Tokenmaxxing was the hottest trend in Silicon Valley earlier this year, with CEOs encouraging employees to push AI usage as far as it would go. Then the bill came due. Uber reportedly blew through its annual AI budget in a few months, some companies cut Claude licenses for parts of their org, and Meta killed its internal leaderboard.
This tension between adoption and accountability is now forcing a reckoning across corporate America. Tiffany Luck, a partner at New Enterprise Associates (NEA), recently told Reuters that the majority of enterprises are still in the early stages of measuring their return on AI investments. The challenge, she said, is not a lack of use cases but a lack of frameworks for evaluating whether those use cases actually deliver value.
Also read: Anthropic becomes first AI startup to join Frontier's $915M carbon removal push
The tokenmaxxing hangover
The tokenmaxxing trend encouraged employees to use AI tools like ChatGPT, Claude, and GitHub Copilot as much as possible, often without clear cost controls. The idea was to accelerate experimentation and find unexpected productivity gains. But the financial consequences arrived faster than many companies anticipated.
Uber’s internal spending on AI APIs reportedly exceeded its entire annual budget within the first quarter of 2025, forcing the company to implement usage caps and review its AI procurement process. Other firms, including several Fortune 500 companies, began restricting access to premium AI subscriptions after realizing that a small percentage of power users were generating the vast majority of costs.
Also read: Android 17 arrives with bubble bar multitasking and deeper Gemini AI integration
Meta, which had maintained an internal leaderboard ranking teams by AI tool usage, quietly removed the board after it became clear that high usage did not correlate with high output. The company now focuses on measuring outcomes rather than raw consumption.
What enterprises are missing
Luck argues that the problem is not unique to any single company. Many enterprises rushed to deploy AI tools without first defining what success looks like. Productivity gains are often anecdotal, cost savings are difficult to isolate, and revenue attribution is nearly impossible when AI is used across hundreds of workflows.
She recommends that companies start with a small number of high-impact use cases, establish clear baseline metrics, and measure outcomes over a defined period before scaling. Without that discipline, Luck warns, enterprises risk repeating the cycle of overspend and backlash that has already hit early adopters.
The venture capital perspective is telling. NEA, which has invested in AI companies including Anthropic and Scale AI, is now advising portfolio companies to help enterprise customers build ROI frameworks. The message is clear: the era of blind AI adoption is ending, and the era of measurement has begun.
What comes next
The shift toward ROI accountability is likely to slow enterprise AI spending in the short term, but Luck believes it will strengthen the market in the long run. Companies that can demonstrate real value will retain their budgets and expand their AI programs. Those that cannot will face cuts.
For investors, the lesson is that AI adoption alone is not a business model. The companies that survive the current correction will be those that can show, in hard numbers, that their tools improve efficiency, reduce costs, or generate revenue. Everything else is just tokenmaxxing.
Frequently Asked Questions
What is tokenmaxxing?
Tokenmaxxing was a trend in Silicon Valley earlier this year where CEOs encouraged employees to push AI tool usage as far as possible, often without clear ROI tracking.
Which companies have struggled with AI spending?
Uber reportedly exhausted its annual AI budget within a few months, some firms cut Claude licenses for parts of their organization, and Meta ended its internal AI usage leaderboard.
What does Tiffany Luck say about enterprise AI ROI?
Tiffany Luck, a partner at NEA, observes that most enterprises are still in the early stages of measuring AI’s return on investment and have not yet established reliable frameworks for evaluating its business impact.
Why is measuring AI ROI difficult for enterprises?
Many companies adopted AI tools rapidly without clear metrics, leading to budget overruns and license cuts when the costs became apparent, revealing a lack of alignment between usage and value.
Be the first to comment