The Compute 100: Erik Bernhardsson, CEO of Modal

Modal operates as a general-purpose AI infrastructure platform, abstracting away the complexities of compute, storage, and containerization for developers. Originally conceived as a pre-GenAI compute platform, the company pivoted to support the explosive demand for AI inference as GPU supply-side constraints intensified. The platform distinguishes itself through high-performance inference, cost-efficiency, and a robust developer experience that allows researchers to deploy directly to production without managing underlying Kubernetes or cloud infrastructure. Scaling to tens of thousands of GPUs, the business relies on dynamic, quantitative supply-demand matching across multiple hyperscalers and NeoClouds. While the inference market remains competitive, the company focuses on premium, high-value use cases—such as computational biotech and reinforcement learning—rather than commoditized, low-margin inference, leveraging a team of deployed engineers to solve complex technical challenges and drive product discovery.