- A lot of money and energy are being poured into the AI layer of the network
- At the bottom: electricity and water utilities to fuel the revolution and keep everything cool
- At the very top are new applications and services
AI doesn’t exist in a vacuum. It needs a four-layer stack. At the bottom: electricity and water utilities to fuel the revolution and keep everything cool. Above that: the network, with connections that are both fat and fast (low latency). On top of the network: the AI layer. And sitting at the very top — like a magnificent top hat — are new applications and services, those with a solid business case or why bother with AI at all?
Right now, nearly all the money and energy are being poured into the AI layer, as hyperscalers and carriers engage in a full-scale AI arms race to build the largest data centers imaginable, packed with as many GPUs as possible — regardless of cost.
But without the other three layers, AI can’t deliver the economic revolution that McKinsey, Gartner, and the other big-name analyst firms are desperately over-hyping (at the same time as ChatGPT’s $200-a-month Deep Research service replaces them… now that’s what I call a killer app!).
The danger, particularly in the US, is that we end up with a potent AI layer — powered by American innovation, wealth, and hubris — with nothing above or below it.
I’ve covered the utility component of this situation extensively. It’s the weakest layer of the stack — an existential problem for the US economy.
President Trump could solve America’s utility infrastructure problems with a few of those executive orders he loves so much, but that would involve spending federal dollars, and federalism is verboten on his watch (except when it’s used to commandeer the National Guard to suppress protests in LA, obviously). Also, this is a long-term project, which is less exciting to him and his Looney Tunes cabinet than brokering peace between Iran and Israel.
I’m less worried about the network layer of the AI stack now that incumbent vendors have begun rolling out infrastructure solutions explicitly tailored to handle AI traffic demands.
What's in the application layer?
What about the application layer sitting atop our AI layer-cake? Well, it depends on what you’re after. If you run an enterprise or carrier network and your goal is eliminating headcount, boosting efficiency, and cutting OPEX, then, my friend, you’re in luck.
But if, on the other hand, it’s new and diversified revenues you seek, then, um, can I interest you in an annoying and generally useless AI customer service chatbot?
It’s not that AI’s ability to support or birth new applications and services in carrier and enterprise networks is in doubt; it’s that the shape, timing, and nature of those opportunities are still largely a mystery (analysts Leonard Lee and Patrick Kelly are exceptionally knowledgeable on this topic; see my conversation with them here).
Alianza is one of the few companies currently offering real examples of new services that can actually move the carrier revenue needle. It uses AI to level-up carrier voice traffic with bells and whistles consumers will happily pay for. Clever.
Most other vendors are still at the “if we build the framework/ecosystem, the revenue-generating applications will come” stage of development.
The arrival of agentic AI
Within this context, the arrival of agentic AI, with its own hype microbubble, is an unnecessary distraction. Carriers need to follow the money, but at the recent DTW show in Copenhagen, agentic AI was all most vendors wanted to talk about. And sure, I get why autonomous agents doing autonomous agent things are important, but in the short term, they’ll be more expensive, harder to manage, and potentially more prone to making catastrophically ill-advised decisions than first-generation LLMs.
For the rest of this decade, the real opportunity for AI isn’t in enterprise or carrier networks. It’s in heavy industry, where the combination of AI, automation, and robots is utterly transformative. I’ve seen real-world examples firsthand — and poked them with my virtual HB pencil of reporting. They’re impressive, and the United States is years behind other developed nations in deploying them.
The four-layer AI stack, which I invented specifically for this op-ed — you’re welcome — is immutable. Without all four levels in place, AI isn’t going anywhere fast.
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.