Arrcus Cites Growth Surge with AI Inference Focus

Arrcus, a long-running member of the Futuriom 50, this week announced record 2025 bookings growth of 3X (200%) across datacenter, telco, and enterprise customers for switching and routing applications deployed in production across thousands of network nodes globally.

The company also introduced a new product, the Arrcus Inference Network Fabric (AINF), targeted at the new challenges of delivering high-performance networking and policy-aware routing for AI inference applications across highly distributed networks.

Let's recap why this is important.

Growing with AI Infrastructure

Arrcus officials say the company's growth is being driven by enormous demand for AI infrastructure, both on the training and inference sides—but infrastructure is now pushing harder toward the inference side. Arrcus’s ArcOS network operating system and ACE platform can connect AI infrastructure with high-performance routing and switching, including policy-based control using advanced protocols such as segment routing (SRV6), EVPN-VXLAN, and BGP-LN.

AINF is an innovative, policy-aware networking platform that can steer traffic between inferencing nodes, caches, and datacenters. This will be helpful for both cloud and communications operators that want to prioritize the performance of AI inferencing services. It should also be popular with content delivery networks, which are becoming important players in AI inferencing markets.

Arrcus says it can boost AI infrastructure efficiency by boosting throughput Tokens per Second (TPS), reducing Time to First Token (TTFT), and improving End to End Latency (E2EL).

“AINF extends Arrcus’s leadership in distributed networking by delivering the first fabric designed to meet the latency, sovereignty, and power constraints of large-scale AI inferencing,” said Shekar Ayyar, Chairman and CEO of Arrcus, in a statement.

Riding the Inference Wave

With its new products and growth numbers, Arrcus is riding a nice wave as AI infrastructure demand moves from training, which builds the popular AI Large Language Models (LLMs), toward the action at the edge, where inferencing infrastructure serves up responses and results to queries to LLMs.

Futuriom has highlighted this trend as part of the "Age of Inference,” which is fueling the need for distributed, high-performance power and specialized hardware like GPUs, ASICs, and vector databases on the software side. Additionally, the rise of "agentic AI" requires complex, multi-step queries that make inference critical for real-time, user-facing applications.

Arrcus believes that Agentic AI is “bottlenecked by challenges in the speed of delivery of inference results, diversity in inference models, and bringing smart inference decision making closer to edge nodes.”

Indeed, as Inferencing infrastructure is deployed in distributed clusters, it will need to address the requirements of the many bottlenecks, which currently include latency, availability, constraints in power grid capacity, data sovereignty, and cost.

Arrcus believes that AINF solves many of these challenges by enabling inferencing with an intelligent "AI policy-aware" network fabric that dynamically routes AI traffic between inference nodes, caches, and datacenters to the most appropriate site. Operators can define business policies such as latency targets, data sovereignty boundaries, model preferences, or power constraints, says Arrcus.

The company says this technology will help operators deliver a better experience for inferencing and built-in quality of service, including service level objectives. The company’s release includes supporting comments from many partners and customers, including Broadcom, Lightstorm, 1Finity Inc., UfiSpace, Hitachi Ventures, and others.

Futuriom Take: Arrcus is making the right moves to expand its portfolio to address the needs of inferencing. But it won't be alone, as the large incumbent Cisco is also on the same path.