Tokenomics: The Hidden Economics of AI Adoption

中文标题

代币经济学：AI 采用的隐性经济学

Nomura Connects09/17/2026updated 10/17/2026

Unknown economy

Citable research brief

Core conclusion · Tlines extraction

Global corporate AI investment reached $582bn, with generative AI alone at $110bn in trailing 12-month revenue, yet only 13% of organizations report significant enterprise-level value and 40% of projects miss targets. Nomura's AI head argues tokens behave like volatile input costs, effective cost per task stays flat or rises despite falling unit prices, and discipline in token governance now matters more than deployment speed.

Institution: Nomura Connects
Published: 09/17/2026
Research topic: Not explicitly stated
Time horizon: 2025

Key arguments

Token consumption is non-linear: each 10% price cut drives 12-18% more tokens consumed, so total spend keeps rising even as unit prices fall.
Effective cost per task stays flat or rises because tokens consumed per task surge and agentic workflows multiply consumption exponentially.
Over 85% of query cost is hidden: a single enterprise query can burn 6,000+ tokens in system prompts and retrieved documents before the user question is processed.
Rightsizing models is the practical hedge: compact models on dedicated hardware cost about $50/day per 100M tokens versus $1,560/day on a frontier API, a 32x gap.

Key numbers

Global corporate AI investment: $582bn
GenAI trailing 12-month revenue: $110bn
Organizations reporting significant GenAI value: 13%
Projects failing to meet planned results: 40%
Daily cost per 100M tokens, small model on dedicated hardware vs frontier API: $50 vs $1,560 (32x)

Main risks

Organizations built on artificially low token prices face a significant adjustment as the 2023-2025 pricing war corrects.
Frontier deployment is increasingly constrained by cost, capacity and marginal returns, shifting the focus from model capability to input price and scarcity.
Self-hosted inference offers potentially over 50% cost savings versus APIs over three years, but only for high-volume, predictable workloads where economics are compelling.

Corresponding source evidence

Kieran White International Head, AI Center of Excellence Massive spending: AI investment surge, yet only 13% of organizations report signi ficant enterprise-level value from their GenAI investments and 40% of projects are failing to meet goals.
40% of implemented projects fail to meet planned results .
11 12 8, 13 7 11 What does this mean for the adoption race Controlling token economics Organizations should manage tokens like capital— tracking return on intelligence for each AI project and allocating spend to opportunities with the highest returns .

Context: the conclusion is Tlines' automated extraction, not the institution's wording. Quoted evidence is publisher wording. Scope and date must travel with the conclusion.

Verify at the original source ↗

Download institution PDF