The Most Important Chart In AI Right Now

Market Dynamics and Token Share - Open weights models have achieved higher token share usage compared to closed weights models, with DeepSeek capturing 25.2% of total token share compared to Anthropic's 24.5% [7] - Total model spend remains heavily skewed toward premium closed-source models, where Anthropic captures 64.6% of total token spend dollars compared to DeepSeek's 2.8% [8] - Frontier closed labs such as Anthropic and OpenAI lead in revenue generation, with annualized run rates exceeding 65 billion dollars and 40 billion dollars respectively [21] Industry Trends and Corporate Adoption - Major technology and enterprise companies are adopting open-source and open weights models, such as Thomson Reuters and Airbnb utilizing Qwen, and Perplexity utilizing DeepSeek [24] - Open weights models provide critical business advantages including complete data ownership, enhanced privacy, bargaining power, and extensive customization capabilities [32][35][38] - Enterprise AI adoption is transitioning toward building internal knowledge and equity through open-source model training rather than purely renting intelligence from closed labs [44][45] Investment Opportunities and Geopolitical Risks - Open-source AI popularity drives higher demand for AI infrastructure and hardware, resulting in significant commercial benefits for chip manufacturers like Nvidia [15][16] - Pricing differentials between models are substantial, ranging from 50 dollars per million output tokens for premium models down to 18 cents per million output tokens for flash models [19] - The heavy reliance of US enterprises on Chinese open-source models presents a potential geopolitical risk regarding future dependency on foreign chip co-designs [49][50]