Delos Data has raised more than $100 million to accelerate its Nonstop AI architecture, a full-stack approach to AI infrastructure that treats data movement between heterogeneous processors, memory and storage as a central constraint on large-scale inference. The company said the capital will expand its software and hardware engineering teams while accelerating product development and sales.
Investors include Matrix, Playground, Socratic Partners, Capricorn's Technology Impact Fund, Matter Venture Partners and IAG, along with individual investors from the compute, hyperscale, networking silicon, infrastructure software and optical-connectivity sectors. Alongside the financing, Delos introduced its Nonstop AI Data Interface and Nonstop AI Reference Architecture. The Data Interface sits between compute or data endpoints and the network, allowing GPUs, XPUs, accelerators, CPUs, memory and storage to participate in a common data domain. Delos plans three implementations: an I/O chiplet delivering more than 30 Tbps for GPUs, XPUs and accelerators; a near-packaged-optics implementation exceeding 10 Tbps; and a card implementation exceeding 400 Gbps for CPUs, flash and memory.
The architectural focus is persistent agentic inference rather than conventional request-and-response AI serving. Delos argues that these workloads increasingly span different processor types, models, memory systems and physical network topologies, making data movement and availability increasingly important to accelerator utilization. The company targets 10x lower latency and 10x higher efficiency, while its Reference Architecture is designed to scale toward hundreds of thousands of devices and handle endpoint, link, network and software failures. Nonstop AI Clusters are available today for customer co-design work, while the Nonstop AI Server is expected to begin customer sampling at the end of 2026.
What remains unknown is whether Delos can deliver on its vendor-stated 10x performance and efficiency figures in real deployments. The financing gives the company capital to develop silicon, systems and software concurrently, but customer results and independent evaluations have not yet been reported. Delos is demonstrating the architecture at AI Infra Summit in a heterogeneous cluster that continues operation through simulated failures, yet the path from demonstration to production-scale adoption remains unproven.