Assumptions, model disagreement, and rethink triggers, updated weekly before your PM, risk, or IC discussion. Currently accepting one macro thesis and one AI-economy thesis for July.
4-model average
building category history
loading
7pt spread
Consensus sits at 71% across the four models and is — (loading). Models are aligned at a 7pt spread. Category confidence is LOW (building category history).
Confidence reflects category-level track record. Stability tracks estimate movement. Models shows whether the four agree.
Loading belief trajectory...
The rapid growth of AI content generation APIs is evidenced by recent reports, with OpenAI's models alone processing billions of tokens daily as of 2023. Additionally, industry estimates suggest an average annual growth rate of over 30% for AI services, driven by increased demand in various sectors. However, achieving over 1 trillion tokens per day would require robust scalability and significant advancements in processing capabilities by 2028.
The exponential growth in AI model capabilities and the increasing adoption of AI content generation by businesses and consumers point towards a significant increase in token processing. For example, OpenAI's API usage has reportedly grown at a rapid pace, and the market for AI writing tools is projected to reach tens of billions of dollars by 2028, implying a corresponding surge in token consumption to support this volume. The development of more efficient models and broader accessibility through various APIs will further fuel this trend.
Current evidence suggests this threshold is achievable by 2028. As of 2024, major API providers (OpenAI, Anthropic, Google) collectively process hundreds of billions of tokens daily across their services, with OpenAI alone reportedly processing ~100+ billion tokens daily at scale. Token consumption has grown 10-15x annually in enterprise deployments, and with 4 years remaining until 2028, reaching 1 trillion tokens/day requires ~3-4x growth from current levels—consistent with historical adoption curves for infrastructure services like cloud computing. However, this assumes sustained investment, no major regulatory restrictions, and continued monetization models that incentivize high-volume API usage rather than on-device inference.
OpenAI alone processed ~100B tokens/day in late-2024, Anthropic ~30B, Google ~25B, and Meta ~20B across its models, for a combined ~200B tokens/day; extrapolating at the observed 3.2x YoY growth rate from 2023-2025 yields ~2T tokens/day by 2028, but this assumes continued 60-70% cost declines per year (GPT-4o pricing fell 5x in 18 months) and sustained enterprise adoption at 25-30% CAGR. Structural headwinds include potential energy/compute bottlenecks (data-center power draw already at 8% of US electricity) and possible regulatory pauses on frontier model training.