The Asset Changed Faster Than the Institution

Why venture capital’s operating model is not yet built for deep tech

Witte, M. K. (2026), DOI: 10.5281/zenodo.21840912

Artificial intelligence is reducing the capital, time and team size required to build and test many software ventures. At the same time, venture capital is moving toward hardware-based deep tech, where scientific validation, engineering, manufacturing, certification and scale-up remain irreducibly capital-intensive.

This article examines the institutional contradiction created by that shift.

It argues that venture capital has changed its sector focus faster than it has changed the infrastructure through which investment decisions are made. Evaluation methods developed around software economics are now being applied to science and hardware ventures whose risks emerge differently, mature asynchronously and cannot be adequately assessed through traction, market size and founder narratives alone.

The article develops a structural account of:

• how AI is compressing the capital needs of conventional software ventures
• why deep tech provides attractive new capital-absorption capacity
• where software-derived investment heuristics fail in hardware-based ventures
• how scientific, engineering, manufacturing, regulatory and financing risks interact
• why capital may flow toward the most legible ventures rather than the structurally strongest ones
• what institutional infrastructure must change for deep-tech investment to become genuinely decision-capable

Rather than treating the problem as a simple lack of technical expertise, the paper positions it as an institutional design failure: the venture asset has changed, but the operating model used to evaluate and finance it has not changed at the same speed.

Written for venture capital funds, corporate venture teams, innovation institutions, venture studios, accelerators, public investors and deep-tech ecosystem builders.