Larridin, an AI spend analytics startup backed by a16z, published its first benchmark in August 2026 showing the median engineer incurs $213 in billed AI coding costs weekly, equating to about $920 monthly. The data sample covered engineers who merged code and had billed AI-coding spend in the four weeks ending August 2, 2026. At the 90th percentile, costs rise to $911 weekly, indicating a more than 10-fold spread in AI token consumption per engineer, according to saastr.com.
The benchmark was built from production billing and engineering telemetry data, capturing real-world AI usage across engineering teams. Larridin’s platform connects AI spending to individual engineers and AI agents, providing insights into adoption, fluency, and automation opportunities. The startup’s founder is on his sixth company, and the product aims to clarify the often opaque AI spend line items for B2B companies investing in ChatGPT, Claude, and autonomous coding agents.
This data is significant as AI coding costs become a growing budget item for enterprises. With median engineers billing nearly $1,000 monthly in AI tokens and top users approaching $47,000 annually, companies face a wide variance in AI consumption. Larridin’s insights help businesses understand where AI spend delivers value and where it may plateau, addressing a critical need in the SaaS sector as AI integration deepens across engineering workflows.
Larridin’s benchmark report, published in August 2026, offers the most detailed view to date on AI coding token costs per engineer, highlighting the financial impact of AI adoption in software development teams.