Enterprises across the United States are grappling with a rapidly expanding AI token economy. After Stripe announced its planned acquisition of OpenRouter in August, the deal highlighted how tokens—once a purely technical metric—are becoming a core resource that companies must allocate, finance, and manage.
Token Basics and Enterprise Growth
Tokens are the units of text that AI models process and generate when completing a task. While many consumer‑facing AI services still rely on monthly subscriptions, most enterprise and API‑based offerings charge by the number of tokens used. The rise of AI agents has dramatically amplified token demand. According to cybersecurity firm Human Security, agentic traffic grew a staggering 7,850% in 2025.
Cisco found that an AI agent performing a research task generated 450% more network traffic than a person doing the same work manually, and all of that traffic is billable in tokens. This surge in usage has forced businesses to confront both the cost and predictability of token consumption.
Falling Token Prices, Rising Bills
Token prices have been on a downward trend. The LLM Token Expenditure Index, compiled by Silicon Data, fell to 97 cents per million tokens at the start of September—its lowest level since the index launched late last year, according to CNBC. The index has more than halved from its summer peak, driven by improved model efficiency and intensified competition among AI providers.
Despite the price drop, enterprise AI bills are still climbing. A McKinsey study found that 93% of enterprises reported AI cost overruns in the past year. The same study showed that overall enterprise LLM spending tripled within a 12‑month period, even as inference costs fell sharply. Goldman Sachs projects token consumption will increase 24‑fold by 2030.
Predictability Challenges
Accurately forecasting token usage remains difficult. A 2026 study that included Stanford researcher Erik Brynjolfsson tested eight frontier models on software‑engineering tasks and discovered that the models consistently underestimated their own token consumption. The study also noted that identical tasks could consume tokens at rates up to 30 times different, highlighting the volatility of token demand.
Because of this uncertainty, many businesses set budgets well below actual needs. OpenAI recently reported that its heaviest users of AI coding agents were consuming more than $7,000 worth of tokens per day. In response, AI infrastructure providers such as Google Cloud have introduced spending caps that automatically halt eligible API traffic when limits are reached.
Beyond Simple Caps: Emerging AI FinOps
While caps can prevent overspending, they do not guarantee that money is spent wisely. KPMG found that only 15% of business leaders have formal metrics for measuring returns on AI investments. To close this gap, companies are turning to AI FinOps—a discipline that integrates spend, usage, performance, and business outcomes.
McKinsey’s May 2026 Enterprise AI FinOps survey indicated that just 20%‑25% of firms have mature AI FinOps capabilities. Recognizing the need for standardized tools, the Linux Foundation launched the Tokenomics Foundation in August to develop benchmarks and specifications for measuring and managing AI token costs.
J.R. Storment, executive director of the Tokenomics Foundation, emphasized that CEOs are being asked to demonstrate returns without a shared method for counting AI expenses. The foundation aims to create pre‑competitive frameworks that will bring transparency to the emerging market.
Industry Responses
Accenture introduced its own tokenomics offering in July, combining usage monitoring, model routing, budget controls, and tools for aligning AI costs with business outcomes. Stripe’s acquisition of OpenRouter further illustrates the growing interest in infrastructure that can price, route, and finance token consumption.
These initiatives suggest that the token economy, while still in its infancy, is moving toward a more mature market structure similar to commodities like oil or gold. As tokens become a fundamental input for business operations, firms that can effectively price, allocate, and measure them will become essential partners in the AI ecosystem.
Original reporting: KRDO (Colorado Springs metro) — read the source article.
