The Rise of AI-Native Global Networks

In this episode, Gary and Scott explore how global telecom networks are undergoing the biggest architectural shift since the birth of the internet. 

For decades, carriers have delivered connectivity as a static, reactive utility. But AI workloads are fundamentally breaking traditional network designs. Large enterprises running global inference pipelines, real-time analytics, digital twins, and distributed training now require deterministic latency, workload-aware routing, and transport layers that can predict and self-optimize in real time. 

Gary and Scott explain why the next era of global connectivity will be defined by AI-native networks — intelligent, autonomous systems that continuously sense, anticipate, and orchestrate data flows based on model behavior, compute availability, energy conditions, and regulatory constraints. 

They break down how this shift will transform: 
• Enterprise architecture and global WAN design 
• Latency-sensitive AI applications and GPU cluster connectivity 
• Data governance and cross-border regulatory compliance 
• The business models and competitive landscape of Tier-1 ISPs 

Finally, the episode introduces a real-world blueprint of this future: the emerging partnership between Lumen, one of the world’s largest Tier-1 global networks, and Palantir, an AI-driven decision platform built for national-scale complexity. Gary and Scott explain how this collaboration hints at the telecom industry’s next decade — one where networks become intelligent participants in the AI ecosystem rather than passive transport. 

If you’re a CIO, CTO, global network architect, cloud strategist, or enterprise AI leader, this is a must-listen episode that will reshape how you think about connectivity in the AI era. 

This is the beginning of the intelligent network revolution — and the Macro AI Podcast will keep you up to date on the roadmap

About your AI Guides

Gary Sloper

https://www.linkedin.com/in/gsloper/

Scott Bryan

https://www.linkedin.com/in/scottjbryan/

Macro AI Website: 

https://www.macroaipodcast.com/

Macro AI LinkedIn Page:  

https://www.linkedin.com/company/macro-ai-podcast/

Gary's Free AI Readiness Assessment:

https://macronetservices.com/events/the-comprehensive-guide-to-ai-readiness

Scott's Content & Blog

https://www.macronomics.ai/blog

Transcript

00:00
Welcome to the Macro AI Podcast,  where your expert guides Gary Sloper and Scott Bryan navigate the ever-evolving world of artificial intelligence.  Step into the future with us  as we uncover how AI is revolutionizing the global business landscape  from nimble startups to Fortune 500 giants.  Whether you're a seasoned executive,  an ambitious entrepreneur,

00:27
or simply eager to harness AI's potential,  we've got you covered.  Expect actionable insights,  conversations with industry trailblazers  and service providers,  and proven strategies to keep you ahead in a world being shaped rapidly by innovation.  Gary and Scott are here to decode the complexities of AI  and to bring forward ideas that can transform cutting-edge technology  into real-world business success.

00:57
So join us,  let's explore, learn and lead together.  Welcome back to the Macro AI podcast. I'm Gary Sloper and with me is my good friend and cohost Scott Bryan. Today we're stepping into a topic that honestly, Scott, you and I have been talking about and waiting for years  to discuss this context. Yeah, we really have. This is one of those episodes where our combined experience,  know, global networking, telecom infrastructure.

01:27
large scale systems engineering,  and now we have artificial intelligence and all of those really finally converge into a single story.  And that's the big story that it'll take some time to unfold, but it's already moving. And it's the story of how the entire architecture of global networks is about to transform into something really fundamentally new and different. Back when you and I were designing the first internet networks during the dot-com boom,

01:56
I think we were configuring Cisco routers, provisioned with T1s, which for those of you that are not familiar with the T1, it's a meg and a half of bandwidth. Yeah, not quite broadband yet. And some of the fat pipes that we were selling for businesses were DS3s at 45 megabits per second. And then you had some of the early tier one ISPs like UUNet had an OC12 core, which is only 622 megabits per second. And that was their...

02:25
poor network as a tier one ISP. So we can go on and on with stories about how the global telecom networks continue to grow,  you know, year over year and expand over the last few decades.  would be a long story. Yeah, no doubt. for decades, the network has really been the silent partner of the digital world.  If you think about this, it's carried everything from data, applications,  video, voice,  even financial transactions. Yet it never really

02:55
thought never really had a mind of its own. It never predicted. It never adapted in real time. It was just plain transport. Yeah, definitely. It was just, just transport.  And here we are now  in 2025 producing an AI podcast and talking to our clients about AI workloads and, AI workloads, they just don't tolerate reactive only networks. They really demand networks that are intelligent, self-optimizing, latency aware, uh driven by prediction.

03:25
So really networks that that behave more like living systems than a static point to point or static infrastructure. And that's exactly what we're diving into today. The evolution of global telecom networks  into AI native, intelligent driven systems that will ultimately and fundamentally  reshape how enterprises architect their strategies today. Yeah. So I think let's, let's start by just kind of stating the obvious.

03:55
For the last 30 years, the global internet has basically been the same model. So obviously bandwidth grew exponentially  and yes, more routes were added and latency got better. ah There were better optical systems installed for uh multiplexing, more subsea cables, more internet exchange points all over the world. So, and eventually there were multiple data centers in every city. The growth of the cloud hosted model in SaaS  exploded over the last decade. uh

04:23
And then of course, you can't forget about the wireless internet working and internet things. But really the architecture of it all  didn't really change that dramatically. just expanded. Yeah, you're so right. I mean, if you think about the internet, it was really designed for email or file transfers or really static content. Even streaming video, which today drives enormous traffic, was layered on top of a model never intended for

04:52
think about millisecond sensitive compute synchronized workloads. I mean, it wasn't that long ago. I'm probably dating myself, but you sat there and waited with your AOL dial up for an image to download.  I can hear the noise now. ah But yeah, and so now we're looking at enterprise scale artificial intelligence and it's it's it's already happening. You and I are both consulting for clients on this these types of architectures. So just imagine you're a global bank

05:22
running inference  on a fraud detection model across five continents. You can't have your inference latency rising from 15 milliseconds to 70 milliseconds during East Coast peak traffic. ah Or just another example, if you took  a biomedical company that's training a new drug discovery model across 12 GPU clusters distributed globally,  your inter cluster synchronization would, if it slips even slightly, ah

05:52
entire training efficiency just really goes in a hole. Yeah, exactly.  The networks we built for cloud and SaaS can't support AI at scale, not reliably, not predictably, not efficiently. Think about it. Yeah, and I think most enterprises already know this,  the ones that are starting to think about what they need to be achieving.  And even if they don't have the  industry jargon,

06:18
to articulate what they mean by that. They're feeling it in the performance of their AI pipelines. They're feeling it in connectivity to GPU clusters.  They're feeling it in the unpredictability of  training jobs. Yeah, agreed. I mean, we should probably talk about what large multinationals are actually experiencing today because many enterprises are seeing network bottlenecks  like never before. ah We hear this weekly,  we added a

06:47
second GPU cluster in Europe, for example, our training efficiency dropped 30%. Another example, our inference traffic to our cloud model spikes during the workday and falls off a cliff at night. And the network just simply can't keep up. These aren't cloud problems. These aren't model problems. These are network problems masquerading as everything else. Yeah. That's a good way to put it.

07:15
I think AI is really forcing enterprises to treat the network as part of the compute stack now. So not adjacent to it, uh but really part of it.  And this is really new for CIOs. They've spent the last 15 years focusing on cloud migrations,  DevOps pipelines, modernization of apps. But now the plumbing underneath everything suddenly matters again. They need AI native core networks at scale.

07:43
Yeah. And we should probably define what an AI network actually is because this term gets  thrown around in marketing materials. I've seen it with carriers.  Um, but what we're talking about is not superficial. An AI native network is a network that changes state continuously, automatically, really in response to workload conditions, uh, you know, based on your environment and your overall topology. Yeah,  exactly. Yeah. So instead of, uh

08:13
operators logging into routers and making manual changes, configuration changes,  the AI native network makes those decisions itself. So it can anticipate congestion. can predict cross region demand patterns. It knows where GPU capacity is across the core, if it's increasing or decreasing.  And it even knows when energy pricing shifts and  workloads should be shifted around. Yeah, I think

08:42
for a lot of business leaders and even,  you know, DevOps teams and cloud engineering teams really need to think of the network as a living organism. So break it down this way. Every packet becomes a signal. Every flow becomes a context. Every latency spike becomes a predictive indicator. So if you think about that, you know, we probably should cover some characteristics of the future network that we've listed out so far, Scott. Yeah.

09:11
I think you just uh really summarized it pretty well there. I think the five characteristics  that we had listed out.  The future network is observant. That's characteristic one.  Really, deep telemetry across optical IP application and compute layers. The future network is observant. ah Two, it's predictive. has models that can forecast failures,  forecast congestion.

09:37
imbalance  and even regulatory violations because those are getting to be complex. ah Three was that we listed out was  self-optimizing. So automatically rewriting routing policies and traffic engineering. ah It's governance aware. So at the high level, countries with data residency rules will have enforcement at the transport layer. So that's a new concept. ah

10:05
And the last one, the future network is workload driven. So it's adjusting paths based on model requirements,  not just a  network utilization parameters. Yeah, that's  a great way to break it down. So if you were to summarize that list, AI native networks collapse the boundaries between cloud,  edge,  compute and transport. So there is no distinction anymore. It becomes a single orchestrated system. And I think

10:33
you know, anyone listening should be excited about this, especially if you were at the dawn of the early days of global networking, you know, everybody got fired up about, you know, the, the BGP protocol and how that could help optimize  and reroute traffic. So when you were talking about it's self optimizing, you know, and I think, you know, to use your words automatically rerouting routing policies and traffic engineering for those environments, that's where

11:01
I think teams will really start to see the benefit of this. Yeah. Layers of intelligence on top of BGP. ah Yep. So I think that's good summary.  Let's kind of pivot a little bit and talk about what this means for the Fortune 500 and Global 2000. So they're really, you know, they're the ones that are running the most, the world's most complex networks today. I think for them, uh the AI native network is a huge upgrade. So instead of predictable global routing, they get

11:30
Terministic paths, optimized for inference and training. Instead of siloed monitoring tools, they get unified visibility across all layers. And instead of manual capacity planning, you know, 40%, we'd need to install a new circuit. The network allocates itself dynamically. Yeah. I mean, this also redefines the role of the network architect, right? Instead of just designing circuits and

11:58
CPE, also known as customer premise equipment, you the routers or layer three devices that you have on site. They'll also be designing AI assisted workload fabric. So networks that respond to the needs of distributed compute, not the other way around. If you think about that. Yeah, definitely.  And then on the user side, which is really the key concern of CIOs that we talked to, know, the user experience is top of mind for them. Applications become faster.

12:26
collaboration tools become smoother,  analytics become instantaneous across the business. uh And then huge investments like investments in digital twins, they can actually update in real time. you know, complex simulations that they're running, which are gonna become normal,  they  don't stall out halfway through processing. Exactly, I mean, the network ceases to be the cost center, becomes a performance engine, instead of just that OPEX line item that many teams  have.

12:55
It's been really the bane of their existence. And in the wider network that realistically has been less of a concern at the CIO level for the last, I don't know, decade or so,  you know, it's now working its way back up to the top level of their attention  and  critical to their business. Yeah. 100%. ah Yeah. So let's jump into uh the introduction of  a critical idea that's really at the heart of this episode.

13:25
You can't just build an AI native network with routers and cloud services alone. You need something above the network  and intelligence layer. It really a system that integrates the telemetry runs machine learning models, enforces policies and automates decisions. Something that has a real time model of the complete environment. Yeah, exactly.  This, is the missing puzzle piece. Every carrier knows they need this layer, but historically they've

13:54
Lack the platform that can do it at massive scale with real world constraints, compliance,  routing complexity, multi-cloud deployments and global regulatory regimes.  So a Palantir like solution becomes the brain of the next generation network. Yeah, exactly. So, uh, you mentioned the Palantir like solution. That's the intelligence layer. So it ingests data from every layer in the network.

14:23
fuses it with context from applications and compute. It learns the patterns over time. It decides what to do next automatically. And it's not an add-on. It is the orchestration plane. Right. Right. And enterprises will soon be looking to their carriers to provide this layer as part of their service catalog. Yep. They won't, they won't settle for just straight pipes anymore. They'll ask questions like how intelligent is your network? What optimizations does it make for our GPU clusters?

14:53
Um, does your network enforce policy at the transport level? Can it predict bottlenecks before training jobs fail? So this  is the new frontier of tier one telecom competition and, and telecom execs are thinking about how to reinvent themselves. Um, this is a tall task, especially when you have a lot of legacy systems within the carriers, but it will be table stakes for any CIO. Yeah, definitely a tall task, just like any.

15:22
large organization,  legacy organization. And I think that big telecoms, big carriers are facing the biggest architectural shifts since MPLS was invented really even  more so. And I think that,  Gary, you and I would agree that they'll evolve into a kind of a three layer model. m the way that I think that this is gonna roll out, you'd have obviously won the physical layer, the fiber.

15:51
spectrum subsea cables, tier one routing edge, tier two, well, layer two would be observability and control layer. So that'd be, know, telemetry platforms, optical monitoring, real-time analytics. And then the third layer would be the intelligence layer. That'd be the brain. So AI models, data infusion, governance, at, at, at layer three. Yeah, I agree. And only a handful of carriers today even have early prototypes of this architecture.

16:21
So because of that, think the race is on, ah similar with previous carrier tech life cycles, whoever builds the stack first  will certainly set the global standard. Yeah. And get some of the  big global logos on the network too. Yeah. And we had,  we had Michael Reed, the CEO of Megaport. He was on the podcast just a few months ago  and he talked about how they've really built their massive global network from the ground up.

16:48
essentially free of the legacy systems that other major global telcos are burdened  over years and acquisitions, et cetera. Yeah, that's a good point. And I just gave a quote for light reading, if anybody subscribes to that,  on Packet Fabric's  AI-driven design quotation implementation environment. It's early stage, their first beta. And that to me is definitely a game changer when you're trying to design and quote

17:19
and implement workload connectivity between clouds or even third party data centers. So there's a lot coming out that I think will force tier one carriers alike to meet that demand of what customers are going to be asking for, which these two providers are providing. Because I do a lot of work with clients who need to solve for multi-cloud connectivity today. We've talked about that.

17:47
on episodes and clients are amazed at how easily we can provision services ah between cloud providers. For example, Oracle Cloud Infrastructure, OCI, and AWS, for example. ah Or even from their data center workloads to OCI and a GPU as a service provider or providers. So the network as a service model is clearly where things are headed.  I

18:13
think it's going to also force some of the tier two and CLEC providers to really take a hard look at  their core infrastructure and say, need to keep up. Right. Yeah, no doubt. um Yeah. And I think now that we've kind of painted the conceptual picture, let's talk about the partnership that's already been announced to modernize a big global telco.  And that's where Lumen Technologies, which is,  they rank

18:42
per KEDA as the world's largest tier one ISP. And they have roots that go back to Quest Communications and Level 3 and a number of other acquisitions. And so Lumen is making a big and expensive bet. Yeah, it is big and expensive. And for anyone not familiar, Lumen operates one of the largest and most important global backbones now. It's AS3356 or autonomous system.

19:11
is what AS stands for, for those of you that are not network geeks like us. And they've decided it's time to evolve from a telecom carrier into an AI native infrastructure provider. So in order to do that, they partnered with Palantir and we've talked about Palantir on the show. We had an episode or two on about them. And one of the few companies that are capable of building that intelligence layer is Palantir at the national and enterprise scale. So we talked about

19:41
this and I think it was episode 50. So if you're not familiar with Palantir, yeah, I think you should definitely go back and listen to episode 50. It'll give you a really good sense of who they are and what they're doing. think there's been some other specials like 60 minutes have covered them as well, but some really interesting technology that they have. Yeah. Yeah. So just to kind of reiterate that Lumen brings the global network, the edge footprint, physical and IP backbone.

20:07
Palantir comes in and brings  the intelligence, the decision-making frameworks, data fusion  and automation.  so together, their concept is that they're going to build a blueprint for the AI native carrier stack that we've been talking about in this episode. Yeah. And this is a partnership.  It's not an experiment. It's a multi hundred million dollar multi-year commitment. It signals to the industry, this is the future and we're doing it now. Yeah.  Agreed.

20:36
And if Lumen and Palantir succeed, I think  every major carrier will need to follow with something similar. So all the big tier one ISPs, AT &T, Verizon, Aurelion, GTT, British Telecom, DT, and so on,  all of them will require an intelligence layer to remain competitive for global enterprise networks. Yeah. And that's a lot of investment. I think- a lot of investment.

21:04
Yeah, so I think anybody that feels AI is just vaporware, ah you're going to see this in the telecom environment. And I think it's, you know, something that the industry has been asking for for multiple decades now. So ah it's really the beginning of the telecom arms race and the AI native network is really the next battlefield.  The carriers that adapt will have a big advantage in the enterprise market. And those that don't will rely and remain. uh

21:32
as transport utilities. So that's what I was mentioning before, some of these tier two or CLEC environments probably stay as transport utilities. It may be too expensive for them to make that leap. Yep. Totally agree. um Yeah. So just getting toward wrapping it up. I think takeaway would be the world is shifting from networks that carry data to networks that really understand data. Yeah. From pipes to intelligence. uh

22:02
and  predictive autonomous systems.  And the partnership between Lumen and Palantir is the first true example of this evolution really making it into the new cycle. If you go ahead up there and Google a little bit, you'll see. Yeah. Yeah. So if you're  a global business leader,  a network architect, cloud strategist, or really anyone shaping AI adoption, it's time to start planning for AI native connectivity.

22:29
and understanding what's out there and what tools you can be taking advantage of in your wide area networking. Yeah, exactly. as those offer offerings evolve, we'll keep breaking it down for you right here on the macro app podcast. Thanks for joining us.  Uh, please like and subscribe, share it with your network. I want to thank all of our listeners.  Never thought we'd get to thousands of downloads.  Uh, but  the show's really taken off. So thank you so much.  Yeah. Thank you.