NEA VC warns: AI tokenmaxxing era ends as ROI reckoning hits enterprises

Tokenmaxxing — the Silicon Valley trend of encouraging employees to push AI usage to its limits — has given way to a harsh ROI reckoning. NEA partner Tiffany Luck, speaking on TechCrunch's Equity podcast, points to real-world consequences: Uber reportedly exhausted its annual AI budget in just a few months, some organizations have cut Claude licenses for parts of their teams, and Meta shut down its internal AI leaderboard. Luck, who previously helped companies embrace e-commerce, now sees a parallel cycle in AI where the hype must be justified by measurable returns.

Luck argues that the next wave of value lies not just in model improvements but in every layer of the AI stack — from infrastructure to applications. She highlights the rise of 'personal agents' that can autonomously handle tasks, and the growing role of forward-deployed engineers as 'Trojan horses' for AI adoption within enterprises. Startups are stepping in to help companies track and optimize AI spending, while enterprises increasingly mix and match models from different providers rather than committing to a single vendor. The episode also touches on AI IPOs and how public markets are pricing these companies.