Uber exhausted its entire 2026 AI budget in four months. Employees at Disney logged 460,000 Claude interactions in nine days. Across the enterprise landscape, AI adoption leaderboards designed to drive usage have created a new problem: tokenmaxxing—gaming AI metrics without measuring business value.
Now, a counter-movement is emerging. CFOs and CIOs who championed unlimited AI access are pulling back, capping usage, and demanding a new metric: valuemaxxing. The shift signals the end of the AI adoption spending spree and the start of financial accountability.
The Budget Shock
In June 2026, Bloomberg reported that Uber burned through its entire annual AI budget by April—four months into the fiscal year. The culprit: widespread use of AI coding assistants like Claude Code and GitHub Copilot, with no visibility into cumulative costs until the bill arrived.
Uber executives responded by capping AI tool usage mid-year, a dramatic reversal from the "deploy everywhere" strategy that dominated 2025.
Uber isn't alone. At Disney, one employee interacted with Claude AI 460,000 times in a nine-day period, according to Business Insider. The company had deployed an AI adoption dashboard to track usage, inadvertently creating a leaderboard that encouraged employees to maximize token consumption rather than business outcomes.
This is tokenmaxxing: treating AI adoption as a vanity metric—lines of AI-generated code, number of prompts, or tool interactions—divorced from whether those interactions created value.
"The metric for success is not if I have 99% of my organization using Microsoft Copilot," says Becky Trevino, Chief Product Officer at FinOps vendor Flexera. "It's now, show me the money. Show me what's changing in your organization."
Why AI Budgets Are Exploding
Three structural forces are colliding:
1. AI Is Underpriced (For Now)
"AI is underpriced right now, and that's masking the real cost," says Cliff Jurkiewicz, VP of Global Strategy at Phenom. "Behind every prompt is expensive infrastructure—data centers and energy—and providers aren't absorbing that forever."
Anthropic and OpenAI are both moving toward IPOs, which will force pricing discipline. Early 2026 saw both vendors roll out metered pricing models, replacing flat seat fees with usage-based billing. That shift makes costs variable and unpredictable.
2. Model Costs Vary 100x
"A lot of companies are turning on AI features everywhere, riding the productivity wave until they are blindsided by a huge bill," says Andy Sen, CTO of AppDirect.
The cost difference between AI models can be 100x. Premium models deliver answers in 5 seconds instead of 15 seconds, but for most enterprise work—writing emails, summarizing documents—that speed premium isn't worth the cost.
Many companies let employees choose any model without visibility into the financial impact. A team using GPT-4o for every task instead of GPT-4o-mini can blow through budgets 20x faster.
3. No Guardrails on Usage
"There's very little discipline around what actually drives business value versus what's just convenience," Jurkiewicz says. "Every interaction has a cost—input and output—and these systems tend to overdeliver, which means you're paying for more than you need. Multiply that across an entire workforce, and costs scale fast."
Adoption leaderboards incentivized volume, not value. Employees gamed the system, generating tokens to rank higher on internal dashboards. The result: runaway spend with no business justification.
From Tokenmaxxing to Valuemaxxing
In response, a new FinOps discipline is emerging: valuemaxxing.
Valuemaxxing combines three elements:
- Visibility: Real-time tracking of AI spending across models, teams, and use cases
- Governance: Guardrails that limit premium model access to high-value work
- Financial Accountability: Tying AI costs to business outcomes (revenue, EBITDA, tickets resolved)
Flexera, which provides FinOps platforms, reports that AI budget surprises have become "a serious talking point among customers." The companies that win will treat AI like a finite resource, not an unlimited utility.
The Linux Foundation Steps In
In early June 2026, the Linux Foundation announced the Tokenomics Foundation—a new initiative to establish open industry standards, benchmarks, and best practices for AI cost management.
The foundation's goal: create transparency around the economics of AI infrastructure so enterprises can compare costs across vendors and optimize spending.
This is a structural shift. When AWS launched in 2006, it took years for FinOps practices to mature. AI is moving faster—enterprises are demanding cost discipline within 18 months of mass adoption.
What CFOs Need to Know
Budget Reality Check
If your organization deployed AI tools in 2025 with flat pricing, 2026 is the year metered billing arrives. Token costs are now variable, and without governance, a single team can blow through annual budgets in a quarter.
Action: Audit AI spending monthly. Track cost per team, per use case, per model. If you don't have visibility, you don't have control.
Tokenmaxxing Is a Red Flag
If your internal dashboards celebrate AI adoption volume (tokens generated, prompts submitted, tool usage rates), you're incentivizing the wrong behavior.
Disney's 460,000-interaction employee wasn't creating 460,000 units of value. They were gaming a metric. That's waste.
Action: Replace volume metrics with outcome metrics. Did AI reduce ticket resolution time? Increase qualified leads? Cut manual review hours? If you can't tie AI usage to a business outcome, it's not value—it's spend.
Valuemaxxing Frameworks
Trevino recommends prioritizing AI budget allocation by department:
"If you have a limited amount of budget and you have 20 departments, which of those departments matter the most? Allocate most of your AI budget there. If three of those 20 are the ones you need to drive growth or competitiveness, ensure those three business units have the access they need."
This is AI portfolio management. Not every department needs premium models. Not every use case justifies unlimited access.
What CTOs Need to Know
Model Cost Governance
Sen's advice: research the cost of individual models and encourage employees to use the most economic choice available.
Turn off premium features for low-stakes tasks. Writing emails, summarizing Slack threads, and formatting documents don't need GPT-4o. Use GPT-4o-mini or Claude Haiku—models that cost 1/20th as much.
Reserve premium models for high-value work: code generation, complex reasoning, multi-step analysis, customer-facing interactions.
Usage Policy Frameworks
Jurkiewicz predicts that "the winners will be the ones that put guardrails in place early—prioritizing high-impact use cases and treating AI like a finite, strategic resource, not an unlimited utility."
Example guardrails:
- Limit premium model access to specific teams (engineering, data science, customer success)
- Set department-level token budgets with monthly alerts at 75% and 90% thresholds
- Require justification for tool usage above certain thresholds (e.g., >100K tokens/month per user)
- Default to cheapest effective model unless user opts into premium tier
These aren't restrictions—they're optimization frameworks. You want AI adoption, but you want efficient adoption.
What CIOs Need to Know
Usage Visibility Infrastructure
Trevino emphasizes that FinOps platforms are now essential for enterprises with significant AI spend.
Without centralized visibility, you can't see:
- Which teams are consuming the most tokens
- Which models are driving costs
- Whether high-spend users are creating high value
- When budgets will be exhausted
FinOps tools integrate with AI vendor APIs (OpenAI, Anthropic, Google, AWS Bedrock) to aggregate usage across platforms. Some provide cost allocation by department, project, or user.
Role Model the Right Behavior
Sen recommends identifying employees who use AI well—not just frequently.
"If you want to see more people leveraging AI across the board, point to the people who are already using it successfully. Role models can be a great way to show people how they can be more effective at their job and encourage AI use."
This flips the leaderboard model. Instead of celebrating volume, celebrate outcomes:
- Engineer who reduced code review time by 40% using Claude Code
- Sales team that increased qualified leads by 25% using AI-powered outreach
- Finance team that cut month-end close from 10 days to 6 using AI reconciliation
Volume is noise. Outcomes are signal.
The Reckoning
"As pricing shifts to metered models, companies are going to get a wake-up call," Jurkiewicz says. "The winners will be the ones that put guardrails in place early."
Uber's budget crisis is a preview. Every enterprise that deployed AI in 2025 with unlimited access policies will face the same reckoning in 2026 as vendors shift to usage-based pricing.
The question isn't whether you'll hit a budget wall. The question is whether you'll have visibility and governance in place before you hit it—or whether you'll be capping usage mid-year after the damage is done.
The AI adoption spending spree is over. Valuemaxxing is the next chapter.
Continue Reading
Sources
- The AI adoption spending spree is over. Time to focus on value. - CIO.com, June 12, 2026
- Uber caps usage of AI tools like Claude Code to cut costs - Bloomberg, June 2, 2026
- Disney AI Adoption Dashboard Tokens Tokenmaxxing - Business Insider, April 2026
- Linux Foundation Announces the Intent to Launch the Tokenomics Foundation - Linux Foundation, June 2026
- What is Valuemaxxing? - Flexera Blog, 2026
