Two figures drew the attention of enterprise finance teams in the first half of 2026. Uber reportedly exhausted its annual budget for agentic AI coding tools within four months, while Meta employees consumed 73.7 trillion AI tokens in just over 30 days, according to reporting based on internal company information.*
These cases do not prove every organization faces an immediate token cost crisis. They do, however, highlight a growing structural challenge. Usage-based AI costs can expand far more quickly than conventional software budgets, particularly when adoption incentives, financial accountability and measures of business value are misaligned.
Token based pricing breaks the way enterprises budget
Token-based consumption pricing does not behave like the software line items CFOs know how to model. The gap between engineering consumption and finance expectations is no longer hypothetical. Organizations have spent the past two years encouraging employees to adopt AI tools, often without recognising every prompt, agent loop and parallel task ultimately contributes to a real invoice.
According to reports based on comments by Uber CTO Praveen Neppalli Naga, Claude Code was introduced to Uber’s engineering organization in December 2025. Adoption reportedly climbed from 32 per cent of engineers in February to 84 per cent classified as agentic coding users by March, with approximately 95 per cent of engineers using AI tools monthly during the rollout. Uber also reported significant productivity gains from the programme. As adoption accelerated, AI consumption expanded much faster than the company’s allocated budget, illustrating how quickly usage-based pricing can outpace traditional budgeting models.
Meta experienced a similar challenge. According to reporting based on internal company communications, employees consumed 73.7 trillion AI tokens in just over 30 days*. Internal reporting also described an employee-created leaderboard known as “Claudeonomics”, which appears to have encouraged some staff to maximize visible token consumption rather than demonstrate measurable business outcomes.
The company’s CTO, Andrew Bosworth, subsequently reminded employees “nobody should be using AI tools just for the sake of using them” and “token usage alone is not a measure of impact of any kind.” The warning followed Meta making AI use a core expectation within employee performance reviews.
The industry is starting to build the governance infrastructure
The market has recognised this is not a one-company issue. The Linux Foundation recently announced its intention to launch the Tokenomics Foundation, a new initiative focused on establishing open industry standards, benchmarks and best practices for AI cost management, operating in partnership with the FinOps Foundation.
As J.R. Storment, Executive Director of the FinOps Foundation, observed, “Token costs and efficiency have become a CEO-level concern, not an engineering footnote.”
At the same time, Goldman Sachs Research forecasts global AI token usage will increase 24-fold between 2026 and 2030, reaching approximately 120 quadrillion tokens per month. As enterprises move generative and agentic AI workloads from pilot programmes into production, governance disciplines have struggled to keep pace with consumption growth. AI is becoming one of the fastest growing areas of enterprise technology expenditure, yet many organizations still lack mature approaches for measuring and governing spend.
Governance is a people and culture question, not just a finance one
The Uber and Meta examples share a common feature. AI adoption accelerated faster than the organisations’ budgeting processes, governance structures and accountability mechanisms. Teams were encouraged to use AI extensively, while financial ownership and visibility often remained elsewhere within the organization.
This is precisely what an outcome-led approach to AI adoption is designed to prevent. Before deploying agentic tools broadly, organizations need clear answers to fundamental questions.
* Which tasks justify agent-level token spend?
* Who owns the AI budget?
* What measurable business outcome justifies the ongoing cost?
Available research suggests governance continues to lag adoption, although estimates vary between studies. A PEX Network survey of more than 200 professionals found 43 per cent of respondents said their organization had an AI governance policy. Separately, Deloitte reported only 21 per cent of surveyed organizations had a mature governance model for agentic AI. While these studies measure different populations, both point to a meaningful gap between AI deployment and organisational oversight.
Meta is reportedly responding by introducing a centralized AI Gateway to improve visibility into AI usage and spending across teams, with formal token budgets expected from 2027. The move illustrates how governance capabilities often emerge only after rapid adoption exposes weaknesses in existing financial controls.
Access without governance helps no one, including nonprofits
While enterprises grapple with AI spending, another conversation is taking place around equitable access to AI technologies.
The OpenAI Foundation has committed an additional US$50 million in 2026 through its People-First AI Fund, building on the programme launched in 2025. Applications for the current funding round close on 15 July 2026 and are designed to support eligible nonprofit organizations using AI to strengthen their communities.
The initiative demonstrates expanding access remains an important priority. It also reinforces a broader lesson. Whether an organization is a global enterprise or a nonprofit, sustainable AI adoption depends on the same fundamentals: clear purpose, appropriate oversight and accountability for ongoing AI expenditure.
What adopting AI with confidence actually requires
The examples above are not arguments against AI adoption. Uber has reported meaningful engineering productivity improvements, while Meta continues to invest heavily in AI-enabled ways of working. Rather, they illustrate the importance of sequencing.
Organizations deploying AI tools before establishing token governance, model routing strategies and clear links between AI consumption and measurable business outcomes risk finding themselves in an uncomfortable position. They may either slow adoption to regain financial control or continue investing without demonstrating proportional business value, neither of which is an easy boardroom conversation.
At Insentra, our AI Momentum approach puts those foundations in place first through clear ownership of AI expenditure, governance policies connecting AI usage to measurable business outcomes, and organisational change programmes ensuring incentives do not outpace financial controls.
If your organization is scaling AI tools and the token bill is already surprising people, it may be time to assess whether the right governance foundations are in place before your next rollout.
Contact us about where your AI governance stands and what needs to be in place before your next rollout.
*Source: Tokenminimizing: Meta Moves to Curb Employee AI Usage as AI Costs Reach Billions
