I like that his followup is truthful about the very real short comings of his solutions. For any real-time system, if there is a hard requirement of 95% accurac

I like that his followup is truthful about the very real short comings of his solutions. For any real-time system, if there is a hard requirement of 95% accuracy, I can understand the initial pricetag. The vendor carries all the risk. Making proof of concept resilient and usable in the real world is not easy nor cheap. His examples of camera placement and image distortion alone are non trivial problems to solve. Maybe they need custom made cameras for them to be fast enough and small enough, with good enough resolution. That said, $86M is a lot of money. If you assemble a team of 5 people. Assume they use $1M a year each in for salary, office, equipment, ect, they can work fulltime on this for 5 years and still have more than $60M left for other stuff.