing orbital infrastructure at scale, not by having the slickest slide deck.

If you’re a supply chain strategist, your next step is to map your enterprise’s compute intensity and model what 10x cheaper energy at the infrastructure layer actually does to your network design. If you’re an investor, your next step is to trace capital flows from launch capacity to infrastructure to orchestration software.

Orbital logistics is no longer a thought experiment. It’s a new modal. The question is not whether it’s coming.

The question is whether you’re building for it now—or scrambling in 2028 when Google lights up its first gigawatt-scale orbital data center.

Share this with your supply chain team, your CIO, or your portfolio manager. The next 18 months are critical.

FreightFA Brief drops daily. Subscribe for the supply chain intelligence that actually moves your business.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit freightflowadvisor.substack.com/subscribe