Posted by James Burgess on 13th Nov 2026
AI Is Reshaping Infrastructure. What Does It Mean for Your Network?
From AI factories and accelerated computing to intelligent networks, NVIDIA GTC Berlin highlighted a significant shift in how organisations will design, deploy and scale their infrastructure. But what do these developments mean for businesses planning their next network investment?
NVIDIA GTC Berlin 2026 brought together some of the latest developments in artificial intelligence, accelerated computing and next-generation networking.
While the announcements focused on technologies supporting increasingly sophisticated AI workloads, the implications extend beyond organisations building large-scale AI systems.
For IT decision-makers, resellers and systems integrators, the message is becoming clearer: network infrastructure is playing an increasingly important role in determining how effectively organisations can deploy new technologies, manage growing workloads and scale their operations.
Key announcements and what they could mean for future infrastructure planning.
- AI factories: Infrastructure is becoming part of the competitive advantage
One of the major announcements at GTC Berlin was NVIDIA DSX POD, a modular approach to deploying complete AI factories based on NVIDIA Vera Rubin infrastructure.
By integrating and validating computing, networking, cabling, power and cooling before delivery, NVIDIA says deployment at a suitably prepared site could be reduced from approximately six months to as little as 14 days.
The significance goes beyond faster installation.
It demonstrates how infrastructure is increasingly being designed as an integrated system, rather than a collection of independently specified components.
What does this mean for customers?
For organisations planning data centre expansion or accelerated computing projects, it reinforces the importance of considering infrastructure requirements at the earliest stages.
Network architecture, connectivity, power, cooling and future scalability all influence the success of a deployment.
Even where organisations are not building AI factories, the underlying principle remains relevant: infrastructure designed around clearly defined workloads can reduce complexity and help avoid costly changes later.
The Gbics perspective: The most effective infrastructure investments begin with understanding the application and its requirements, rather than selecting hardware based on headline specifications alone.
- Networking is becoming more critical to AI performance
The continued development of NVIDIA's Vera Rubin platform reinforces the importance of high-performance networking in accelerated computing environments.
Technologies including NVIDIA Spectrum-X Ethernet, ConnectX networking and BlueField DPUs are designed to support increasingly demanding data movement, processing and infrastructure management requirements.
As computing environments become more distributed and workloads more intensive, network performance can become a limiting factor.
Investing in additional computing capacity delivers less value if the supporting network cannot move data efficiently between systems.
What does this mean for you?
Organisations considering infrastructure upgrades should evaluate more than bandwidth alone.
Key considerations include:
- Bandwidth: Does the network have sufficient capacity for current and anticipated workloads?
- Latency: Are applications sensitive to delays in data transmission?
- Compatibility: Will new switches, NICs and connectivity integrate with existing infrastructure?
- Scalability: Can the network accommodate future requirements without unnecessary replacement?
- Cost efficiency: Can the required performance be achieved while maintaining control over project budgets?
Importantly, not every organisation requires an AI-optimised network.
For many businesses, the right investment may be a targeted upgrade to 25G or 100G connectivity, rather than a complete infrastructure overhaul.
The Gbics perspective: The objective should always be to match network capability to business requirements. Higher speeds create value when they address a genuine performance or capacity need.
- AI is moving closer to where data is generated
Another important theme from GTC Berlin was the development of distributed AI infrastructure.
NVIDIA announced further collaboration with Nokia and major telecommunications operators to bring accelerated computing capabilities into networks, including distributed locations closer to users and devices.
This approach supports applications that benefit from processing data nearer its source rather than relying entirely on centralised infrastructure.
Potential use cases include industrial automation, intelligent communications and real-time analytics.
What does this mean for you?
As organisations consider more distributed computing models, they may need to reassess how their network connects central systems, remote locations and edge devices.
The right architecture will depend on the application, but considerations may include:
- Reliable connectivity between distributed locations.
- Appropriate bandwidth for data-intensive applications.
- Low-latency communication where real-time processing is required.
- Network resilience and the ability to expand capacity as requirements change.
For businesses with existing distributed infrastructure, these developments also highlight the importance of understanding where current network limitations may affect future projects.
The Gbics perspective: Network planning increasingly needs to account for where applications run, where data is generated and how information moves between systems.
- Accelerated computing is becoming more accessible
GTC Berlin also highlighted NVIDIA's expanding range of local AI computing platforms, including DGX Spark, DGX Station and RTX-based systems.
These technologies provide different options for developers, engineers and enterprise teams looking to run AI workloads locally.
Not every organisation needs a large-scale AI data centre to explore accelerated computing.
Smaller systems can support development, testing and specialised applications, with larger infrastructure considered as requirements evolve.
What does this mean for you?
Organisations evaluating AI should begin with the use case rather than the technology.
Questions worth considering include:
- What business problem are we trying to solve?
- What computing resources will the application require?
- Can existing infrastructure support the workload?
- Will the solution operate locally, centrally or across multiple environments?
- What additional networking or connectivity might be needed as adoption grows?
This approach can help organisations make proportionate investments rather than committing to infrastructure before requirements are fully understood.
The Gbics perspective: Accelerated computing creates new opportunities, but successful adoption depends on choosing an appropriate combination of computing, networking and connectivity.
- The bigger picture: Infrastructure planning needs to become more joined-up
Perhaps the most important takeaway from NVIDIA GTC Berlin is not any individual product announcement.
It is the growing relationship between computing performance, network architecture, connectivity and the wider infrastructure supporting them.
As technologies evolve, organisations face a balancing act.
They need to support new applications and increasing data demands while managing budgets, protecting existing investments and avoiding unnecessary complexity.
This creates an opportunity to take a more strategic approach to infrastructure planning.
Rather than asking simply, “Do we need a faster network?”, IT teams should be asking:
“What does our network need to support today, and how can we prepare it for tomorrow without overinvesting?”
For some organisations, the answer will involve higher-speed networking and accelerated computing.
For others, it may mean improving existing connectivity, removing bottlenecks or choosing more cost-effective alternatives to OEM components.
The important point is that infrastructure decisions should be driven by operational requirements and long-term value.
What this means for Gbics customers
As an NVIDIA Partner with Networking Competency, Gbics is ideally positioned to help customers and partners explore evolving networking and accelerated computing requirements.
Alongside NVIDIA networking and computing technologies, our compatible transceivers, AOCs, DACs and fibre cabling provide options for building and upgrading network connectivity while managing costs.
Through compatibility testing and technical support, we help organisations identify suitable connectivity solutions for their existing and future infrastructure.
Whether you're planning a targeted network upgrade, evaluating higher-speed connectivity or exploring accelerated computing, the starting point remains the same:
Understand the workload. Assess the infrastructure. Invest where it delivers real value.
Planning your next infrastructure project?
Explore NVIDIA solutions and compatible connectivity with Gbics, or speak to our team about your networking and infrastructure requirements.
