The Multi-Sector Acceleration of AI

Business Model and Compute Shifts

The AI sector is moving away from the subsidy-heavy flat-rate pricing era toward usage-based billing as enterprises confront the true costs of scaling token-hungry agents. Anthropic has projected its first profitable quarter, marking a milestone for AI labs, while OpenAI reported strong Q1 revenue growth driven by high-volume coding models. Concurrently, compute infrastructure is expanding through non-traditional channels, most notably with SpaceX positioning itself as an AI compute provider by scaling up access to its Colossus data centers for partners like Anthropic.

Model Capability and Research Breakthroughs

OpenAI demonstrated a significant leap in model reasoning by using a general-purpose LLM to disprove a long-standing geometry conjecture by Paul Erdos. The model solved the problem without specialized training or complex prompting, suggesting that LLMs are becoming capable of autonomous scientific discovery. This trend is further supported by Andrej Karpathy joining Anthropic to lead research into recursive self-improvement, focusing on using models to accelerate their own pre-training research.

Consumer Integration and Policy

Google is integrating agentic capabilities directly into Search, allowing users to move from one-time queries to persistent, ongoing information gathering. This shift is accompanied by a broader push to integrate AI into existing product suites like Docs. Meanwhile, the political landscape remains volatile; while California is exploring labor disruption policies, federal efforts to mandate AI safety benchmarking were recently derailed by executive intervention, reflecting a prioritization of maintaining a competitive lead over China over regulatory oversight.