Hims & Hers CEO urges data-rich firms to adopt open-weight AI models

Andrew Dudum, chief executive of Hims & Hers, said companies with large proprietary datasets should switch from big proprietary AI systems to open-weight models to cut costs.

During one outlet Squawk Box interview, Andrew Dudum, chief executive of Hims & Hers, advocated for companies that own extensive, independent datasets to move away from large, closed AI models and adopt open-weight alternatives. He highlighted his company's own patient data repository as a "real asset" that can be used to train bespoke AI, potentially lowering expenses by as much as 70-80 percent. Open-weight models, which allow users to modify trained parameters, are not only cheaper at scale but also deliver better performance when tailored to a firm’s specific needs.

Dudum claimed that Hims & Hers' initial deployment of such a model already surpassed the capabilities of commercially available solutions and that its performance will continue to evolve quickly as it ingests new patient information. His remarks arrive as many firms are curbing token-maxing practices and seeking more efficient AI strategies, including model routing and the adoption of lower-cost models from providers like Moonshot AI and Chinese offerings. The story also notes recent data-acquisition moves by Google and Handshake AI, underscoring the growing value placed on high-quality training data.