Frontier data market projected to reach $100 billion in three years

Handshake AI leaders project that the frontier data market, valued at $10 billion today, will reach $100 billion within three years. This growth is driven by the increasing complexity of AI training data, which has shifted from simple labeling to expert-authored environments for long-horizon agent tasks. The cost per unit of data is rising as labs require diverse, high-quality, and stylistically consistent data to train models on professional domains like finance, law, and life sciences. Handshake notes that data is now a primary driver of model progress, with labs like DeepSeek attributing gains to data quality.

The company is expanding its workforce to 800 employees, hiring domain experts to author data and build evals. Handshake is positioning itself as a key infrastructure provider for both AI labs and enterprises seeking to fine-tune models on proprietary workflows. Handshake AI believes that data quality and complexity are the new bottlenecks in AI development, and that the market for expert-authored data and evaluation services will expand rapidly as models tackle more complex, real-world tasks.