Lightstorm & Arrcus bring AI-ready networks to Asia-Pacific

Lightstorm and Arrcus have formed a strategic partnership to deliver AI-optimised networking across the Asia-Pacific region for distributed AI inference and training workloads.

The collaboration combines Arrcus' programmable, AI-aware networking software with Lightstorm's Polarin network-as-a-service platform. The integrated offering targets AI cloud providers, enterprises and infrastructure operators running workloads across multiple data centres.

Distributed AI is increasingly common as organisations spread compute and data across locations for resilience, capacity and regulatory reasons. That shift creates new network requirements: predictable latency, consistent bandwidth and faster provisioning as clusters expand across sites.

Network integration

The partnership focuses on connectivity across geographically distributed data centres. The integrated design uses automated, policy-driven controls, aiming to make the network a managed layer for AI workloads rather than a static transport service.

Polarin provides on-demand connectivity services across Lightstorm's infrastructure footprint, while Arrcus supplies routing and networking software built around programmability and policy awareness. Together, they use API-driven orchestration so customers can provision connectivity based on workload requirements and business policies.

The system targets two main AI activities: real-time inference, where response times depend on consistent latency between compute, storage and users; and distributed training, which can require deterministic bandwidth between nodes for efficient synchronisation and predictable job completion.

Amajit Gupta, Group CEO & MD of Lightstorm, framed the partnership as a response to how AI systems are built and operated across large regions.

"AI infrastructure requires the network to behave like part of the compute fabric. By integrating Arrcus' routing intelligence with our Polarin NaaS platform, we're enabling inference and training workloads to scale across Asia-Pacific without compromising performance or compliance with data sovereignty requirements," said Amajit Gupta, Group CEO & MD, Lightstorm.

AI workload demands

Latency consistency, bandwidth predictability and provisioning are the main operational issues the partnership aims to address. These factors affect inference performance, training efficiency, and the complexity of deploying and expanding multi-site AI clusters.

Policy-driven networking is central to the approach. It uses centrally defined rules to govern traffic flows between locations, allocate capacity and roll out changes. Customers can adjust connectivity as demand changes rather than rely on longer network change cycles.

Data sovereignty is also a key design constraint for regional AI deployments. Many Asia-Pacific organisations operate under rules that limit where data can be processed or stored, shaping AI pipeline design and where different stages of a workload can run.

Shekar Ayyar, Chairman & CEO of Arrcus, said the partnership is intended to make networking more responsive to changing AI demand.

"Distributed AI workloads-from real-time inference to large-scale training-require policy-aware, programmable networking that adapts to workload demands in real time. Our partnership with Lightstorm brings AI-optimized connectivity to operators across Asia-Pacific, helping them scale infrastructure efficiently while improving performance and utilization," said Ayyar.

Regional footprint

The solution is available for deployment across Lightstorm's fibre network spanning India and other key Asia-Pacific markets. That footprint supports AI operators running multiple sites that need predictable connectivity, especially as workloads move between locations or expand to additional data centres.

The companies also linked the partnership to broader AI infrastructure investment across the region. Asia-Pacific has seen rising demand for data centre capacity, interconnection services and specialist infrastructure to support AI compute at scale. As AI clusters become larger and more distributed, networking suppliers are increasingly focusing on automation and programmability.

Lightstorm operates digital network infrastructure and connectivity services across Asia-Pacific. Arrcus develops networking software and is headquartered in San Jose, California.