How does Envision Group’s operational and planned renewable powered AI infrastructure in Inner Mongolia challenge SpaceX CEO Elon Musk’s proposal to place AI data centers in orbit as a solution to electricity constraints, and what are the key differences between the two approache

Envision’s Inner Mongolia project is a practical counterexample to the claim that AI’s electricity constraint requires moving compute into orbit: it places large-scale AI hardware directly beside low-cost renewable generation on Earth. But it does not prove terrestrial renewables will always beat space; the economics of both approaches, especially at future scale, remain unproven.

The core contrast

Dimension	Envision Galaxy Campus, Ulanqab	SpaceX orbital-data-center proposal
Location	Terrestrial: Inner Mongolia, near renewable resources and in China’s emerging western/northern computing regions. 3	Low-Earth orbit: a proposed constellation of solar-powered AI-computing satellites. 5
Operational status and scale	Envision has commissioned an initial phase of a planned gigawatt-scale campus; reporting describes a planned 2 GW facility. 352	No operational AI-data-center constellation is identified in the evidence. SpaceX has applied for up to 1 million satellites, and Musk has said first AI satellites could launch in Q4 2027, with “significant scale” in 2028. 54
Energy source	Local wind and solar directly connected to compute, with batteries intended to smooth variable output. 52	Near-continuous orbital solar power, avoiding terrestrial grid bottlenecks in principle. 5
Basic technology	Conventional, serviceable terrestrial data-center infrastructure combined with renewable generation, storage, intelligent power management, and naturally cool climate conditions. 352	Distributed, solar-powered satellites using AI chips, linked as an orbital computing network. 45
Electricity constraint addressed	Moves AI workloads to places where clean power is abundant rather than waiting for constrained urban-grid interconnections. 352	Removes the facility from Earth’s grids altogether by generating power in orbit. 5

Why Envision challenges Musk’s premise

• Envision shows that the immediate answer to constrained grid capacity can be geographic and infrastructural rather than extraterrestrial: colocate compute with wind, solar, storage, and cooler climates already available on land. 352

• Its advantage is maturity and deployability. A commissioned terrestrial campus can use standard servers, cooling, networking, maintenance crews, and replacement cycles, while orbital computing remains dependent on a future satellite architecture and launch cadence. 34

• The comparison is not “renewables versus solar”: both proposals rely on solar-derived renewable energy. The difference is that Envision uses land-based wind/solar and storage, while SpaceX would use space-based solar generation and must launch, cool, operate, network, and eventually replace computing hardware in orbit. 52567

Reliability and engineering trade-offs

• Envision must manage intermittency: wind and solar output varies, so its reliability rests on storage, power management, possible curtailment/flexible workloads, and any backup or grid arrangements not specified in the available evidence. Battery smoothing is reported, but public evidence here is insufficient to establish its firm 24/7 renewable capacity or outage performance. 52

• SpaceX could receive much more consistent solar exposure in suitable orbits, but space is not inherently easy to cool. In a vacuum, a data center cannot shed heat through air convection; thermal control and radiators become fundamental engineering constraints. 67

• Orbital systems also face harsh operating conditions and difficult repair or hardware-replacement cycles. Experts cited by Reuters point to unresolved cooling, launch-cost, and component-environment issues. 7

Cost comparison: known advantage versus speculative advantage

• Envision’s potential economic case is straightforward: avoid or reduce the need for scarce grid interconnection, use abundant local renewables, and lower cooling needs in a cold region. A reported estimate of roughly $0.05 per kWh is not independently verified in the supplied evidence and should therefore be treated as indicative, not an established delivered cost. 52

• SpaceX argues that orbital solar and reusable launch could ultimately make space compute cheaper. Yet the European Parliament’s assessment identifies economics—especially future launch cost—as the primary barrier, noting current reusable-launch costs remain on the order of several thousand euros per kilogram. 1

• A separate, non-peer-reviewed analysis assumes $200/kg Starship launches and estimates that placing 1 GW of AI compute in low-Earth orbit could require 10,000 tonnes of satellites, about 50 Starship V3 launches, and $2 billion in launch cost alone. Those are assumptions, not an achieved SpaceX cost or a complete project-cost estimate. 2

• Neither approach has a settled all-in cost per AI compute unit. For Envision, missing variables include server/accelerator prices, storage duration and replacement costs, transmission/network costs, utilization, and the realized price of renewable electricity. For SpaceX, the unknowns additionally include launcher reuse and cadence, satellite mass, solar arrays, radiators, radiation hardening, insurance, failures, communications, replacement launches, and depreciation of rapidly obsolete AI chips. 167

Deployment timeline and strategic meaning

• Envision is already operating an initial phase of its Inner Mongolia campus, making it a near-term response to AI-power constraints; the full multi-gigawatt ambition remains a plan rather than completed capacity. 352

• SpaceX’s schedule is aspirational: Musk’s stated target is initial satellites in late 2027 and significant scale in 2028, contingent on execution, approvals, manufacturing, and low-cost high-cadence launch capability. 45

• In Chinese strategic terms, Envision fits a broader terrestrial pattern: concentrate computing in regions with plentiful energy and land, then connect that capacity to demand centers. The supplied evidence supports Inner Mongolia’s role as an AI-infrastructure location, but it is insufficient on its own to quantify Envision’s contribution to China’s national computing strategy or to attribute specific national-policy outcomes to this one campus. 3

What remains uncertain

• Server costs and demand: AI accelerators depreciate quickly, and future demand for training versus inference, model efficiency, and utilization rates could change whether multi-gigawatt campuses—or a million-satellite constellation—are economically justified.

• Terrestrial electricity economics: Envision’s advantage depends on sustained cheap renewables and effective storage/power management; variable generation means nominal renewable capacity is not the same as guaranteed compute availability. 52

• Orbital launch economics: SpaceX’s case depends heavily on future, not current, launch-cost reductions. The $200/kg estimate is an assumed future benchmark, while the European Parliament notes that launch cost remains the decisive economic constraint. 12

• Technical reliability: Envision confronts familiar terrestrial problems; SpaceX must demonstrate reliable high-density AI operation, heat rejection, component durability, networking, and replacement in space at unprecedented scale. 67

Thus, Envision does not disprove the long-run possibility of orbital AI compute. It does undermine the notion that Earth-bound electricity scarcity leaves no scalable alternative: large renewable-powered terrestrial AI campuses are already being deployed, whereas orbital data centers remain a high-risk, capital-intensive future proposition.