NEA investor Tiffany Luck: companies still haven't figured out their ROI from AI

In early 2025, "tokenmaxxing", an unofficial corporate race to maximize AI usage, became the dominant trend in Silicon Valley. A few months later, the bills arrived.

What is tokenmaxxing

The term tokenmaxxing emerged in corporate settings: CEOs directly urged teams to use AI tools without limits and to push token consumption to the maximum. The logic was straightforward: whoever uses AI the most adapts faster and outpaces competitors.

The race heated up at every level, from individual teams to entire corporations, from startups to tech giants. Companies rolled out corporate subscriptions to ChatGPT, Claude, Gemini and other tools en masse, effectively competing on the scale of consumption. Internal AI usage rankings became part of corporate culture.

When the bill came due

Reality turned out to be harsher. Uber, according to publicly available sources, spent its entire annual AI budget in a few months. Several large companies cut corporate Claude licenses, in some cases for entire divisions that could not show measurable results. Meta went further: the company shut down an internal leaderboard of AI usage among employees.

This does not mean abandoning the technology. But the first phase of unrestrained experimentation is coming to an end. In its place comes a question that should have been asked first: what exactly did we get for this money?

The NEA investor's view

Tiffany Luck, a partner at NEA, a venture fund focused on technology investments, states the problem directly: the enterprise market for AI tools is only beginning to work out its real return on investment.

"Enterprises are still figuring out what their real return on AI is," said

Tiffany Luck, NEA.

In her observation, most companies entered the AI race under pressure from competitors and fear of missing the trend, without a clear method for measuring effectiveness. Success metrics were either never defined or proved unworkable in practice. As a result, companies spend significant sums but cannot answer the basic question: is it worth it?

What comes next

The enterprise AI market is entering a phase of normalization. Companies are starting to look for specific use cases with measurable returns:

• Process automation with a clear KPI
• Lower operating costs by partially replacing manual labor
• Faster development, from idea to release
• Better customer service through AI agents
• Intelligent search and analysis of internal company data

Companies that learn first to measure AI returns accurately will gain not only a technological advantage but also a more sustainable economic model in the next stage of the race.

What this means

The first wave of corporate AI enthusiasm has collided with accounting reality. The coming year will show who managed to turn AI experiments into measurable value, and who simply burned their budget on tokens.

*Meta has been designated an extremist organization and is banned in Russia.