← @bgSponge

Something Is Changing in the Unit Economics of Software

Summary

Nicolo argues that AI removes software’s historical near-zero marginal cost by attaching an inference bill to each user interaction.

  1. Inference makes usage a direct cost.

    Each model call adds compute expense to the core experience, weakening software’s historical ability to serve another customer for almost nothing.

    AI makes them large, sticky, and inseparable from the core experience. The bill of materials finally matters.

  2. Model choice couples quality and margin.

    Cheaper models protect gross margin while more capable models can improve the product at a higher per-interaction cost.

    For the first time, margin and quality are in direct conflict on a fundamental per-unit basis.

    The pricing model is changing because the cost model changed first.

  3. Flat pricing hides unequal consumption.

    When power users generate far more inference work than casual users, usage-based pricing better aligns revenue with cost to serve.

    The cost per call drops but the calls per user multiply.

  4. Efficiency may increase total demand.

    Falling inference prices can fund deeper integration, more calls, and background agents instead of becoming retained margin.

  5. The SaaS growth assumption no longer holds automatically.

    Founders need viable unit economics before scale because additional customers may carry persistent costs that growth does not erase.

Read the original ↗Save to your Sponge

Summarized with Sponge. Quotes are the author’s own words.

More from @bg

All readings →