What’s hot can come and go just as quickly as the wind changes. From planking to fidget spinners, skinny jeans to Gangnam Style, the same can be said for the tech world where a concept or technology can rule the roost before being dropped like Woody in that scene in "Toy Story 2."

Big data, 3D TVs, and pictures of board apes are just some of the myriad hot topics that had their time in the sun. Just this year, Meta shuttered its entire metaverse division despite spending a staggering $80 billion on the project and centering its entire rebrand from Facebook on the concept.

But one of those areas that didn’t quite follow the other’s lead was IoT. While the initial "smart everything" hype of the late 2010s seems to have given way to the unstoppable juggernaut of AI, IoT has, for the most part, steadily developed, largely becoming an almost unnoticed layer of the infrastructure stack.

During Nvidia’s recent GTC event, amidst all the fanfare about OpenClaw and low-latency "agentic ready" computing, it was surprising to uncover talk that agentic AI could be the thing that, after years of hype and disappointment, unlocks real value in IoT.

Leading this argument was Shawn Hakl, head of AT&T’s business products team. And who better to hear it from than an operator with one of the largest IoT networks in the world, recently topping more than 200 million potential customers covered.

On the resurgence of IoT, Hakl told SDxCentral: “As you think about AI and agentic AI, you can simplify that flow. … If you make the communications element and the security elements agent consumable, coupled with the use of agentic AI tools for workflow, you can capture more knowledge, you can scale it, and you can execute it much more automated than you could before.”

The difference between that initial wave of making everything smart only to subsequently bust, to a potential IoT resurgence today, in Hakl’s view, owes to AT&T being that little bit friendlier when it comes to playing with others.

“Before AT&T and most of the industry overall, we tried to do it all ourselves. Like, if I want to do mHealth (mobile health), I'd hire doctors and nurses. Pretty tough to justify that,” the exec said.

“You’re going to see AT&T take a much more friendly stance toward partnering with the hyperscalers, the neocloud[s], the LLM (large language model) makers, because if I harmonize policy from the edge of the device, I have packing control of the converged network into these facilities," Hakl explained. "I've got a customer who can go straight from the device into the compute facilities of their choice. They can control the security. They control the policy. I just make that available as long as I'm smart enough to harmonize my implementation with the endpoint provider, which is what we want to do.”

That harmonized, playing well with others approach has already seen the carrier team up with Cisco and Nvidia to support network-driven edge AI for enterprises.

Specifically, the trio unveiled a solution that centers on AT&T's dedicated IoT core and radio access network (RAN), paired with Cisco's Mobility Services Platform and AI Grid to provide a secure, end-to-end pathway from edge devices into Nvidia-accelerated compute for on-demand AI inference

“I have the edge device. I have a RAN. I can take that packet, put the deterministic routing around it. I can make it a private connection with policy controllers built in. I can transport that to Nvidia compute that sits at the edge of the network. And then our friends at Cisco bring in the orchestration and the packet core layer, which is the intelligence layer that lets us hook all that up," Hakl explained.

"Our friends at Nvidia also bring a developer ecosystem and the software, so I make that consumable," Hakl continued. "And then you've got really smart people, like the application folks we've got that build stuff at the edge, while Cisco helps us glue all up together network intelligence and the orchestration.”

By playing ball via a more ecosystem approach, the AT&T exec outlined an approach that brings the best of each player into one converged stack, which he surmised: “If it makes it easier and it's faster, good.”

From buzzwords to factory floors (and cars)

But rather than announce something at GTC and not have anything to show for it, AT&T worked with its newfound partners – along with computer vision solutions provider Linker Vision – to deploy their offering to process real‑time video streams across a four-block radius as part of AT&T's Discovery District in downtown Dallas.

Another early trial of the conjoined offering sees TanMar, a support company for the oil and gas industry that sets up and runs remote worksites, implementing the network technology to support AI-augmented video surveillance at its Louisiana facility.

While every technology has its cycle, Hakl argued that more “sophisticated” use cases like these and working with partners in such a way will help the promise of IoT scale.

“We're taking all of the stuff that we know today, but also being humble enough and smart enough to listen to our technology partners and approaching the innovation developer ecosystem, not as a ‘hey, I'm going to control everything and deliver it,’ but ‘let's go build. Here are some capabilities. We're going to listen to you and expand our offerings so that you can create innovation on top of that,’” Hakl said.

That ecosystem mindset is already paying off in more mature IoT domains too. Hakl noted that the combination of AT&T’s network and Cisco’s technology already powers around 80% of connected cars on the road today, and said his goal at events like GTC is to “find five or six more Linker Visions” – specialist partners whose applications can ride on top of that converged, AI-ready infrastructure.

“Their technology is amazing. They've been great to work with. I'd like to find five or six more with all sorts of ideas that I can't come up with," Hakl said.

Not every grand vision of autonomous agents will make it off the slide deck and into production. A lot of people are considerably excited about the promise of OpenClaw, just as they were excited about IoT back in its heyday.

But the difference this time, Hakl argued, is that IoT is no longer being sold as a standalone buzzword, but instead turning into a part of a wider infrastructure stack, offering a pervasive layer of connectivity and sensors that agentic AI can finally put to work.

If AT&T and its partners can make connectivity, security, and policy, as he put it “agent‑consumable,” the next IoT boom may not be branded as IoT at all, but will just look like smarter factories, safer worksites, and cars that have been connected for years, but suddenly finding themselves learning a few new tricks.