Technology

Add as a preferred source on Google

DataDirect Networks Sees AI Inference Shifting to Smaller Sites

CEO Alex Bouzari described facilities of one to five megawatts for workloads near users, autonomous vehicles and sensor data as costly equipment and component shortages pressure operators to use hardware efficiently.

Updated 1 min read
Technicians install compact AI computing equipment inside a data-center room in Chatsworth, California. / EconoTimes.ai
Technicians install compact AI computing equipment inside a data-center room in Chatsworth, California. / EconoTimes.ai

DataDirect Networks sees artificial intelligence inference moving toward smaller computing sites near users and data, a shift its chief executive said could support workloads outside large data centers.

Alex Bouzari, the company’s co-founder and CEO, described potential facilities ranging from one megawatt to five megawatts. He said they could serve applications connected to autonomous driving, physical AI and sensor data while supplementing larger data centers.

The proposal comes as costly equipment and shortages of graphics processing units (GPUs), memory and solid-state drives pressure operators to use installed hardware more efficiently. Bouzari said software that works across different hardware types could help companies obtain more value from equipment over time as budgets tighten.

He also said “the data layer” is gaining importance as inference and agentic AI expand. Bouzari said effective AI applications depend on data infrastructure as well as compute resources.

The company’s Enterprise AI HyperPOD puts processing, networking and storage components into a rack that can be installed at corporate locations with spare power. Bouzari said keeping those systems on premises can give companies greater control over their data.

EconoTimes previously covered SK hynix’s memory-pooling plan for AI servers, which addresses capacity limits by adding shared memory alongside existing server systems.

Related articles

View all