SpaceX's AI Endgame: Owning the Infrastructure Layer of Intelligence

TLDR: SpaceX is targeting a mid-2026 IPO at $2+ trillion valuation while making only $24-30 billion in revenue — a price-to-sales ratio over 64x. The thesis: SpaceX is positioning itself as the physical infrastructure layer for AI through orbital datacenters, TeraFab chip manufacturing, and Starlink global connectivity.

Summary: The numbers around SpaceX's upcoming IPO are genuinely staggering. A target valuation above two trillion dollars for a company generating between twenty-four and thirty billion in revenue yields a price-to-sales ratio exceeding sixty-four — more than twice Nvidia's historical peak. For comparison, Amazon, also valued around two trillion, generated seven hundred and seventeen billion in revenue in 2025. SpaceX would have twenty-three times less revenue for the same market cap.

The bull case rests on the infrastructure thesis: SpaceX is not a rocket company with AI ambitions but rather a vertically integrated infrastructure play where Starlink provides global connectivity, orbital datacenters provide space-based compute, and TeraFab provides chip manufacturing. The xAI merger, the orbital datacenter FCC filing, and the Starlink network are not separate bets — they are three parts of the same thesis. If you accept that global datacenter power consumption will roughly double to nearly one thousand terawatt-hours by end of decade, space-based solar arrays beaming energy to Earth starts sounding less like science fiction and more like infrastructure planning.

The execution risks are extraordinary. TeraFab is a proposed twenty to twenty-five billion dollar joint venture between Tesla, SpaceX, and xAI to create a full-stack chip manufacturing facility — design, lithography, fabrication, memory production, advanced packaging, and testing under one roof. It targets the two-nanometer process. They are not TSMC, and building competitive chip fabrication from scratch is one of the most capital-intensive and technically demanding endeavors in industrial history.

Meanwhile, xAI burns at least one billion dollars per month. X still carries roughly twelve billion in acquisition debt while making thirty-five percent less revenue than in 2022. Tesla sales declined fourteen percent from the previous quarter. The proceeds from the SpaceX IPO are expected to fund everything from Starship scaling to a lunar colony to Mars colonization. That is a lot of weight on a single IPO, however large.

Key takeaways:

SpaceX targeting $2T+ IPO with only $24-30B revenue (64x+ price-to-sales)
The thesis: infrastructure layer for AI (orbital compute + connectivity + chips)
TeraFab: $20-25B full-stack chip fab targeting 2nm process
xAI burns $1B/month; X carries $12B debt; Tesla sales declining
Three mega-IPOs in 2026-2027: SpaceX, OpenAI, Anthropic

Why do I care: For developers and architects, the infrastructure layer story matters because it signals where compute and connectivity costs are heading. If orbital datacenters become viable (big if), the constraints around edge computing, latency, and data sovereignty shift dramatically. More practically, the SpaceX, OpenAI, and Anthropic IPOs in the next year will define the financial landscape of AI tooling — the companies whose APIs we build on every day are about to face public market accountability.

SpaceX's AI Endgame: Owning the Infrastructure Layer of Intelligence