SpaceX filed plans with the FCC in January 2026 to deploy a megaconstellation of up to 1 million satellites designed to function as orbital AI data centers. The goal: reach 100 gigawatts of AI compute capacity per year by 2030, powered entirely by solar energy in space.

To put that in perspective, the entire US grid currently supplies roughly 1,200 GW of total capacity across all uses. SpaceX wants to park a meaningful fraction of that kind of power overhead, dedicated solely to AI workloads, within four years.

From rockets to racks in the sky

The plan calls for initial launches aboard SpaceX’s Starship rocket beginning in 2028, with a nearer-term target of 1 GW per year of compute capacity by late 2027. That ramp-up is supposed to be fueled by a new 11-million-square-foot manufacturing facility announced in June 2026, nicknamed the “Gigasat” factory, which SpaceX expects to produce more than 1,000 satellites annually by late 2027.

SpaceX already operates the Starlink constellation, which consists of more than 10,000 satellites providing broadband internet. The AI satellite initiative represents a fundamentally different use case: instead of beaming internet to rural Montana, these birds would crunch neural network computations in orbit.

Solar power in space is abundant and uninterrupted by weather or nighttime cycles. Cooling is easier in the vacuum of space than in a warehouse in Texas. And scaling up doesn’t require negotiating with local utilities for gigawatt-scale power connections, a process that currently bottlenecks terrestrial data center expansion across the US.

The xAI merger and the trillion-dollar bet

SpaceX completed an all-stock acquisition of xAI in February 2026, creating a combined entity valued at approximately $1.25 trillion. The merger effectively fused Elon Musk’s rocket company with his AI company, consolidating the hardware layer (rockets and satellites) with the software layer (AI models and training infrastructure) under one corporate roof.

Commercial interest appears to be materializing quickly. Google reportedly signed a capacity agreement worth $920 million per month starting in October 2026 for satellite compute access. In English: Google is willing to pay roughly $11 billion per year for orbital AI processing power before a single dedicated AI satellite has even launched.

For context, Google’s parent company Alphabet spent around $50 billion on capital expenditures in recent years, much of it on data centers and AI infrastructure.

What this means for crypto and decentralized compute

The crypto angle here is less obvious but potentially significant. Decentralized compute networks like Render, Akash, and io.net have built their value propositions on the premise that AI compute is scarce, expensive, and concentrated in the hands of a few hyperscalers. SpaceX’s plan, if it works, could dramatically expand the total supply of available compute, which complicates the scarcity thesis that underpins much of the decentralized AI infrastructure narrative.

Bitcoin miners have increasingly positioned themselves as flexible energy buyers who can co-locate with data centers. If a meaningful chunk of AI compute migrates to orbit, that could free up terrestrial power capacity and grid connections for other large-scale energy consumers, including mining operations.

The risks are substantial enough to keep skeptics busy. Deploying a million satellites creates an orbital debris problem that regulators haven’t fully grappled with. The financial investment required for full operational capacity could run into the trillions of dollars. Regulatory approvals across multiple jurisdictions remain uncertain. And the sheer manufacturing challenge of producing satellites at automotive-industry scale has no precedent in aerospace history.

Investors watching the decentralized AI compute space should be tracking SpaceX’s execution timeline closely. If the Gigasat factory hits its production targets by late 2027 and early Starship launches succeed in 2028, the competitive landscape for AI infrastructure could look fundamentally different by the end of the decade.