AI is becoming one of the largest capital and operating spend categories in many companies. The CFO's job is to allocate it like any scarce capital: to the highest risk-adjusted returns, with discipline to stop what doesn't work.
The scale of AI investment
Microsoft said in January 2025 that it was on track to invest approximately $80 billion in fiscal 2025 to build AI-enabled datacenters. Few companies spend at that scale, but the same questions apply at any size.
A portfolio approach
- Run (productivity in existing processes): lower risk, faster payback.
- Grow (better decisions, new revenue, customer experience): higher value, longer payback.
- Transform (new operating or business models): highest uncertainty, staged funding.
Fund each bucket against its own hurdle rates and stage gates, and rebalance quarterly.
Stage-gated funding
- Discovery: small budget, prove the use case and data.
- Pilot: measured against a baseline, with kill criteria.
- Scale: funded on pilot evidence, with value tracking.
- Run: absorbed into the business budget with ongoing unit-cost targets.
Evidence of return
JPMorgan Chase CEO Jamie Dimon said in October 2025 that the bank spends about $2 billion a year on AI and saves at least about the same amount. Company-reported, but it shows the comparison boards expect: AI spend against measured benefit.
Build, buy or partner, and central versus distributed budgets
These are allocation decisions too; see the decision library. The common pattern: central funding for platforms, data and governance; business-unit funding for use cases, with central standards.