State of Data — Sean Cai, Independent / State of Data

Market Trends & Industry Dynamics - Data annotation scale in Manila images reaches 10 to 15 billion dollars a year per lab, though considered the least interesting part [2] - Industry shifts from vertically integrated giants to specialized suppliers, with labs mandating vendor diversification across 20 to 30 different vendors [4][6] - AI industry experiences exponentially increasing capital expenditure spend while AI revenues fall far behind [10] - Data markets serve as an upstream indicator, where vendor spending shifts on cybersecurity and biological data precede lab product launches by 2 to 3 months [32] Investment Opportunities & Potential Risks - Data represents the underfunded leg in the compute, data, and talent equation, presenting significant market opportunities [9][10] - Data companies face the "Goodhart's law with a profit motive" risk where contrived benchmarks and snake oil data fail to test long-dependent episodes [23][24] - Financial benchmark testing reveals that Opus 4.8% scores worse than 4.7% on various rubrics, and GPT 5.5% and Opus 4.08% score within 3 points while showing opposite strengths in arithmetic versus methodology [29][30] - Historical infrastructure parallels show that no infrastructure pioneer holds more than 10% of the market in the long run [39] Corporate Performance & Business Strategy - Successful data companies pivot to enterprise business models, recognizing that durable value accrues to application layers rather than raw data [43][46] - Application layer companies successfully decouple from model layers, proven by cases like GLM 5.2% surpassing GPT on real-world rubrics [41]