- The successor to the LAN, WAN and MAN is IAN — the Intelligent Area Network
- Vendors are developing AI architectures that decouple the concept of networking from geography and scale — and help drive revenue
- I’m calling on the industry to help build a complete understanding of the new Intelligent Area Network market
"AI is the most significant and transformative technology we’ve ever seen," says Bill Gates, and who would argue with him?
However, along with the AI transformation comes complexity and risk, particularly for the carriers tasked with creating the digital infrastructure to support artificial intelligence.
Currently, the communications industry and Big Tech are hyper-focused on solving the most obvious of these challenges: building massive data centers that can house the immense processing power required to expedite training and inference of artificial intelligence models.
But that’s just one piece of the puzzle. The network that connects the data centers also needs to be entirely rethought and rebuilt.
Well, friends, I bring you glad tidings of great joy: All of the world’s leading vendors — plus HPE — are independently working on integrated solutions that will address the AI network conundrum.
Enter IAN — the Intelligent Area Network, and successor to LANs, WANs, and MANs.
Ciena, Cisco, Ericsson, Huawei, Juniper, Nokia and ZTE are all developing IANs to address the AI network challenge, employing a diverse array of next-generation technologies and architectures. Understanding the differences between their strategies must be job No. 1 for carriers; their survival depends on it.
What’s the problem?
Until now, comms network design has followed the same rule as real estate: location, location, location. Local area networks (LANs), wide-area networks (WANs), and metropolitan area networks (MANs) use software, hardware, and best practices that are designed to handle the data being carried at the location they are deployed.
AI traffic is different and has no intention of adhering to the rules that have defined the transmission of data over networks for the past 40 years. In addition to being the most critical traffic traveling over carrier networks, AI is also the least predictable and most demanding.
For network architects, AI’s tumultuous traffic profile — driven by real-time inference workloads, huge data backups, autonomous systems, and edge AI — is a worst-case scenario. AI traffic patterns are highly variable, ranging from tiny, time-sensitive inference exchanges of just a few bytes that must be delivered within milliseconds, to elephant flows of hundreds of gigabytes or even petabytes that can dominate the network for several minutes.
Traditional LAN/WAN/MAN designs can’t handle the AI, and trying to ameliorate the problem by throwing chunks of raw capacity at it using the latest expensive 400- and 800-gig optical technology improves the issue but doesn’t fix it.
Economies of scale
And from a business perspective, it’s a terrible deal for carriers.
That’s because the 20th century telecom economic model is dead. Carrier revenues no longer scale in proportion to the volume of traffic that they carry. “Data value” is now where the big money comes from, not data haulage.
Today, the most valuable data carries AI applications and services, which is almost entirely generated by over-the-top (OTT) providers and hyperscalers who earn billions in profits from it. Carriers, on the other hand, are currently viewed as the steady Eddies of the industry. Data utilities provide a large enough pipe to quickly and reliably transfer AI traffic between hyperscaler data centers, but do not directly profit from the information carried within the data itself or its value to the companies sending it.
And, of course, they are also expected to bear the cost of significant infrastructure upgrades when network demands increase. And that bill is growing larger by the day; AI traffic is increasing at approximately 50% annually but is expected to rise by a multiple of that in the next few years, with much of the growth coming from AI bots crawling the internet for LLMs.
An intelligent answer
Tackling the AI revolution requires a revolutionary approach — one that decouples the concept of networking from geography and scale, while also integrating a strategy to generate revenue end-to-end from AI traffic, regardless of its source.
That’s precisely what all of the world’s leading communications vendors are doing. Cisco’s Agile solution will be a centerpiece of its Cisco Live event in San Diego next week.
Juniper timed the market with the launch of its Mist AI-native architecture in January 2024, based on technology it acquired when it bought Mist Systems in 2019. The fact that it took five years to integrate it into its portfolio speaks to the complexity of the IAN mission.
There are significant differences in the approaches taken by vendors to delivering their IAN, but they also have three attributes in common:
- They all provide multi-domain solutions that enable end-to-end control over network traffic and activity, from the AI source in customer premises, factory floors, or data centers, all the way through the cloud.
- They come with the tools and support to solve the carrier’s existential crisis by diversifying revenues, both by identifying OTT and hyperscaler AI traffic and charging a premium for it through service-level agreements (SLAs), and by creating new AI-enabled digital services via API-plus-developer ecosystems, premium APIs, and innovative billing models.
- They fight fire with fire, embedding AI ubiquitously throughout the carrier network to manage the surge of AI traffic using AI-enhanced monitoring, routing, automation, anomaly detection, and predictive maintenance, thus enhancing network efficiency, performance, security, and user experience.
Item three is essential. As the global hyperbole around AI reaches ever more absurd heights, it is refreshing to see AI being put to use in a pragmatic and much-needed business case. AI will not cure cancer in the next 10 years, as Google has claimed, and there is currently no resolution to the meta-issue of how to power the AI revolution in North America, or to regulate it. However, with IAN’s help, it can absolutely provide carriers with a path to a profitable future.
IAN, meet AI. AI, meet IAN. I think you’ll get along.
A call to action …
Deciding which IAN to work with is the most critical decision carriers will face in this decade, determining their ability to diversify revenues and transcend the 20th century data utility role. But there are key differences between the approaches taken by each vendor in this market. So far, no one has created a sensical comparison of these products, based on consistent, quantitative information. Why don’t we do that, together?
I’m inviting all vendors working on Intelligent Area Network solutions to get in touch with me. But I’m also excited to talk to carriers about what you are looking for in your IAN. Feel free to drop me a line over here.
Steve Saunders is a British-born communications analyst, investor and digital media entrepreneur with a career spanning decades.
Op-eds from industry experts, analysts or our editorial staff are opinion pieces that do not represent the opinions of Fierce Network.