The brief

Analyst view

• Launched Nov 2025 with $6.2B; Series B June 2026 raised $12B at $41B valuation.
• Builds 'artificial general engineer' trained on physics and data, not text. Automates design-to-manufacturing for jets, drugs, chips.
• ~150 people from OpenAI, DeepMind, xAI, Meta. Offices: SF, London, Zurich.

Technical approach

As reported

Physics training

Real data and physical laws

Trained on experimental results and physics equations, not text. Grounds models in real constraints.

Design automation

End-to-end CAD to manufacturing

Optimizes designs by simulating candidates against physics and cost constraints.

Multi-domain

Aerospace, pharma, semiconductors

Trains on diverse domains to learn physical optimization principles across industries.