The 90% Gross Margin Era Is Over: How AI Startups Should Pitch Their Economics
Konstantine · x · 2026-10-29
AI companies can't sustain the 80-90% gross margins of classic SaaS: every query runs through a foundation model, so COGS is heavy—frontier-model products may see 40%, near-zero, or negative margins. The author argues low margins aren't inherently bad; what matters is whether they're improving.
- Industry tailwind: cost per token of the same intelligence has fallen 98%+ in 18 months, and outcome-based pricing captures more value
- Company level: blended gross margin hides the trend; break it out two ways:
- By customer: customers live a year should show much higher margins than those still in implementation
- By pricing cohort: usage-based customers should out-margin legacy seat-based ones with a widening gap
The pitch: fully implemented customers have strong margins and the new pricing is materially better—far stronger than promising margins will improve.
More from Venture
- Lightspeed raises $750M+ for AI Renaissance Fund to acquire and revamp businesses with AI — KateClarkTweets · 2026-10-29
- Indie dev back on the shipping train: 450 active customers in 28 days, chasing conversions — rudrank · 2026-10-29
- Juan Benet launches PLC Neurotech fund after 4 years of BCI investing, likening moment to 2019 AI — Scobleizer · 2026-10-29
- ENGY claims ~30% cheaper inference via Bittensor with no quality loss on open models — markjeffrey · 2026-10-29
- Cognition reportedly mulls funding round at $70B-80B valuation, per Insider — zephyr_z9 · 2026-10-29
- AI as a selective information shock: Hayek's 1945 insight predicts where AI complements vs substitutes — robleclerc · 2026-10-29