Defining Physical World AI – and Transition Ventures’ €128 Million Bet
Physical world AI refers to artificial intelligence systems that directly sense, control, or transform real-world infrastructure, machinery, and environments, combining software intelligence with hardware, robotics, and industrial processes to improve how energy, logistics, manufacturing, and safety-critical operations work in practice. This is the space where Transition Ventures has chosen to concentrate. The London-based early stage investment firm has closed a €128 million (approx. RM652 million) Fund II focused on early-stage companies at the intersection of AI and the physical world, bringing its assets under management to over €257 million (approx. RM1.31 billion). According to EU-Startups, the firm’s team has founded companies worth over €12 billion (approx. RM61.2 billion) across software, hardware, and DeepTech. For founders, this is more than another AI funding round: it is a targeted pool of capital for companies turning algorithms into physical outcomes.
From Software-Only AI to Infrastructure, Robotics and Energy Systems
Transition Ventures’ strategy reflects a wider shift in AI funding rounds from pure software toward AI infrastructure automation, robotics, and industrial applications. The firm invests from inception to Series A in businesses working on energy systems for AI workloads, robotics for industrial efficiency, and next-generation solutions for critical minerals refining. The thesis is that the most important companies in the coming decades will replace legacy physical systems with cleaner, more efficient AI-enabled alternatives. Portfolio companies show what this looks like in practice. Olix is developing photonics-based computing hardware that aims to outperform traditional semiconductors and has reached a valuation of €858 million (approx. RM4.38 billion) after raising over €188 million (approx. RM960 million). Applied Atomics is building small modular nuclear power plants designed to deliver reliable, clean electricity to data centers and large industrial users, tying AI demand directly to new physical infrastructure.
Why Early-Stage Investors See Opportunity in Physical World AI
Transition Ventures argues that “the classic VC model of backing more of the same, incremental improvements, has run out of road.” Instead, Fund II focuses on early stage investment into physical world AI where problems are hard, capital needs are higher, and competitive moats are built around hardware, regulatory insight, and long-term customer contracts. For institutional investors, this promises more durable differentiation than yet another software-only AI tool. The team’s background underscores this ambition: founders and partners from Unity, Index Ventures, Atomico, Balderton and others are targeting companies that can matter “for generations to come.” Their investments stretch from full-stack physical AI for wildfire suppression, via Seneca’s autonomous drone systems, to platforms such as Upway, which has grown revenue more than 30x since Transition’s initial backing by focusing on refurbishment and reuse of physical assets.
A Wave of Specialist Funds and What It Means for Startups
Transition Ventures’ Fund II sits within a broader wave of specialist funds raising capital for DeepTech, climate, AI, energy and industrial systems. EU-Startups reports that more than €1 billion (approx. RM5.1 billion) has been disclosed across similar or adjacent vehicles in 2026, including funds from 2150, Lansdowne Partners, Eka Ventures, Passion Capital, The Footprint Firm, Montis VC, 360 Capital and Vanagon Ventures. Together, they signal strong investor confidence that AI tied to physical infrastructure will shape the next generation of large companies. For founders, this has two implications. First, capital is becoming more available for AI infrastructure automation, robotics, and industrial applications that once struggled to attract early financing. Second, specialist investors now bring technical, regulatory and go-to-market expertise that generalist funds often lack, raising the bar for quality but also improving the odds of success for credible physical world AI startups.