Trends in Frontier AI Model Count: A Forecast to 2028

2025-04-21

AI Summary

Regulatory thresholds based on training compute (FLOP) in global AI governance—such as the EU AI Act’s 10²⁵ FLOP threshold and the U.S. AI diffusion framework’s 10²⁶ FLOP threshold—are rapidly becoming obsolete due to exponential growth in model training costs, undermining regulatory coverage and sustainability. Method: We employ time-series extrapolation, uncertainty quantification (90% confidence intervals), empirical scaling laws linking compute to model size, and threshold sensitivity analysis to project the number of frontier foundation models exceeding each threshold annually from 2025 to 2028. Contribution/Results: By 2028, we project 103–306 models will exceed the EU threshold and 45–148 the U.S. threshold. Crucially, we propose a novel, stable threshold paradigm—“relative maximum training compute”—which anchors regulation to the empirically observed annual peak training cost across active models. This yields a steady annual coverage of 14–16 models, markedly enhancing regulatory foresight, adaptability, and long-term viability.

Technology Category

• Philosophy and Ethics of AI: AI & Law, Justice, Regulation & Governance
• Humans and AI: Other Foundations of Human Computation & AI
• Machine Learning: Deep Neural Architectures and Foundation Models

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Governments are starting to impose requirements on AI models based on how much compute was used to train them. For example, the EU AI Act imposes requirements on providers of general-purpose AI with systemic risk, which includes systems trained using greater than $10^{25}$ floating point operations (FLOP). In the United States' AI Diffusion Framework, a training compute threshold of $10^{26}$ FLOP is used to identify"controlled models"which face a number of requirements. We explore how many models such training compute thresholds will capture over time. We estimate that by the end of 2028, there will be between 103-306 foundation models exceeding the $10^{25}$ FLOP threshold put forward in the EU AI Act (90% CI), and 45-148 models exceeding the $10^{26}$ FLOP threshold that defines controlled models in the AI Diffusion Framework (90% CI). We also find that the number of models exceeding these absolute compute thresholds each year will increase superlinearly -- that is, each successive year will see more new models captured within the threshold than the year before. Thresholds that are defined with respect to the largest training run to date (for example, such that all models within one order of magnitude of the largest training run to date are captured by the threshold) see a more stable trend, with a median forecast of 14-16 models being captured by this definition annually from 2025-2028.

Problem

• Estimates number of AI models exceeding EU and US compute thresholds by 2028.
• Analyzes superlinear growth of models surpassing fixed FLOP thresholds annually.
• Compares trends for absolute vs. relative compute-based regulatory thresholds.

Innovation

• Forecasting models exceeding FLOP thresholds
• Analyzing superlinear growth in AI models
• Comparing absolute and relative compute thresholds