DataDirect Networks Sees AI Inference Moving to Smaller Data Center Sites
CEO Alex Bouzari said inference needs to run near users and their data, while equipment costs and shortages are increasing pressure to use hardware efficiently.

AI inference is prompting a shift toward smaller computing sites closer to users and their data, DataDirect Networks CEO Alex Bouzari said, as costly equipment and component shortages increase pressure to use hardware efficiently.
Bouzari, the company’s co-founder and chief executive, described potential sites ranging from 1 megawatt to 5 megawatts. He said they could serve workloads located near autonomous driving, physical AI and sensor data.
The smaller sites would supplement large data centers, Bouzari said. He described a software layer that can accommodate different types of hardware and help operators get more use from expensive equipment over time.
GPU, memory and solid-state drive shortages are coinciding with tighter budget constraints, Bouzari said. He argued that a data layer capable of doing more with less infrastructure is valuable under those conditions.
DataDirect Networks’ Enterprise AI HyperPOD packages CPUs, GPUs, networking and storage into a rack for deployment at existing corporate sites with available power. Keeping the systems on premises can give companies more control over their data, Bouzari said.
Bouzari said the data layer is becoming more important as inference and agentic AI grow. He said successful AI applications need both computing capacity and data infrastructure.