NEA Investor: Uber Burned Through Annual AI Budget in Months — Market Demands ROI

Token maximizing — a race by companies for maximum AI utilization without regard for costs — has resulted in the first serious hangover. Uber burned through its annual AI budget in just a few months, Meta* shut down its internal rating, and some companies cut Claude licenses. Venture investor Tiffany Lak from NEA calls what's happening "the hour of reckoning" on ROI — and believes the market is only maturing.

What is token maximizing

In early 2025, a trend spread among Silicon Valley CEOs: demanding that employees use AI tools as broadly as possible. The idea was intuitively clear — the more experiments, the higher the chance of finding a breakthrough application. Companies increased subscriptions to GPT, Claude, and Gemini without setting clear KPIs. Employees were encouraged to ask AI for everything — from code to strategic presentations. Meta* introduced an internal rating that tracked who "used" AI most actively. Corporate token budgets grew exponentially, while questions about return on investment were put off until later. But "later" came sooner than expected.

When the bill came

Uber, according to sources, spent its entire annual AI budget in just the first few months. This is not an isolated case: a number of large companies quietly cut Claude corporate licenses — primarily for departments that failed to demonstrate measurable value. Teams that "used AI" but couldn't tie it to concrete results lost access.

Meta* went even further: the company shut down the very internal rating it had launched to incentivize employees. The metric of "who generates more requests" turned out to be meaningless once the cost of each request became significant. Tiffany Lak from NEA calls this inevitable: the market is entering a phase where for every dollar spent on AI, a company must be able to account for concrete numbers — cost reduction, process acceleration, or revenue growth.

IPO and personal agents

Lak sees the current rethinking not as a crisis, but as normal market maturation. Companies that have gone through the experimental phase now understand where AI actually delivers value and where it was just window dressing for shareholders. Among the topics she highlights as key:

• AI IPO — the first public offerings of AI companies will take place amid heightened skepticism: investors will look not for growth rate, but for unit economics
• Personal agents — the next wave of value will not come from corporations, but from individual users with agents customized to their needs
• Shift in metrics — "usage" as a KPI is fading, being replaced by cost reduction, process acceleration, revenue growth

For investors like NEA, this means a new filter: they increasingly look at how well an AI company can prove ROI of its product to the customer, rather than simply showing attractive reach numbers.

What it means

The era of token maximizing did not kill AI in corporations — it killed mindless AI. Companies that have gone through this cycle and can demonstrate real results will come out of it stronger. The market has matured — and this is good news for those building real products rather than chasing beautiful dashboards.

*Meta is recognized as an extremist organization and banned in the RF.