- TM Forum CEO Nik Willetts isn’t sold on the edge business case
- He warned operators not to get distracted by the edge and instead focus on getting “fit for AI”
- Willetts also highlighted China’s leadership on network autonomy and dished on why operators there are so far ahead
DTW IGNITE 2026, COPENHAGEN, DENMARK – As the promise of AI inference looms, there’s been a question lurking in the shadows of telco conversations: Is the edge opportunity really real this time? For his part, TM Forum CEO Nik Willetts appears to be firmly in the skeptic camp.
“Our position right now is definitely don’t bet the farm on it, and we’re skeptical,” he told Fierce. “Until we see the use cases, we think everything needs to be based on actual demand.”
According to Willetts, the problem with the edge is threefold. First, there’s the question around use cases. While there are some niche physical AI use cases where edge processing might make sense – in manufacturing, for instance – Willetts said these would likely be served by on-prem deployments.
Second, there’s the simple question of whether it makes economic sense to pursue edge compute deployments.
“It to me is unlikely that the economics add up for us to provide GPUs at any kind of scale in most countries – there are certain exceptions to that – and that the need for the customer can’t be solved through other means,” he said. Not only is it hard to match the scale of what hyperscalers can do, but telcos would likely be further down in the GPU procurement line. And in any event, GPUs could soon evolve to be able to deliver compute capabilities on-device (hello, Microsoft and Nvidia PC collab), negating the need for edge computing.
Which leads into problem 3: timing. Willetts said customers who might be interested in edge computing would likely want the ability to leverage global capabilities. The question is whether telcos can build out that kind of infrastructure in the time it will take for potential use cases like physical AI to take off.
His concern is that telcos could get distracted chasing a phantom opportunity and be sidetracked from efforts to make themselves and their networks “fit for AI.”
Autonomous levels and leaders
Willetts also sounded off on the industry’s progress on network autonomy in his interview with Fierce, noting half the battle to date has been proving that autonomous operations are even possible.
“We’re proving things like it’s way more reliable than the traditional human way of doing it,” he said. “Somewhere around 70% of faults in network that are introduced are introduced by human error.”
TM Forum’s recent assessment found that most operators today have achieved between Level 2 and Level 3 autonomy. Willetts noted that the level scores aren’t for an operator’s entire network but rather network domains or processes. And while some leading operators have validated scores in the high Level 3s, most of the market “is fairly behind” with scores in the low Level 2s.
Asian operators, and more specifically those in China, are farthest ahead. Asked why there’s such a large gap, Willetts said Chinese operators have benefitted from some “systemic advantages.”
For starters, the country had an earlier commitment to AI, and Chinese telcos also dominate what in the U.S. is known as the hyperscale cloud space. Thus, they’ve been early to seeing the benefits of compute and the network working together.
Willetts added there are also fewer “human barriers.” Western countries tend to have more unions and concern from individuals about the impacts of AI and autonomous operations on their roles. That’s not necessarily the case in China, which is more open to the idea of human roles changing.
That kind of mindset shift is something telcos the world over will need to adopt as they shift from selling access to selling network outcomes (i.e. performance-based services that generate incremental revenue).
“There’s a large volume of human beings running networks today who are doing repetitive mundane work, and that’s not really adding the human value,” he said. “In the world we’re talking about where we’re selling outcomes, we need humans to come up with those ideas and feed them in, not to be typing in command line codes to the networks to reconfigure them. So, it’s not that the humans aren’t needed, it’s that they’re doing something that’s more valuable.”