Delos Data raises over $100 million to cut AI data center bottlenecks

Delos Data, founded by Intel veterans, raised more than $100 million for networking chips and software aimed at AI data-center bottlenecks. With the round it introduced a Nonstop AI Reference Architecture and Data Interface that it says deliver 10 times lower latency and 10 times greater efficiency, joining earlier Nonstop AI Cluster and Server products. The money will expand engineering, product development and sales. The tech is built to work with mixed Nvidia, AMD, Cerebras and other hardware. CEO Ed Doe called a GPU, CPU or accelerator waiting on the network the most expensive idle asset in a data center. CTO Dan Daly said agentic AI performance is decided by how data moves between devices and needs an interface that does not break under load. Mezha.net separately quoted him, as Dan Daley, saying Delos can provide the fastest way to move that data. Pat Gelsinger of Playground Global said inference performance is a network problem now. Delos cites projections of a roughly 300-fold rise in token demand by 2030, tokens being the text units AI models process. What is contested: Sources disagree on two investor names, listing either IAG and DYNAMIQ or AIG and Dynamic. No left-rated or center-rated coverage exists in the record; the sole right-rated outlet is Independent Journal Review.