📉 AI Training Data Shortage
Research from Epoch AI predicts that the supply of high-quality human data for training LLMs could be exhausted by 2026–2032. This is driving the formation of a new market for specialized data collection: startups like Kled AI pay for video, Silencio AI collects audio, and Mercor hires experts for labeling.
🌍 The shortage of public text is forcing companies to move from mass web scraping to creating controlled, paid, and multimodal datasets, which is changing the economics of model training.
👤 New ways of earning money are emerging through the provision of personal data (video, sounds, professional knowledge).
Source 1: https://epoch.ai/publications/will-we-run-out-of-data-limits-of-llm-scaling-based-on-human-generated-data Source 2: https://www.businessinsider.com/avi-patel-kled-ai-pays-for-data-7-million-funding-2026-3