Delos Data builds the connections that let AI processors, memory and storage work together, targeting cloud operators, AI infrastructure builders and enterprises running AI models. On September 15, the company announced more than $100 million in financing to expand its Nonstop AI portfolio. Its premise is that processors can spend valuable computing time waiting for data to arrive across the network.

The company’s software helps infrastructure teams see where that waiting occurs. Delos says its Mosaic software detects inference bottlenecks and helps tune workloads across a cluster—a connected group of computing devices. In May, it offered early software deployments to selected customers. The intended benefit is to get more useful work from installed processors by improving how data moves between them.

Delos is extending that approach into silicon with its newly announced Data Interface, which sits between a computing device and the network. The company says the interface detects failures beyond the device and manages recovery in hardware, including when an accelerator or link fails. Planned forms include a chiplet integrated into processors, an implementation using optical connections, and a card for connecting CPUs, memory and storage.

The financing announcement names Matrix, Playground, Socratic Partners, Capricorn’s Technology Impact Fund, Matter Venture Partners, IAG and individual industry investors as backers. Delos says the money will expand its software and hardware engineering teams and accelerate product development and sales. The announcement does not assign the financing a round label.

Customers can also work on a cluster before building it. Delos says its design platform lets infrastructure teams and model builders simulate different network layouts, test workloads and evaluate failures before committing to equipment. Its shared architecture is intended to accommodate different processors, models and network arrangements within one system, giving customers a way to design infrastructure around the workload they need to run.

The products are at different stages of delivery. Delos says Nonstop AI Clusters is already in production on existing infrastructure and its design platform is available now; the Nonstop AI Server is scheduled for customer sampling at the end of 2026. That timetable makes the next business test concrete: whether customers using the software and design tools carry those relationships into deployments of Delos’s new hardware.