Capex from the five largest big-tech companies on data centres passed $400 billion in 2025, with a further +75% expected in 2026 (IEA) — an unprecedented pace of investment in the history of digital infrastructure. But anyone who allocates real capital knows a data centre, on its own, is only the last link in a much longer, much more capital-intensive chain. Capital is already following that chain upstream — into semiconductors, robotics, autonomous systems. That's where the next phase is decided.
The clearest signal comes from semiconductors, not software. In March 2025 TSMC announced an additional $165 billion investment in the United States — three new fabs, two advanced-packaging facilities, one major R&D centre — described as the single largest foreign direct investment commitment in American industrial history. It isn't an isolated announcement: PwC estimates global investment in new semiconductor fabs will reach $1.5 trillion between 2024 and 2030, more than the previous two decades combined; TSMC alone has guided 2026 capex to $52-56 billion, up from $40.9 billion actually spent in 2025. This is capital that becomes concrete, steel, permitting, EPC contracts, supply chain — exactly the kind of capital-intensive programme that requires institutional-grade governance, not just conviction in the thesis.
The same shift shows up in robotics and autonomous systems: $27.6 billion flowed into robotics and "physical AI" venture capital in 2025, across more than 1,000 deals (PitchBook) — with autonomous drones and defence/security systems the single largest segment. Further out on the same spectrum sits quantum computing: investment in quantum-tech start-ups rose from $2 billion in 2024 to $12.6 billion in 2025, 6.3 times as much in a single year (McKinsey), with private capital overtaking public funding for the first time. AI, in other words, is ceasing to be purely a software story: it's becoming a physical-asset story — factories, hardware, machines that move and compute in the real world.
Read through the discipline of the CGE Alignment Index — Capital, Governance, Execution — the point is clear: the further capital moves into hardware and physical infrastructure, the further risk shifts from financial to technical and executive. A $165 billion fab programme isn't governed like a software venture round: it needs clear decision rights, real engineering oversight on the ground and across the supply chain, boards with genuine authority, institutional-grade reporting — the same discipline I've applied to a roughly €1 billion European industrial programme, 15 plants, circular economy and energy transition. The AI thesis is now consensus, in every boardroom. The return, as always in capital-intensive real assets, will be made in execution — not in the thesis.