AI spending shifts from unlimited 'tokenmaxxing' to ROI accountability, prompting companies to measure AI costs and mix multiple models instead of committing to a single provider.

What happened: Companies that aggressively pushed AI usage earlier this year—including Uber, which burned through its annual AI budget in a few months—are now facing cost pressures. Some firms have cut Claude licenses for parts of their organization, and Meta shut down its internal leaderboard. NEA partner Tiffany Luck is focused on how startups are helping enterprises track return on AI spend.

Why it matters: The tension between AI hype and measurable business results is reshaping how companies evaluate their AI investments. Enterprises are moving away from betting everything on a single AI model provider and instead mixing and matching models. Forward-deployed engineers (technical specialists embedded at customer sites) are becoming a key lever for AI adoption in the enterprise sector.

What to watch: Value is being created at multiple layers of the AI stack, not only at the model layer itself. The shift from enthusiasm to ROI measurement will likely determine which AI investments survive budget scrutiny in coming months.