AI industry sees record infrastructure growth amid falling token prices

The artificial intelligence sector is experiencing a divergence between infrastructure growth and operational costs. Global semiconductor equipment billings rose 23% year-over-year to $40.53 billion in Q2 2026, marking a second consecutive record quarter driven by sustained investment in AI computing capacity.

Conversely, the economics of AI model usage are shifting. Token prices for frontier models have fallen significantly due to increased competition and the rise of open-source models, with some indices reporting record lows. While this benefits end-users, it creates pressure on the profitability of major providers like OpenAI and Anthropic as they prepare for potential public listings.

For enterprises, managing AI spend remains a challenge. Despite falling unit prices for tokens, many companies report that AI budgets are exceeding projections. This is attributed to complex factors such as reasoning models generating hidden thinking tokens, agentic workflows triggering cascading calls, and the lack of unified visibility across multiple AI vendors. Companies like Salesforce are attempting to address these concerns by offering more predictable, outcome-based pricing models to prevent a mismatch between AI expenditure and actual productivity gains.