Podcast Episode 2 Insights: Deciding the Future: Building New Networks for the AI Era
April 28, 2026
AI Factory
TL;DR
AI is shifting networking from a background function to a core enabler of scale.
Multiple networks, higher complexity, and accelerated compute are redefining infrastructure.
Leaders must engage early, design for openness, and future-proof their networks.
Introduction
For years, networking sat in the background of technology conversations. It was essential, but rarely the headline. Now AI has changed that. Conversations are shifting from traditional networking to AI-driven and open networking.
In this discussion, Vishal Shukla and Alan Weckel explore what it really takes to build AI factories and AI fabrics. This conversation focuses on how the market is evolving, how leaders should think about infrastructure, and what decisions matter today to prepare for the future.
Why is networking becoming central in the AI era?
The network is no longer just connectivity; it is the unifying layer that enables AI scale.
We are moving from one network to multiple networks to support AI factories. With the increasing use of GPUs and XPUs, networking becomes the glue that connects and scales these systems. This shift is unlocking a massive opportunity, with data center networking expected to grow significantly while also increasing complexity.
What makes AI networks fundamentally different from traditional networks?
AI introduces multiple coexisting designs and workloads, increasing infrastructure complexity.
There are different types of compute for training, inference, and traditional applications. This creates a fabric where multiple architectures must work together. Unlike previous waves like software-defined networking or disaggregation, this shift is fundamentally changing how networks are built and operated.
How should leaders think about ROI and future-proofing in AI networking?
Designing open and standardised networks is key to achieving long-term ROI and avoiding lock-in.
As AI adoption grows, cost models such as token serving and application cost will become critical. Decisions made today will directly impact future costs. Standardisation allows faster adoption of new hardware, better price performance, and the ability to adapt as the ecosystem evolves.
Checklist
- Focus on software standardisation
- Avoid dependency on single-vendor systems
- Enable faster hardware adoption
- Design for long-term cost efficiency
Is AI networking only relevant for organisations already deploying GPUs?
AI readiness starts now, even if GPUs are not yet deployed.
Applications are already evolving and becoming AI-driven. Traditional networks cannot absorb the scale and complexity required. Organisations that delay preparation risk being left behind as infrastructure and workloads shift rapidly.
Key Takeaway
AI is transforming networking into a core strategic layer. Leaders must engage early, design for openness, and align infrastructure with future workloads. The decisions made today will determine scalability, cost efficiency, and competitiveness.
The AI wave is accelerating rapidly, with infrastructure, applications, and compute evolving together. Networking will remain at the core of this transformation, enabling integration across GPUs, storage, and software. Organisations that prepare now will be better positioned to adapt as new applications and architectures emerge.
Conclusion
AI is reshaping how networks are built and operated. Networking is becoming the foundation for scaling compute, applications, and infrastructure. Leaders who engage early, embrace openness, and design for the future will be better positioned to navigate this shift.
Frequently Asked Questions
1. Why has networking become a central part of AI infrastructure conversations?
For years networking sat quietly in the background of technology decisions. AI changed that. With GPUs and XPUs now driving compute at scale, the network has become the unifying layer that connects and scales those systems. Data center networking is expected to grow significantly, and the complexity coming with it means network decisions now carry the same weight as compute and software decisions.
2. How are AI networks different from traditional networks?
Traditional networks were built around a single architecture serving general workloads. AI introduces multiple coexisting designs at the same time, covering training, inference, and traditional applications all running together. These different types of compute need to work across a shared fabric. Unlike past shifts like software-defined networking or disaggregation, AI is fundamentally changing how networks are built and operated from the ground up.
3. What should leaders prioritise when designing networks for AI workloads?
Four things matter most: standardising software, avoiding single-vendor dependencies, enabling faster hardware adoption, and designing for long-term cost efficiency. Open and standardised networks give teams the flexibility to adopt new hardware faster, get better price performance, and adapt as the ecosystem evolves without being locked into one vendor's upgrade cycle or pricing model.
4. How do network decisions made today affect future AI costs?
As AI adoption grows, cost models like token serving cost and application cost will become the metrics organisations are judged on. The infrastructure decisions made now directly shape those future numbers. Teams that standardise early can adopt better hardware faster and keep operational costs predictable. Teams that lock into proprietary stacks today will pay more to change later when workloads shift.
5. Do organisations need GPUs deployed before starting to prepare for AI networking?
No. AI readiness needs to start before GPUs arrive. Applications are already becoming AI-driven and traditional networks cannot absorb the scale and complexity those applications will require. Organisations that wait until GPUs are in place to think about network infrastructure risk falling behind as workloads shift faster than a reactive approach can keep up with.
6. What does moving from one network to multiple networks actually mean for IT teams?
It means managing more complexity at the same time. AI factories need separate fabrics for training, inference, and traditional workloads, each with different performance and latency requirements. IT teams that used to manage one unified network now need to design and operate multiple architectures that still need to work together. Getting ahead of that complexity early, rather than retrofitting later, is what separates teams that scale smoothly from those that struggle.
7. Why is openness so important when building AI network infrastructure?
The AI ecosystem is evolving fast. New hardware, new architectures, and new applications are arriving constantly. Proprietary stacks slow teams down because every change has to go through a single vendor. Open networks let organisations adopt new silicon faster, integrate across GPUs, storage, and software without vendor friction, and stay competitive as the landscape shifts rather than waiting on a vendor roadmap that may not match their needs.
Explore how your network strategy aligns with the AI era and start preparing your infrastructure for the next phase of growth.
