Humanoid robots promise to operate in the same environments as humans, but their success depends on far more than hardware. This roundtable explores the critical role of data in enabling humanoids to perceive and interact with the physical world. We will discuss the types of data required to train movement, dexterity, decision-making, and social interaction, as well as the challenges of sourcing, labeling, and validating data at scale. The conversation will focus on what it truly takes to move humanoids from impressive prototypes to reliable, real-world systems - and which types of data make the difference.