Why AI Compute Comes in Rectangles — Erik Bernhardsson, Modal CEO

Modal, the AI infrastructure platform founded by Erik Bernhardsson, has announced a $750 million Series D round that lifts its valuation to $16 billion and pushes annualized revenue past the $1 billion mark. The company has built nearly 3 billion “sandboxes,” isolated environments where AI models and agents run, and these account for roughly a third of its revenue—an outcome the founders did not foresee when launching the product.

Modal’s business model turns the fluctuating demand for GPUs into a shared pool for thousands of customers. Rather than selling fixed capacity contracts that create a rigid capacity‑time rectangle, the firm offers elastic capacity that is allocated based on actual need, cutting idle hardware costs and preventing resource shortages during usage spikes. This approach also carries financial risk: if GPU prices fall after capacity has been secured, the provider may end up with overvalued resources.

To enable rapid response, Modal has built a suite of infrastructure techniques, including GPU availability buffers, custom file systems, process checkpoints, and memory snapshots, allowing environments to be started and restored in seconds.

This development illustrates how serverless architecture is reshaping how AI compute is consumed and paid for, enabling scalable generative applications and lowering cost barriers for developers and enterprises. Modal’s success signals that demand for AI resources will continue to grow, driving further innovation in infrastructure management.

Read the original article on Latent Space

This summary is an informational synthesis produced by dataqbs.com. All rights to the original content belong to its author and the cited media outlet. We act solely as curators of technology news and claim no authorship.