# Industrial AI Unit Economics The unit economics of industrial AI are governed by one ratio: deployment cost vs. annual contract value. If a deployment costs 320 engineer-hours (~$50-80k fully loaded) and the ACV is $46-60k, you lose money on every customer in year one. The business model depends entirely on two things: 1. ACV expansion from pilot to production (the [[Land-and-Expand in Enterprise AI]] motion) 2. Deployment cost reduction through automation (see [[Deployment Velocity]]) The services trap: each customer adds ~0.2-0.5 FTE of ongoing model maintenance if maintenance is manual. At 5 FTEs, you cap at 10-15 active deployments. Margins stay at 30-40%. This is a $3-5M ARR ceiling that no amount of sales can fix. The scaling equation: - 30-40% of codebase reusable = services company - 60-70% reusable = emerging platform (if deployment hours prove it) - 80%+ reusable = true SaaS (rare in industrial AI) Each new vertical (process manufacturing, airports, pharma) requires a domain hire + 6 months + reference deployment. Scaling across verticals is O(n) in headcount. Scaling within a vertical is where leverage exists. The breakout: if deployment drops below 80 hours AND at least one customer expands from pilot to $500k+ production contract AND model maintenance is autonomous, then software economics kick in. All three must be true. This is why industrial AI is hard. Related: [[Deployment Velocity]], [[Consultancy-to-Platform Transition]], [[Incumbent Bundling Risk]], [[Industrial AI MOC]] --- Tags: #deeptech #investing