ASUS launches full-stack agentic AI offering for enterprise datacenters

ASUS has introduced a full-stack offering for businesses that want to run agentic AI on their own systems, according to The Register. The company announced the package, called its Enterprise Agentic AI Solution, at NVIDIA's GTC show in Berlin on October 21 and 22.

The launch comes as spending on AI hardware keeps climbing. IDC reports that the global server market expanded by 30 percent in a year, reaching $122.6bn in the first quarter of 2025, and the analyst firm now describes AI-driven investment as durable rather than cyclical.

The offering has three layers. One is a set of AI servers and supporting infrastructure, including several new units. Another is a software stack for connecting and managing AI workflows. The company says the package combines products it already had for enterprise AI customers with new additions.

IDC expects inference to make up more than 70 percent of computing demand by next year, with the fastest growth happening at the edge. The analyst firm also notes that adoption is moving past public cloud providers into enterprise settings, where organizations want to operate agentic AI themselves.

Owning infrastructure appeals to some companies for reasons of data control. Regulated industries, among them banking, health provision, factory operations and government bodies, often face sovereign data requirements. Others want costs for delivered tokens to be predictable, which tunable in-house systems can support.

ASUS executives argue the market shift makes an end-to-end product timely. One senior figure at the company, Paul Ju, who leads its infrastructure solution business group, said the servers are built to handle demanding task execution and large-scale deployment of AI agents.

Ju also pointed to a change in what AI servers must do. In his account, these machines were once tuned mainly for GPU work, but agentic workloads now run a repeating cycle of planning, code execution, tool calls and correction that leans heavily on the CPU. That added CPU load, he said, is creating bottlenecks in conventional AI servers, starting with single-thread performance that leaves GPUs waiting.

ASUS says it expects the surge to continue in datacenters and also at edge sites, particularly in industrial settings, and on developers' desks.