AI’s new economic reality: Why tokens are reshaping enterprise technologyAs artificial intelligence moves from experimentation into everyday workflows, those traditional frameworks are breaking down. A new financial reality is emerging, defined by tokens.Stu ScotisTokens are at the centre of the AI-based economyGift this article4 min read4 hours agoAI is rapidly becoming one of the most significant and unpredictable costs in modern business. For decades, companies could forecast their technology budgets through familiar models like per-user licensing, predictable subscription tiers and cyclical hardware refreshes. But as artificial intelligence moves from experimentation into everyday workflows, those traditional frameworks are breaking down. A new financial reality is emerging, defined by tokens.In recent analysis from Deloitte, global AI leaders warn that enterprise AI spending is accelerating faster than most organisations had contemplated. This is both in terms of initial build cost but also the ongoing running driven by deeper usage, more complex workloads and the shift toward agentic systems that consume exponentially more consumption of Tokens – the unit measure of AI. Gartner predicts global AI investment will hit US$2.5 trillion in 2026. That means the urgency to understand the economics behind this surge AI tokens consumption has never been greater.The token: AI’s new unit of economic valueAt the centre of this transformation is the token. A token is the basic unit of data an AI model processes when generating text, analysing images or reasoning across tasks. Every prompt, workflow and agent interaction consumes tokens, each carrying a measurable cost that links technology choices directly to business value. Token consumption can fluctuate based on model size, prompt design, depth of reasoning, software, demand and the infrastructure they run on. As a simple example, even a small change in prompt design or agent workflow can consume drastically more tokens creating big differences in cost. Multiply this across hundreds or thousands of agents running that workflow every day and the costs scale with it.The paradox of falling costs and rising spend AI follows a version of Jevons’ paradox: as models become more efficient, organisations use them more so total spending rises. Even though the average cost of inference may fall significantly in coming years, deeper reasoning, multi-agent workflows and expanding adoption drive higher token use. Stu Scotis is Global Agentic AI Leader at DeloitteDeloitte’s analysis shows that while the cost per token declines with hardware improvements, overall enterprise AI investment continues to climb as usage spreads across the organisation. More employees, more use-cases and more sophisticated applications all contribute to higher consumption. Three buying models, three different cost curvesEnterprises today source AI through three main channels, and each has its own financial implications:1.Packaged AI (traditional SaaS) – predictable pricing but opaque token consumption, often masking inefficiencies under simple subscription models. 2.API-based AI – transparent token pricing gives clarity but introduces volatility, with real-time fluctuations tied to usage. 3.Self-hosted “AI Factories” – capital-intensive to build, but the most economical at scale, giving organisations full control over GPUs, storage, security and token efficiency. As token consumption scales the choice is significant. At small volumes (around 10 billion tokens annually), APIs remain more cost-effective, but by 300 billion tokens, AI factories rapidly become cheaper. At full maturity, when workloads reach the trillion-token range, cumulative API costs can exceed AI factory costs by a factor of four. Governing AI as an economic system The shift to tokens demands a new governance model. AI can no longer be treated purely as a technology deployment; it must be managed as an economic system with real financial consequences. Organisations that excel in AI governance treat tokens the way manufacturers treat energy or capital. They track them, forecast them and link their consumption directly to business outcomes. This approach is reinforced by emerging AI FinOps practices, which extend financial accountability across GPU hours, token usage and workload performance. These disciplines give leaders real-time visibility, enable chargebacks to business units, and create guardrails that prevent silent overspend.The hybrid future of Enterprise AI Deloitte’s research indicates most organisations will adopt a hybrid architecture — conducting early experimentation on APIs, moving high-throughput workloads to owned infrastructure, and maintaining flexibility across general cloud and specialist GPU cloud providers.This model balances agility with cost control, reducing vendor lock-in while optimising for performance and sovereignty.Leadership imperatives for the AI era To navigate AI’s changing economics, leaders should govern AI as an economic system, not just an IT investment. They should understand workload scale to know when to shift away from APIs and right-size models, saving frontier models for tasks that truly need them. At the same time, they should embed FinOps and token governance across every team and architect hybrid by default to manage volatility and maximise value. AI is reshaping enterprise economics. Tokens, not licences, should be core in every technology strategy. Organisations that master token governance will innovate faster, scale more confidently and create durable competitive advantage. Those that don’t risk watching their competitors accelerate with runaway benefits while they struggle to realise value from their own AI investments.Stu Scotis is Global Agentic AI Leader at Deloitte.  -DisclaimerThis publication contains general information only and Deloitte is not, by means of this publication, rendering accounting, business, financial, investment, legal, tax, or other professional advice or services. This publication is not a substitute for such professional advice or services, nor should it be used as a basis for any decision or action that may affect your business. Before making any decision or taking any action that may affect your business, you should consult a qualified professional adviser.  Deloitte shall not be responsible for any loss sustained by any person who relies on this publication. About DeloitteDeloitte refers to one or more of Deloitte Touche Tohmatsu Limited, a UK private company limited by guarantee (“DTTL”), its network of member firms, and their related entities. DTTL and each of its member firms are legally separate and independent entities. Please see www.deloitte.com/au to learn more.Copyright © 2025 Deloitte Development LLC. All rights reserved. -More CoverageTalent ramps: Invest in new pathways or face a workforce cliffARTIE GINDIDIS AND AMANDA FLOUCHAI goes physical: navigating the convergence of AI, roboticsJim Rowan, Tim Gaus, Franz Gilbert and Caroline BrownDeloitte leaders on the tech shifts that will define 2026Deloitte Team