Hyperscalers are pouring hundreds of billions into AI this year. Mid-market companies are right behind them, spending hundreds of thousands to millions on their own AI initiatives. Different scale, same accounting headache.

Adam Olsen and Nicole Harger break down how to actually account for AI development costs, and why the playbook finance teams have used for traditional software doesn't map cleanly onto this technology.

In this episode:
  • Why framework selection comes first: internal-use software (ASC 350-40), externally marketed products (ASC 985-20), and pure research (ASC 730) all carry different capitalization rules
  • The three-stage model under ASC 350-40 and why AI's iterative development cycle makes stage-tracking harder than it looks
  • Data costs: the most overlooked, most material line item, and the "alternative future use" judgment that determines its treatment
  • ASU 2025-06: FASB's principles-based overhaul of internal-use software accounting, and the new "significant development uncertainty" concept that changes when AI projects can start capitalizing
  • What this all means for budgeting, useful life assumptions, build-vs-buy decisions, and the auditor conversation