Larridin Blog

Larridin's CTO Says AI Spend Should Be CapEx. Here Is What His Own Numbers Show.

Written by Larridin | Sep 11, 2026

Larridin CTO Ameya Kanitkar told Business Insider that AI costs should be treated as capital expenditure. In a separate Forbes piece, he calls token spend a variable operating expense and shows how quickly usage and cost mix can change. Those are two different arguments. Here’s what each tells finance.

Key Takeaways

  • The CapEx framing is about how leaders evaluate the investment over time, not an automatic accounting classification.
  • Volatile token usage and model mix make AI spend difficult to forecast regardless of how leaders frame the investment.
  • Better spend visibility helps finance manage that volatility, but the Forbes data doesn’t prove the case for CapEx treatment.

What the Two Statements Actually Mean

In Business Insider, Kanitkar says organizations should evaluate AI as a capital investment and judge returns over a year or two rather than expecting immediate gains.

His Forbes Technology Council article addresses a different problem: token spend behaves as a “variable operating expense” that can be difficult to track and forecast.

The distinction matters because the Forbes data can explain the spending and budgeting problem without establishing the case for CapEx treatment.

What Kanitkar’s Forbes Data Actually Shows

The Forbes article identifies several reasons AI spending can be difficult to forecast.

  • Costs are spread across many sources. Token consumption can come from cloud model providers, AI subscriptions, browser plugins, desktop agents, custom connectors, and API gateways, each with its own billing and reporting.
  • Usage and spend do not move together. Across Larridin’s enterprise customers, Kanitkar reports that Claude Sonnet accounts for 70% of API prompts but 51% of API spend. Claude Opus accounts for 27% of prompts but 48% of spend because each Opus prompt costs three to four times more.
  • Usage can change quickly. Over a 10-week period, active Claude users increased 41% and sessions increased 76% across the enterprise customers in the dataset. That kind of movement can make an annual budget assumption stale quickly.

A finance team looking only at total spend or prompt volume can also miss changes in model mix that materially affect cost.

What the Data Doesn’t Show

The Forbes data does not establish that AI spending should be treated as CapEx.

The article explicitly calls token spend an operating expense. The data shows why that expense is unusually variable and difficult to track. It doesn’t make an accounting argument for reclassifying it.

Kanitkar’s CapEx position comes from his separate Business Insider comments. That argument asks leaders to evaluate AI against a longer investment horizon rather than judging it only by short-term cost.

Those ideas can be discussed together, but they should not be treated as the same claim. The volatility data explains the measurement and budgeting problem. It does not, by itself, prove the CapEx case.

For the practical argument for a longer-term investment framing, see AI Costs as CapEx, Not OpEx: The Framing Shift That Changes the Budget Conversation.

Why Measurement Matters Under Either Framing

Whether finance ultimately treats a specific AI cost as an operating expense or a capital investment, leaders still need to know where the money is going.

Larridin’s Token Spend & Insights consolidates AI spend across tools, models, agents, and teams and connects that spend with the work and outcomes behind it.

That gives finance a clearer answer to the underlying questions Kanitkar raises in both pieces: What’s driving the cost? How quickly is it changing? Which models and teams account for it? And what is the organization getting in return?

How AI spend is classified doesn’t tell finance what’s driving it or what the organization is getting in return. That requires measurement.

Frequently Asked Questions

How can finance manage AI spend when usage changes quickly?

Track spend by model, tool, team, and usage over time rather than relying on a single annual estimate. That makes changes in model mix and consumption easier to spot before they turn into unexplained budget variance.

Doesn’t Kanitkar’s own Forbes article call AI spend an operating expense?

Yes. The Forbes article explicitly describes AI token spend as a variable operating expense. His separate Business Insider comments make a broader case for evaluating AI as a longer-term capital investment. The two statements address different questions.

Does treating AI spend as CapEx have accounting implications?

Yes. Formal treatment depends on the type of investment, applicable accounting standards, and jurisdiction. Organizations should confirm the appropriate classification with their accounting advisors rather than treating the strategic framing as an accounting determination.

Track the Spend Before You Debate the Framing

Larridin’s Token Spend & Insights shows where AI spend is coming from, how it changes over time, and which teams, tools, and agents are driving it.

Book a discovery call to see what is driving your AI spend.