Ask HN: How should I price an AI infrastructure platform?

I'm building an event-driven infrastructure layer for AI agents. I may have my first customer: a consulting firm wants to build AI automation solutions for its clients using my platform and license it as part of those solutions.

For example, they might sell a $20k solution to a client, with my infrastructure powering the agent system.

I'm considering a flat platform fee, usage-based pricing, or a platform fee + revenue share.

For those who've sold developer infrastructure through consulting/integration partners, how would you structure the first deal?

I'm especially interested in what you'd do differently in hindsight.

2 points | by aqibkhan026 3 hours ago

1 comments

  • deadcatfound 2 hours ago
    For the first deal, I’d use a flat platform fee with a clear usage allowance. Revenue share gets messy before you know support and compute costs.
    [-]
    • aqibkhan026 1 hour ago
      That's helpful. One wrinkle I'm running into: most of the underlying code is open source, and I've built a minimal platform/MVP on top of it that ships with the open-source library.

      So I'm debating between:

      1. A self-hosted version where the customer uses their own compute, and we license the platform layer. 2. Hosting it myself and charging for the managed service.

      For a first enterprise/consulting customer, would you lean toward self-hosted or hosted?

      Also, on the support side, what terms would you look out for? I'm especially concerned about accidentally committing to unlimited support/custom development or infrastructure costs before I know the economics.