20VC: Who Wins the Inference Market | Has Open-Source LLM Inference Peaked? | What Happens to the Cost of Compute | The Real Limits of Post-Training and RL with Erik Bernhardsson, Modal CEO

Inference is rapidly becoming the foundational layer of the modern cloud, driven by an insatiable demand for scalable compute. Modal, led by CEO Erik Bernhardsson, positions itself as a comprehensive platform by integrating sandboxes, training, and batch processing alongside raw inference to move beyond the race-to-the-bottom commodity model. Success in this capital-intensive market requires more than just hardware access; it demands a robust, developer-centric brand and the ability to manage variable demand across a fragmented GPU landscape. While frontier models dominate headlines, the long-term value lies in specialized, agentic workflows and the ability to provide consistent, high-performance infrastructure. As the industry shifts toward post-training and agentic applications, the ability to build efficient, multi-product toolsets will determine which companies survive inevitable market pullbacks and emerge as generational leaders in the AI era.