While the cost-per-token is decreasing as models become more efficient, this efficiency gain drives a massive increase in new use cases and overall consumption. This economic principle, Jevons Paradox, explains why total enterprise spending on model inference is skyrocketing, even as the unit cost falls.

The era of 'token maxing,' where enterprises used AI models without cost constraints, is ending. Companies like Microsoft are now scrutinizing the ROI of their AI spend, leading to budget cuts and a potential deceleration in the hyper-growth seen by model providers.

A paradox exists where the cost for a fixed level of AI capability (e.g., GPT-4 level) has dropped 100-1000x. However, overall enterprise spend is increasing because applications now use frontier models with massive contexts and multi-step agentic workflows, creating huge multipliers on token usage that drive up total costs.

The explosive AI revenue growth stems from corporations re-categorizing the spending. It's no longer a line item in a constrained IT budget but a strategic investment in labor augmentation and replacement. This unlocks a vastly larger pool of capital from operational budgets, fueling hypergrowth.

The return on investment for enterprises adopting LLMs is exceptionally high. A typical complex task that might save $55 in human labor costs consumes a fraction of a million tokens, which cost about $5. This massive economic incentive is what fuels the surging demand for AI compute from corporate adopters.

Even as enterprises optimize AI spending for better ROI, overall spend will continue to grow rapidly. The adoption curve for new use cases and new enterprises is so steep that it overwhelms any efficiency gains from optimization, ensuring continued growth for model providers.

While the cost for GPT-4 level intelligence has dropped over 100x, total enterprise AI spend is rising. This is driven by multipliers: using larger frontier models for harder tasks, reasoning-heavy workflows that consume more tokens, and complex, multi-turn agentic systems.

Goldman's CIO predicts that while unit cost per token will decrease, the explosion in token usage from agentic systems will make total AI compute a major corporate expense. He suggests it should be compared to personnel costs, not traditional IT spending.

Dylan Patel’s firm, Semi Analysis, saw its AI spend rocket from tens of thousands to a $7M annual run rate. This personal anecdote illustrates the insatiable enterprise demand for cutting-edge AI, suggesting a willingness to pay that far exceeds initial expectations and even rivals salary costs.