How 'tokenmaxxing' could be distorting AI demand signals

Summary

• Introduces the concept of "tokenmaxxing," where engineers are scored and incentivized to consume AI tokens, potentially inflating demand signals.
• Companies are using token budgets as recruiting perks; Shopify cites AI usage as an employee performance metric.
• Core thesis: This performative demand creates a risk for the over $1 trillion in planned AI infrastructure spending, which assumes real, accelerating demand.
• Questions whether hyperscaler claims of demand outstripping supply (e.g., Amazon's Andy Jassy) reflect genuine enterprise adoption or engineered metrics.
• The AI trade has gone sideways despite rising spending, indicating market skepticism about demand quality.
• Highlights a significant information gap: AI is a major corporate budget item but poorly understood; fintech Ramp calls it a "$1 trillion blind spot."
• Notes a strategic divergence: OpenAI is spending aggressively assuming real demand, while Anthropic is imposing user limits to curb inefficient token consumption.
• Concludes that AI demand is real but may be partially artificial or unsustainable, leading the market to potentially price in inflated expectations.