The Problem with AI Demand

Why It Matters

Mis‑priced AI consumption inflates costs and risks overbuilding infrastructure, forcing firms to rethink budgeting and pricing models to sustain growth.

Key Takeaways

• AI token consumption outpaces budgets, causing overspend across firms.
• Only Anthropic prices AI usage based on actual demand.
• Companies incentivize token use, leading to gaming and inefficiency.
• Flat‑rate plans become unsustainable as agents consume massive tokens.
• Mis‑measured demand risks overbuilding AI infrastructure for future growth.

Summary

The video highlights a broken AI demand signal: companies are burning through tokens faster than they can budget, and only Anthropic is pricing AI based on real consumption. This mismatch threatens the sustainability of the current AI spending cycle.

Tokens are the unit of AI cost, and new autonomous agents can consume millions without user oversight. Uber’s CTO noted its AI budget was maxed out by April, while Goldman Sachs warns inference costs could rival engineering headcount. Leaderboards at firms like Meta and Shopify reward raw token usage, encouraging wasteful behavior.

Anthropic’s CEO warned that flat‑rate plans are untenable; the firm has stopped unlimited subscriptions for third‑party tools and is moving enterprise customers to per‑token billing. Jensen Huang’s comment about engineers needing to spend $250,000 in tokens underscores the pressure to justify spend, while Uber and other firms scramble to control runaway costs.

If token demand is overstated, the massive infrastructure being built—GPUs, data centers, and cloud capacity—may be oversized, leading to stranded assets and higher prices for genuine users. Companies must adopt granular budgeting and shift away from vanity metrics to ensure AI investments deliver real productivity gains.

Original Description

The main demand signal for artificial intelligence looks explosive on paper, but it may be significantly overstated. Anthropic, by pricing its tools for that reality, might be the best-positioned AI company if a correction comes.

CNBC's Deirdre Bosa explains.