AI Isn’t Just Changing Infrastructure - It’s Exposing Its Limits

After discussing the changing fiber market on The Route to Networking podcast and writing our previous article, Has Fiber Reached Its Turning Point?, there was another part of the conversation that kept sticking with us afterwards.

A lot of the discussion with Andrej Danis and Kanishk Raghuvanshi from AlixPartners focused on where the industry is heading operationally. But underneath almost every topic was a much bigger theme around infrastructure itself and the pressure AI is beginning to place on it.

The conversation kept moving between power, connectivity, latency, edge environments, fiber density, and deployment challenges. And the more Andrej and Kanishk spoke about the realities operators are starting to face, the more it felt like the industry may still be underestimating how much physical infrastructure AI is actually going to require over the next decade.

That’s what led naturally into this article.

After hearing their perspective on where digital infrastructure is heading, it felt worth digging further into the infrastructure side of the AI conversation. Not necessarily the AI tools themselves, but the systems underneath them and whether the industry is truly prepared for the scale of demand that may be coming.

Because one thing became pretty clear throughout the discussion: AI isn’t just accelerating infrastructure demand. It’s starting to expose where existing infrastructure models begin to struggle.

AI Growth Is Starting To Collide With Physical Reality

One of the more interesting parts of the conversation was how often infrastructure limitations came up.

Not software limitations or AI model capabilities, but physical infrastructure constraints that operators are already beginning to run into.

• Power availability.
• Connectivity.
• Latency.
• Fiber density.
• Deployment timelines.

As AI adoption accelerates, those challenges are becoming harder to ignore.

At one point during the discussion, Kanishk made a comment that probably summarizes the current state of the market better than most industry reports:

“Infrastructure is becoming increasingly the limiting factor.” 

And honestly, that feels like where the industry is heading.

For years, digital infrastructure was treated as something that naturally scaled behind technology demand. More users meant more capacity. More traffic meant more fiber. More cloud adoption meant more data centers.

But AI feels different because the scale requirements are arriving so aggressively, and across so many parts of the infrastructure stack at once, that infrastructure itself is becoming part of the problem the industry now needs to solve.

And it’s not just about building more data centers.

• It’s power grids struggling to keep up with demand.
• It’s delays around deployment and permitting.
• It’s the need for lower latency environments.
• It’s how all of these systems work together operationally.

That feels like a much bigger shift than simply “AI needs more compute.”

Connectivity Is Becoming More Complex

One of the smartest points Andrej made during the episode was around what he described as “ubiquitous connectivity.”

Historically, infrastructure conversations tended to separate technologies into categories:

• fiber 
• wireless 
• satellite 

But AI workloads may force those systems to work together much more closely than they ever have before.

As Andrej explained:

“You need fiber connectivity, wireless connectivity, and satellite connectivity.” 

That feels important because it changes the conversation around resilience completely.

For a long time, network infrastructure was largely judged on speed and coverage. But AI-driven environments rely much more heavily on synchronization, real-time processing, uptime, and latency management.

And that creates a different type of infrastructure ecosystem entirely.

• Fiber alone isn’t enough.
• Wireless alone isn’t enough.
• Satellite alone isn’t enough.

The infrastructure stack becomes interconnected.

And, it feels like the industry is only just beginning to fully appreciate how complicated that ecosystem could become over the next decade.

AI May Finally Make Edge Infrastructure Matter

Another really interesting part of the discussion centered around IoT and edge infrastructure.

For years, edge computing has been treated as one of the next major shifts in digital infrastructure, but realistically, it never fully scaled the way many people expected.

Andrej acknowledged that directly during the conversation:

“The promise of IoT didn’t materialize.” 

But what stood out was that he didn’t frame that as failure. It sounded more like the industry simply hadn’t reached the point where edge environments became operationally necessary.

AI may now be changing that.

During the discussion, Andrej started talking about physical AI sensors, distributed systems, and environments that require extremely fast response times between devices and compute layers.

That creates a very different infrastructure challenge.

Because suddenly, not every workload can sit inside a massive centralized hyperscale facility. Some applications require compute much closer to where activity is actually happening.

And that raises much bigger questions around:

• where compute sits 
• what gets processed locally 
• what moves back to centralized environments 
• how much latency becomes acceptable 
• how infrastructure gets distributed physically 

It’s difficult not to feel like the industry may be entering a second attempt at edge infrastructure, but this time with a much clearer reason for why it’s needed.

The Physical Infrastructure Behind AI May Become The Most Valuable Part

One of the comments that probably stuck with us most came when Andrej said:

“I would invest in the ducts.” 

It’s such a simple line, but it says a lot about where infrastructure thinking may be heading.

Because while most of the market attention today sits around GPUs, hyperscale growth, and AI models, Andrej’s point was really about ownership of the physical infrastructure underneath everything else.

• The pathways.
• The transport layers.
• The foundational infrastructure future connectivity depends on.

This feels like something the industry may still be underestimating.

• AI increases traffic.
• Traffic increases fiber demand.
• Fiber demand increases density requirements.
• Lower latency creates pressure for new infrastructure environments.

Eventually, the physical systems underneath the AI ecosystem become just as strategic as the compute itself.

And that feels like a major shift in how digital infrastructure may be valued moving forward.

The Industry Still Feels Early In Understanding What AI Requires

One thing this conversation reinforced is that the AI infrastructure discussion still feels incredibly early.

Most public conversations around AI focus heavily on the applications themselves:

• AI assistants 
• automation 
• enterprise adoption 
• large language models 

But this discussion felt much more grounded in the infrastructure reality underneath all of it.

Because eventually, AI growth stops being theoretical.

And when that happens, infrastructure becomes impossible to ignore.

• Power availability matters.
• Connectivity matters.
• Latency matters.
• Physical deployment matters.
• Operational execution matters.

And honestly, that may end up being one of the biggest infrastructure stories of the next decade.

Not just how quickly AI develops.

But whether the systems underneath it can realistically keep up.

How Hamilton Barnes Can Help

Hamilton Barnes works with organizations and professionals across AI infrastructure, telecommunications, networking, cloud, and digital infrastructure globally.

As demand for AI-ready infrastructure continues accelerating, organizations are increasingly looking for talent across:

• Network engineering 
• Fiber infrastructure 
• Data centers 
• Cloud and connectivity 
• Edge infrastructure 
• AI infrastructure operations 
• Digital infrastructure leadership 

We support both businesses and professionals through:

• Specialist recruitment support 
• Career guidance 
• Resume and interview preparation 
• Salary benchmarking 
• Market insight 
• Access to specialist opportunities globally 

If you would like to speak with our team about your hiring plans or career goals, you can get in touch directly.

• Contact our team
• Looking to hire? Upload a Vacancy
• Looking for your next role? Upload your CV

Want to hear more from The Route to Networking podcast?

• Follow our LinkedIn page
• Find our podcast on Spotify
• Find our podcast on Apple Podcasts
• Read all of our podcast Articles

FAQs

Why is AI creating pressure on digital infrastructure?

AI workloads require significantly more power, connectivity, compute capacity, and low-latency environments than traditional workloads. That is increasing pressure on the infrastructure systems supporting them.

Why is connectivity becoming more important for AI?

AI environments rely heavily on real-time processing, synchronization, and constant uptime. That requires highly resilient connectivity across fiber, wireless, and satellite infrastructure.

What is edge infrastructure?

Edge infrastructure refers to compute environments located closer to where data is generated, helping reduce latency and improve response times for applications that require real-time processing.

Why are physical infrastructure assets becoming more valuable?

As AI demand grows, ownership of fiber routes, ducts, transport networks, and connectivity infrastructure may become increasingly strategic because future AI environments depend heavily on those foundational systems.