Buried inside the SpaceX S-1 filing is a definition that has received almost no coverage: Macrohard. Built jointly with Tesla, Macrohard is described in the S-1 as "an agentic AI platform designed to be capable of fully emulating digital workflows and augmenting human operation of computers — from coding and product development to management and entire business processes — using sophisticated autonomous agents."
This is not a coding assistant. This is a platform designed to run entire companies autonomously. The S-1 states SpaceX believes Macrohard "will have the potential to fundamentally transform how companies are structured and operate, thereby allowing dramatic increases in human productivity."
Macrohard runs on both NVIDIA processors and next-generation Tesla processors — a deliberate product of the SpaceX-Tesla vertical integration strategy that also includes Terafab, a joint chip manufacturing facility with Tesla and Intel targeting one terawatt of compute output annually.
The broader picture from the S-1 is a company that has quietly assembled the full physical stack of AI: chips (Terafab), data centers (Colossus I and II, 1 gigawatt combined), models (Grok 4 and 5), distribution (X, 550 million MAUs), applications (Macrohard), and launch infrastructure to move it all into orbit by 2028.
The Anthropic contract — $1.25 billion per month through May 2029 for Colossus compute access — confirms the infrastructure is already monetized at scale while SpaceX builds toward its orbital compute vision.
Source: SpaceX S-1 Registration Statement, SEC EDGAR, May 20, 2026.
Read the S-1 →