Biohub, a biomedical research institute founded by the Zuckerberg family, has launched the ESM Atlas, a groundbreaking protein structure database featuring 1.1 billion predicted protein structures. This release significantly surpasses the AlphaFold database, which contains over 200 million structures. The ESMFold2 AI model, developed by Biohub, is credited with generating these predictions and is claimed to outperform AlphaFold3 in various metrics. Notably, ESMFold2 is fully open-source, allowing unrestricted commercial use, which could reshape the landscape of protein AI research. The ESMFold2 model employs a novel approach by treating protein sequences as a language, similar to natural language processing techniques. This method enables the prediction of 3D structures directly from sequences, incorporating a vast array of microbial protein data absent from AlphaFold's database. The open-source nature of ESMFold2 is expected to mobilize the global scientific community, facilitating rapid iteration and application. While the academic community has responded positively, some experts urge caution, noting the need for independent validation of prediction results.