Dylan Patel's Information Machine (substratemag.com)

During the recent Nvidia GTC conference, founder Jensen Huang highlighted a groundbreaking increase in AI inference performance, showcasing a 50x improvement on the GB300 NVL72. This significant milestone underscores Nvidia's dominance in the AI hardware space and illustrates a broader trend of unprecedented infrastructure investment, with hyperscalers set to invest $600 billion this year alone. At the heart of this surge is SemiAnalysis, a maverick analytics firm founded by Dylan Patel, whose keen insights and data-driven projections have turned him into a pivotal voice for industry giants like Nvidia, AMD, and Meta. SemiAnalysis has evolved from a humble blog to an influential research platform, catering primarily to hedge funds and software developers involved in AI infrastructure. Its models—including the Accelerator & HBM Model—analyze critical supply chain components like chip production and memory allocation, enabling predictive insights for major players in the semiconductor industry. With revenues projected to exceed $100 million this year, SemiAnalysis exemplifies how a new genre of data analysis is reshaping the AI landscape, relying heavily on social media and community engagement for insights. As the AI ecosystem becomes increasingly complex, the firm’s unique approach positions it as a vital resource for navigating the rapidly changing infrastructure needs of AI development.