Contrary to the conventional view, the training phase for advanced AI models, especially those using reinforcement learning, now demands more compute resources for inference tasks than for the actual backpropagation process where the model learns.

The original playbook of simply scaling parameters and data is now obsolete. Top AI labs have pivoted to heavily designed post-training pipelines, retrieval, tool use, and agent training, acknowledging that raw scaling is insufficient to solve real-world problems.

The dominant use of AI compute is moving from training massive models to running inference tasks. This shift fundamentally alters the market, enabling broader enterprise adoption via cheaper, open models and changing the demand profile for compute hardware beyond the absolute cutting edge.

The era of advancing AI simply by scaling pre-training is ending due to data limits. The field is re-entering a research-heavy phase focused on novel, more efficient training paradigms beyond just adding more compute to existing recipes. The bottleneck is shifting from resources back to ideas.

Pre-training on internet text data is hitting a wall. The next major advancements will come from reinforcement learning (RL), where models learn by interacting with simulated environments (like games or fake e-commerce sites). This post-training phase is in its infancy but will soon consume the majority of compute.

To minimize the total cost for a certain level of performance, the compute budgets for a model's lifecycle stages should be balanced. A powerful heuristic is to equalize the costs: the compute spent on pre-training should roughly equal the compute for RL/fine-tuning, and also equal the total compute for user inference.

AI progress was expected to stall in 2024-2025 due to hardware limitations on pre-training scaling laws. However, breakthroughs in post-training techniques like reasoning and test-time compute provided a new vector for improvement, bridging the gap until next-generation chips like NVIDIA's Blackwell arrived.

A study that trained old models on new datasets (and vice versa) found that data improvements accounted for a 9x gain in compute efficiency, while architectural changes only provided a 3x gain. This suggests that data quality, filtering, and synthesis have been the predominant drivers of progress in pre-training, at least at smaller scales.

AI's computational needs are not just from initial training. They compound exponentially due to post-training (reinforcement learning) and inference (multi-step reasoning), creating a much larger demand profile than previously understood and driving a billion-X increase in compute.