As the industry gears up for its biggest annual show at MWC, it is (perhaps) worth looking back at another major 2026 telco gathering: PTC in Hawaii in January. Once a regional telecom meet, PTC has evolved into a global forum for AI infrastructure, and for telcos of all sorts within it. Here were/are the talking points.
New economy – The AI boom is fueling enormous capital deployment, with the top five hyperscalers set to invest $602 billion in 2026 alone, while supply struggles to keep pace.
Tight spot – Telcos, totally critical, are getting the least share of the pie. But fibre operators and mobile carriers are responding, deploying long-haul and metro fibre, plus edge systems.
Inference time – AI workloads are shifting from central training to edge inference, affording telcos a chance to serve hyperscalers and enterprises, and maybe make some money again.
Ahead of the industry’s biggest annual shindig, at MWC in Spain, RCR Wireless figured it might be worth reflecting on another major (three-letter) telco event to size up the big infrastructure discussion that is impacting the whole industry, as it hurtles through 2026: PTC in Hawaii. As such, we have summarised some of the big talking points about how the broad tech market – data centres, fibre networks, mobile networks, plus the rest – is standing up the biggest economic goldrush since before time began: AI. So here are some talking points – from PTC, for MWC; to get a grip on the state of the market.
1 | AI is like nothing before – for AI players, anyway
AI, well shucks. “None of us have seen this before,” said Bill Barney, founder and chief executive at Asian Century Equity, in his intro address at PTC. He detailed its total economic revolution from the start: the big eight US tech stocks, all newly-geared around AI, were up 50 percent in 2025, and are propping up the global economy – to the point US growth would be flat, or marginally down, otherwise; the whole of geopolitics, practically, hinges on their fortunes, notably against their present and future equivalents in China.
Joe Valenti, managing director and co-head for media and telecom at Bank of America, kicked off another session. “The demand profile is enormous,” he said, opening up a discussion about the state of the data centre market in the US. He reeled off some top-line stats: the top five hyperscaler companies (Amazon, Microsoft, Google, Meta, Apple) will invest $602 billion of capital in data centre infrastructure in 2026; four of them will spend $100 billion each – “north of 50 percent” of their revenue in some cases. “It is a step function in historical capital deployment,” he said.
And they can’t go fast enough, it seems. There is virtually zero over-capacity; the vacancy rate for US real estate, whether powered shell facilities or fully fitted spaces, is just 1.6 percent in primary markets, he said – so only 1.6 percent of total data-centre inventory is currently unoccupied in historical AI-infrastructure heartlands. “It is well less than one percent in certain markets – Northern Virginia being the most obvious one. So we have this huge imbalance of demand relative to supply.” So demand is exploding and supply is nearly exhausted.
And AI is the reason, linked mostly to the task to train massive frontier language (LLM) models.
2 | Power is the bottleneck, of course – but so are people
Power is the problem, as everyone knows – and practically every panel at PTC discussed. Barney suggested a 20-30 percent gap in the US “between the power that’s going to show up and the power that’s needed.” He said: “So a big problem is how to accelerate that side of the industry.” Gas? Steam? Nuclear? Everything’s on the table, and nothing is straightforward. Microsoft’s new Fairwater data centre complex in Wisconsin uses more power than the city of LA – monster urban power centre, by any stretch.
There will be seven such campuses in the US by the end of 2026, and 19 by the end of 2027. “This is the beginning of the future,” said Barney. Anthropic has just built a data centre (somewhere) in “less than two years”, he said. “Which is unheard of… It requires a gigawatt of power, and the power industry can’t get to a gigawatt of power in two years. As an industry, we’re becoming highly efficient; the challenge is whether the infrastructure can keep pace.” Late December, Google’s parent, Alphabet, spent $4.75 billion on Intersect Power to bypass grid issues.
Barney reflected: “It is the first time a hyperscaler has bought a power company, and, I think, the first in a series of [such deals] over time.” But the problem is not just power. Speaking during the same session as Valenti, Jason Nance, in charge of client solutions at CBRE Data Center Solutions, a service group within commercial real-estate services and investment firm CBRE, suggested that the migration to secondary, and now tertiary, markets in pursuit of power has created another challenge: how to recruit people and placate communities.
The community issue is well known – as the drain on power impacts local energy prices, and protests sweep parts of the US. But there was much discussion at PTC about community engagement and education, and the more novel idea was just how to draw human power to back-woods markets. Nance said: “The biggest problem is getting people to move to these tertiary markets. People look at the power and the network, but… you still need people at the end of the day. And that may stymie growth in tertiary markets… Sometimes it’s a bridge too far to move to North Dakota.”
3 | Telcos are in a tight spot – but they are holding aces
In his opening keynote, Barney addressed the elephant in the room: telecoms. Because PTC, officially the Pacific Telecom Conference, was once a sunny new-year jaunt for the US telecoms industry, and is now squarely about the broader AI infrastructure market, and the role for telcos within it. “Twenty years ago, we only talked about networks – voice networks, then data networks. This is PTC, right? It was always about telecom. But the telcos have become a very small piece of the ecosystem. They’ve become a utility that transports [the data].”
He went on: “[And yet they] also have to be stronger, faster, and more powerful, [even as] telecoms gets less and less and less of the pie. [It’s a] very difficult spot. Fibre operators and mobile operators are seeing a capital crunch. Even though the tech industry is doing well, [even as] data centers [are] accelerating, it is a hard slog for them. It’s a fight just to keep operating, or even to grow a little… [Telcos] are getting the least amount. But they play probably the most critical role. If you don’t have the networks, you can’t transport the applications.”
He added: “The capital we’re spending doesn’t match what we need to get done… Subscribers aren’t paying what needs to be paid to build these networks.” It is a reality check, of course, and good to have spelled out. But the industry – the fibre industry, anyway – is responding. One hundred and seventy million miles of new fibre in long-haul backbone infrastructure, according to Zayo chief Steve Smith, speaking separately. Up to 60 million miles of new fibre will be deployed in metro areas. That is what it will take to serve AI over the next four years, reckons Smith.
It is quite the workload, so to speak; a “big pickup in our build-out”, he said. His firm started its rollout in earnest last year, he says, deploying “thousands of new fibre miles” just in the US. The company’s backhaul and metro “requirements” have moved from 12-24 fibres, typically, to 48-72, to support data traffic between data centres, mostly to train a handful of large language models to deliver generative AI on a handful of consumer-oriented platforms, which use static data in centralised compute clusters. But times are changing, as well – see the below entry about inference.
4 | So is the subsea sector – if you’re not a hyperscaler
Subsea fibre infrastructure was a major talking point at PTC, with a whole day given over to the subject on the Sunday. A major crunch is coming in the Atlantic, the oldest subsea-stronghold, and the “world’s highest traffic route”. But the system is at full stretch. Seven of its 21 routes will be retired “relatively soon”, said Nigel Bayliff, senior director for global submarine networks at Google. “That is a problem because that starts to reduce that reliability mesh” – meaning network resiliency from network diversity.
Plus, of course, the AI phenomenon is driving traffic demand through the roof. The industry has delivered, so far: capacity per cable has jumped by “279,000 times” since the first PTAT line was landed in 1988, said Bayliff. EXA Infrastructure carried four times as much traffic on its Atlantic channels in 2025 than it did in the previous three years combined (2022-2024), according to Jim Fagan, chief executive at EXA Infrastructure. “We’re seeing the growth, and I’d love to say it’s because we’re crushing it, but all the boats are rising.”
But while the industry has, necessarily, created ‘mesh’ value to pivot subsea lanes towards traffic flexibility, failover optionality, and geographic diversity, the system count has remained static over a period of time, and is set to decline. Its distributed landings have stretched the system. The industry has compensated for fewer routes by being smarter with where it lands them; the mesh masks fragility, in a way – until it is overwhelmed with AI-style demand, and it doesn’t, and the whole thing topples over. “More cables is definitely the answer,” said Bayliff.
More landings, as well – is the instruction. “However they’re built, whoever’s building them, more cables to generally the same location – so we can create these switching meshes – is the way forward.” But the point about who builds new crossings is a bone of contention. For Fagan at EXA Infrastructure, the question is not just theoretical: his firm owns the two newest “non-hyperscaler” routes across the Atlantic, and wants to build more. But for a decade now, hyperscalers have built most of the new trans-Atlantic and trans-Pacific subsea infrastructure themselves.
Since AI has started to spike uncontrollably, their agenda has shifted – to build for their own needs first, keep any excess for internal traffic, and only release it to the wholesale market as a last resort. Fagan stated: “We are facing a severe supply crunch in the Atlantic, probably late 2027. If you think about the [new] data center traffic and flows, it is going to create quite a problem. At this point, the question is how to make sure there is enough wholesale capacity for the rest of the market. And for us, we think maybe it is time for private cable again.”
5 | But even telecoms is not in as tight a hole as some
Barney described a developing “k-shaped economy” in his opening speech, where the tech industry is raking it in, hand-over-fist, while everything else goes south. “The rest of the industries are kind of going like this,” he said, pointing down. It is the AI industry’s defining challenge: beyond infrastructure-building, how to actually deliver value to raise up other industries. “Tech companies are accelerating, while the rest of the stock market… is declining. Bricks-and-mortar companies are not doing that well. And it’s not just the US; it’s around the world.”
Barney went on: “It is causing a lot of the challenges, not only with the economy, but also with politics… [Because] there are two groups: the people that are making money and the people that aren’t… Unfortunately, whether we like it or not, we are at the centre of a lot of the problems that are happening across the US and around the world – as caused by the economic situation… Most of the people in this room are happy because their stocks are going up. But a lot of people around the world aren’t so happy about where things are going.”
What else is there to say: AI needs to pay its way, and prove itself as more than a circular energy-busting economy, which lives in a bubble, and does little in the way of genuinely transformative good.
6 | Inference is the answer – for telcos, as the next buildout
Lynn Smullen, member of the PTC board of governors, opened a PTC session by suggesting that the mission to train AI foundation models will drive total global data centre capacity to 120-150 GW, “even 180 GW” – and that, as AI workloads shift towards inference tasks over the course the next couple of years, “as much as 90 GW of capacity will be at the edge”. This is a major opportunity for telcos, as AI works its way along their network arteries. Smullen said:
“And what’s critical is not just power, but the network, and its efficiency, resiliency, reach, and protection (security).”
New AI inference applications will bring the action closer to the edge, and into the enterprise space, and likely change network deployment strategies along the way. “We’re studying both [trends] closely,” said Smith at Zayo. “We are building nearly 20 new routes right now. We are seeing requirements for both; there’s still a lot of training that we’re all trying to help with, and inference is starting to show up because companies are working on use cases and POCs. [New fibre routes] take two to four years; these are big capital decisions, in growth corridors.”
Maybe the best session, from a traditional RCR point of view, was with Verizon Business, which made the case for well-adjusted telcos to serve both core network demand, geared towards hyperscalers, and metro and edge demand, to support enterprise usage of AI. “It will fuel growth for the time to come,” said Jeffrey Hulse, president of network and partner solutions in the firm’s wholesale arm. “The thing that gets us excited is how you push AI out to the prem. Because you need dense fibre at the edge for enterprise, public sector, and for (GPU/compute) sled opportunities.”
He referenced the firm’s so-called One Fiber project in the US to build a converged network for its wireless and wireline services. “We used to put 12 fibres down a street, wall to wall. When we started One Fiber, we were putting 864 down Main Street [and are now] putting 1,600 on the most important routes – and the use case is to put all of our macro cells on there. Fifty-one percent of our sites are now on-net…. Now we’re saying: how do we take those fibres, splice them, and go into enterprises… to give them private-line services, high speed ethernet, wave services.”
See more coverage from PTC here:
Power plays and digital divides – the hard realities of the AI build-out
Under-supply to over-supply – the generational AI build-cycle
The big dig – how edge AI inference shifts the fibre bottleneck
Atlantic crossings – crunch time for subsea cable in the AI era
Master and commander – Google and Meta chart AI’s subsea voyage
Verizon’s vision for metro fibre and private 5G for enterprise AI inference
