Angel investor Alexander Kardos‑Nyheim sold his AI research firm Safe Sign Technologies to Thomson Reuters in 2024, marking the legal‑data giant’s first pre‑revenue acquisition in its 170‑year history. The Cambridge‑based startup built a legal‑reasoning model that matched leading labs while spending a fraction of their budget. Kardos‑Nyheim argues the deal proved that scientific edge, not product traction, can drive valuation. It showed appetite for AI.

In Q1 2026 foundational‑AI startups collectively raised $178 billion, but 97% gravitated to incumbents OpenAI, Anthropic and xAI. Kardos‑Nyheim warns emerging founders that chasing these giants limits upside, because application‑layer firms depend on pricing and access set by upstream owners. He bets lasting value lies in the underlying model, architecture and inference efficiency that remain unsolved and reshapes future funding.

Kardos‑Nyheim’s checklist for investors focuses on team depth—science talent on par with DeepMind—and whether a problem sits at the model layer rather than a superficial add‑on. He advises founders to prioritize durable, safety‑critical breakthroughs over short‑term revenue signals. The market’s slow capital cycle will eventually surface these deep‑tech winners, rewarding those who built the stack first and secure long‑term relevance.