AI Margins Are Lower Than SaaS: FAL CTO's Hard Truth

The Gist

• AI gross margins are lower than SaaS and unlikely to improve due to rising inference costs
• New AI models (e.g., video) are more expensive to run, offsetting cost reductions in older models
• High-usage AI customers can become unprofitable, unlike in traditional SaaS
• Pricing must reflect real cost to serve from day one to avoid margin erosion

• AI margins are lower than SaaS because every new user or query costs real money to serve, unlike SaaS where marginal costs approach zero.
• As AI models improve, costs don't decrease; they increase, contrary to initial assumptions that hardware improvements would lower costs.
• High-usage AI customers can be high-cost customers, creating a tension between usage and margin that doesn't exist in SaaS.
• FAL tracks wallet share to understand and grow the right accounts, shifting focus from mere account expansion to strategic growth.
• FAL's unconventional hiring process for researchers involves open invitations and paid auditions through research grants, leading to successful hires.
• Positioning as a 'generative media platform' has helped FAL stand out in a crowded market by focusing on a specific niche.

• Build pricing models that reflect the real cost to serve from day one, as AI margins won't improve by waiting.
• Track wallet share to strategically grow high-value accounts rather than just expanding usage.
• Consider shorter-term quotas (monthly or quarterly) in high-growth environments to allow for course correction.
• Adopt unconventional hiring practices, like research grants, to identify top talent through practical demonstrations of skill.

• 80-90% (Gross margins in traditional SaaS due to near-zero marginal costs.)
• $8B (Valuation of FAL, a generative media platform.)
• 50% (Growth of FAL's annual target during the interview process for a head of sales.)
• 4 (Number of people hired through FAL's research grants program.)