CANCOM & Cisco are preparing data centres for the AI era

Artificial intelligence is currently not only transforming business models, but is also fundamentally challenging the technical foundation of many organisations: the data centre. Whilst traditional IT infrastructures were long designed primarily for stability and scalability, modern AI workloads now demand entirely new network, compute and storage architectures.

“AI is only as powerful as the network behind it,” says Thomas Gerbafczits, Senior Director of Network Solutions at CANCOM Austria AG. Performance, latency and lossless data transmission are critical factors, particularly when it comes to large language models and agentic AI.

Particularly when training large AI models, enormous volumes of data need to be moved between GPU clusters. Even the slightest delays or packet loss have an immediate impact on training times and costs.

“Many companies underestimate how crucial the interplay between network, compute and storage is for AI workloads,” adds Michael Perfler, Senior Director of Data Centre Infrastructure Solutions at CANCOM Austria AG. “It is no longer enough to consider individual components in isolation.”

Modular platform

Cisco addresses these requirements with its data centre platform centred around Nexus switches and the ACI (Application Centric Infrastructure) architecture. The aim is to operate networks in a software-defined manner and adapt them dynamically to changing workloads.

For CANCOM, the strength of this approach lies primarily in integration. “Today, businesses need platforms that address performance, security and transparency collectively,” says Gerbafczits. Network, compute and storage are increasingly merging into a single, orchestrated system.

At the same time, many companies face the challenge of gradually modernising existing data centres without jeopardising ongoing processes.

“The reality in most companies is hybrid,” explains Perfler. “That’s why we need architectures capable of supporting both traditional workloads and modern AI applications in parallel.” This is precisely where the collaboration between CANCOM and Cisco, with its modular platform approaches, comes into play.

Performance and security are converging

In addition to performance, operational security is also becoming increasingly important. AI workloads are often business-critical and place high demands on transparency and monitoring.

“Cisco integrates analytics and observability functions directly into the infrastructure,” explains Gerbafczits. The Unified Cisco Nexus Dashboard acts as a central management and visibility platform. Through the integration of Splunk, network, security and application data can be analysed in real time.

CANCOM also sees a growing need for modernisation in the field of storage networks. With the MDS series, Cisco addresses the requirements of high-performance storage area networks.

“AI applications generate enormous volumes of data,” says Perfler. “That is why today it is not just about storage capacity, but about flexible and high-performance storage architectures.”

AI-ready infrastructure as a platform approach

Another key focus is on so-called AI-ready data centre and edge architectures. Here, Cisco is pursuing a consistent platform-based approach, from the central data centre right through to edge locations.

“Many AI applications require real-time decision-making – for example, in industrial or IoT environments,” explains Perfler. This is why processing data directly at the point of origin is becoming increasingly important.

With so-called AI Pods, Cisco goes one step further. These are pre-validated infrastructure building blocks that combine networking, GPU-based compute systems and storage within a coordinated architecture.

“The major advantage of these AI Pods lies in their rapid implementation and lower integration risks,” says Perfler. This enables companies to set up productive AI environments much more quickly.

At the same time, security is also taking centre stage. “Security is no longer just an add-on, but an integral part of the network architecture,” emphasises Gerbafczits. Cisco is focusing on new security approaches such as Hypershield and Hybrid Mesh Firewall.

Holistic transformation rather than isolated projects

For CANCOM and Cisco, one thing is clear: building an AI-ready data centre is not a one-off upgrade, but a strategic transformation. “Companies need to rethink their infrastructure holistically,” says Perfler. The interplay between architecture, operations, security and automation is crucial.

The technological foundation for this is already in place today. “Cisco provides the platforms; together with our customers, we use them to create bespoke end-to-end solutions. As Austria’s leading Cisco partner, we support our customers with the highest level of technical expertise – from strategic consultancy through planning and design to implementation, operation and continuous development,” explains Gerbafczits. Given the rapid developments in the AI sector, the speed of modernisation is increasingly becoming a decisive competitive factor.