Delos Data Unveils Nonstop AI™ to Build and Scale Resilient AI Clusters to Deliver Nonstop Tokens

Nonstop AI portfolio includes Reference Architecture, Data Interface, Software, Server, and Platform

PALO ALTO, Calif., Sept. 15, 2026 /PRNewswire/ -- Delos Data today expanded the Delos Nonstop AI™ portfolio with the addition of the Delos Nonstop AI™ Data Interface and Delos Nonstop AI Reference Architecture for low-latency, high-bandwidth, resilient agentic AI infrastructure. The new offerings round out the previously announced Delos Nonstop AI™ Clusters and Delos Nonstop AI™ Server enabling end-customers to build scalable AI clusters to dramatically lower cost per token. Today's Agentic AI Infrastructure requires putting together a Mixture of "X" Infrastructure (MoXI): a mixture of hardware, a mixture of models, and a mixture of switches, links, and topologies, all serving one workload.

The company also announced it has raised over $100 million from Matrix, Playground, Socratic Partners, Capricorn's Technology Impact Fund, Matter Venture Partners, IAG, and industry luminaries from across compute, hyperscale, scalable systems, networking silicon, infrastructure software, and optical connectivity to launch Delos Nonstop AI.

"There has been real excitement around optical links, new standards, and die-to-die interconnects, each one solving a real piece of the puzzle," said Ed Doe, CEO and co-founder of Delos Data. "What's missing is the network in aggregate, rethought for a workload that didn't exist when these networks were designed. Agentic AI is always on, always running, always being fed. You can't get there by applying the old network to a new problem. Delos Nonstop AI is about building the faster, stronger, more efficient AI capacity and putting some MoXI into the infrastructure."

"The data interface to endpoints in AI workloads is a critical enabler of performance. It stitches everything together in the right way," said Dan Daly, CTO and co-founder of Delos Data. "These engines spend most of their time waiting to be fed. When you're stitching together endpoints from different vendors at this scale, you can't engineer failure away, you must build for it from the beginning. Get those two things right and the efficiency follows: faster frees up capacity, and always available means more time doing real work instead of recovering."

Delos Nonstop AI™ Reference Architecture. A system-level blueprint for agentic AI infrastructure organized around quicker data movement between endpoints, larger scale to the order of hundreds of thousands of devices rather than one rack or a handful and higher resiliency to endpoint, network, software, and link failures. The architecture introduces a data interface between every endpoint in the data domain, so that heterogeneous compute, acceleration, memory, and storage can be composed to the workload to deliver Nonstop Tokens.

Delos Nonstop AI Data Interface. One data interface, delivered in three form factors so it can serve each class of endpoint:

1. I/O chiplet - 30+ Tbps, multi-protocol, for GPUs, XPUs, and Accelerators
2. Near-packaged optics - 10+ Tbps, multi-protocol, integrating Delos Data Interface with leading optical engines
3. Card - 400+ Gbps, multi-protocol, for CPUs, flash, and memory endpoints joining the Nonstop AI data domain

Delos Nonstop AI Data Interface enables each endpoint to have higher bandwidth and lower latency to every other endpoint. Because it sits between the endpoint and the network, the Delos Nonstop AI Data Interface is the first place a failure on the other side can be detected. It manages the recovery in hardware, whether the cause is a dead accelerator, a broken link, or a routine software update. Delos Data targets 10x faster performance, 10x stronger resiliency, and 10x higher scale than what endpoints can achieve today.

Delos Nonstop AI Clusters now include a platform to design and simulate a cluster topology before deployment. The platform combines a PCIe card for validating IP against real workloads across different network topologies, data interface logic for pre-silicon simulation and emulation, and a behavior-accurate, cluster-level performance modeling suite for evaluating scale, failure modes, and reliable transport. It supports a variety of topologies, letting infrastructure teams and model builders co-design infrastructure with their workloads before committing to a build.

Availability and business model
Delos Nonstop AI Clusters, showcased earlier this year at GTC, is in production today running clusters built using existing infrastructure. The platform is available now for customers to co-design their next-generation agentic AI infrastructure. Delos Nonstop AI Server, showcased earlier this year at Computex, will sample to customers at the end of 2026.

See Delos live at AI Infra Summit
Delos Data is exhibiting at booth 1344 at the AI Infra Summit, showcasing its technology and products - including Nonstop AI Cluster Live, a showcasing of heterogeneous endpoints composed into a single data domain, moving data through Delos Nonstop AI Data Interfaces and surviving failures without stopping the workload. Attendees can see the Delos Nonstop AI Clusters in operation and meet with the team.

About Delos Data
Delos Data is building Delos Nonstop AI™ for faster, stronger and more efficient inference, starting from how compute, acceleration and data is distributed and scaled. The team is a mixture of software, systems, and silicon experts focused on delivering the world's most capable and responsive intelligence at scale. To learn more, please visit www.delosdata.com

Press Contact
Kelly Karr
Tanis Communications
[email protected]
(408) 718-9350

SOURCE Delos Data