eric landau

Models and compute have captured trillion-dollar valuations, but data remains an order of magnitude smaller ($14B for Scale AI); as AI moves from demos to production requiring 99.99% reliability, high-quality curated data becomes the critical bottleneck and value driver.

Physical AI companies lack an internet-scale pre-existing dataset equivalent, making them extremely data-hungry to reach baseline model performance; data demand for robotics, autonomous vehicles, and manufacturing will rival or surpass LLM data needs.

Early investment in multimodal data handling (vision, audio, sensor, text) positioned Encord to capture physical AI tailwind, since real-world AI inherently requires fusing multiple sensory modalities unlike text-only LLM workflows.

80% of global economic activity involves manipulating physical objects, and robotics leaders expect more robots than humans within a decade, creating a massive market for multimodal AI data infrastructure.

As physical AI moves from POC to production, companies face a data management crisis requiring specialized infrastructure to curate, annotate, and evaluate petabytes of multimodal sensor data, creating a scalable data layer opportunity.

METR institute metric shows AI task horizon doubling every 7 months (from milliseconds to 2-hour tasks currently), on track to reach 2-week sprint equivalence, indicating model progress is not plateauing despite benchmark saturation narratives.

Models are developing distinct 'personalities' — Anthropic dominating enterprise/coding, OpenAI focusing on consumer assistant, Google leveraging cash flow — creating a multi-winner landscape rather than winner-take-all, as different tasks require different capability profiles.

Users cannot articulate needs for capabilities they don't yet understand (Henry Ford 'faster horses' problem); adoption lags because people lack mental models for what AI can do, making problem-focused development more valuable than tool-first approaches.