EdgePoint’s Suresh Sidhu on Malaysia’s towers, 5G and sovereign AI

• EdgePoint has built close to 2,000 towers in Malaysia in six years — but its founder says the hardest part was never the engineering.
• As parent DigitalBridge heads into SoftBank’s hands, Suresh Sidhu argues foreign ownership matters less than most assume.

The hardest part of building a mobile tower in Malaysia is not the tower. It is being allowed to put it up.

Six years after founding EdgePoint Infrastructure, Suresh Sidhu is candid about what he underestimated, and it was not the engineering. Even with a federal licence from the industry regulator, tower operators in Malaysia must still win authorisation to build state by state.

That gap, between how the tower business looks from outside and how it actually works, runs through EdgePoint’s six years in Malaysia. The company started from zero in 2020, grew partly by acquiring smaller local operators, and now owns close to 2,000 sites in the country and roughly 16,000 across Malaysia, Indonesia and the Philippines, making it one of the region’s largest independent tower operators. 

Of its roughly 2,000 Malaysian sites, EdgePoint built about 500 from scratch, the rest acquired, and Sidhu measures that against the pace of the market as a whole. Nationwide, he reckons, the commercial market—setting aside government-subsidised rural projects—adds somewhere between 500 and 1,000 new sites in a typical year.

Set against that, building 500 over six years is, in his words, “not a bad share.” 

He is also quick to puncture the assumption that a tower is a standardised, repeatable product. “People never really know what the tower business is. Then when they know what it is, they think it’s the same thing every time,” he said. “Actually it’s like 90% the same thing every time. And the 10% that isn’t is a lot of work.”

That 10%, he explained, is where the effort hides: one site needs far more ground preparation than the next; another sits somewhere every neighbour objects and takes months to bring everyone onside; an operator suddenly wants the antennas facing a different way. “So your effort per site can be way different,” he said. “Some are like, you apply, you get it, no one objects. Some take months of just trying to get everybody on the same page.” The business, in his telling, rewards persistence and detail over scale alone—a grind of permits, relationships and patience rather than just pure technology play.

Understanding why that grind matters requires understanding what a tower company actually sells. Independent operators such as EdgePoint—”towercos”—own the physical passive infrastructure: the steel, the power supply, the land rights. Mobile network operators then rent space to hang their antennas, usually on long contracts. The economics turn on one number: the tenancy ratio, or the average number of operators sharing each tower. 

One tenant is a cost; two or three tenants paying rent against broadly the same fixed outlay is where the profit lies. It is why a towerco is valued more like a property business than a technology one, and why Sidhu thinks in decades, not quarters. “You have to have a 20, 30-year view,” he said, “because that’s how long the infrastructure lasts.”

A market that went from one network to two

That grind has just been reshaped by the biggest structural change in the Malaysian telco industry in years. On July 1, U Mobile completed the migration of its customers onto its own ULTRA5G network and exited its wholesale arrangement with the state-owned Digital Nasional Berhad (DNB), bringing the country’s second 5G network fully into being.

The shift caps a long policy reversal. Malaysia originally rolled out 5G through a single, government-owned wholesaler in DNB, a model meant to speed deployment and avoid duplicated infrastructure. After operators pushed back and concerns grew about depending on a single point of failure, the government announced in May 2023 that it would move to a dual-network model, and formally revoked DNB’s single-wholesale status at the end of 2024. 

U Mobile, the smallest of the major operators by subscribers, a choice that drew scrutiny at the time, was appointed to build the second network. The country thus arrived, after several years and several changes of direction, at the very two-network structure policymakers had once rejected.

Sidhu has seen enough cycles not to be surprised by the churn. “What you expect to happen over a longer period generally happens,” he said. “What you can never predict is how it appears. You would not have predicted a single wholesale network, and then converted to a dual wholesale network.”

For a tower owner, more networks should mean more demand, and EdgePoint is positioned to benefit. It was named an in-building coverage partner by U Mobile and has built close to 50 in-building systems—the antenna networks that carry mobile signals inside malls, offices and transport hubs—mostly in Kuala Lumpur, with deployments in Penang and, most recently, a convention centre in Kuching. 

So far, Sidhu said, U Mobile’s demand has come largely through upgrades to existing macro sites, with new-site orders still to come. “From the end of this year onwards, we should start to see a lot more new sites that U Mobile will seek to access,” he said. He credited the operator’s approach, too: where some carriers strip in-building coverage back to the essentials, “U Mobile has gone for great coverage. I think it’s quite courageous. It’s a great business for us.”

He is measured, though, about how much duplication two networks will really create, and pushed back gently on the worry that Malaysia is now overbuilding. “Both models can work, single and dual,” he said, noting the country ran several 4G networks for years. Because there is no strict exclusivity on his sites, a tower already serving one network can serve the other. “We have a 5G site there for DNB. It probably makes a lot more sense for U Mobile to use the same site,” he said. “So I don’t think there’ll be a lot of duplicated physical infrastructure.” 

The duplication that does occur is in the radios, the active equipment operators buy from vendors such as Ericsson, Nokia and Huawei, and the costly part of any network. Two networks mean two sets of it. But the towers, land and power underneath are shared, so the total bill, in Sidhu’s view, rises well short of double.

The tenancy figures bear out the nuance. EdgePoint’s Malaysian ratio sits around 2.2, the highest in its portfolio, and has held roughly flat through the transition, not because growth stalled, but because the company has been adding sites and tenants at the same time. “The numerator and the denominator are moving a little bit in parallel,” as Sidhu put it.

The regional contrast is where he sees the real runway. Indonesia runs at roughly 1.6 to 1.8 tenants per tower; the Philippines sits just above one. Asked whether that gap is his biggest upside or a sign the sharing model has not landed there, he did not dodge: “Probably both.” The Philippines is early, he said; colocation is growing fast, but from almost nothing, and real coverage gaps mean new sites are needed regardless. 

EdgePoint bought 2,800 towers there, has since built another 350, and added around 500 colocations. Malaysia, by contrast, “started in fresh air” decades ago, with tower-sharing roots going back to around 2000. “The Philippines is in its early days,” he said. “The upside is great, but colocations require time for people to get used to how to work with them.”

Beyond adding tenants, the clearer near-term growth is in doing more on each site. EdgePoint runs combined tower-and-power sites in the Philippines end to end, has deployed solar hybrid sites in Malaysia, often where the grid physically cannot feed a heavily tenanted urban tower, and is weighing the same in Indonesia. But Sidhu resists the idea that this turns EdgePoint into an energy company. 

“There’s no end to towers and sites. What a tower or site is may change over the years,” he said. “We used to build the giant 100-metre towers. Now, typically in Malaysia, we’re building 18- to 25-metre street poles, then in-building, and even small cells where we take a lamp-post light box and convert it to an active point.” The second engine of growth, as he sees it, “is maybe not necessarily power, but more services per site.”

The sovereign question

That patient, physical, locally-regulated business is the lens Sidhu brings to a much bigger question, one he had come to his recent World Digital Economy and Technology Summit session to discuss: sovereign AI.

The term has become shorthand for a nation’s push to own and control the infrastructure behind its AI ambitions—the data, the compute, and the networks that carry it. It is an idea gaining currency across the region, and one Sidhu approaches with more caution than most. His view cuts against a debate often dominated by calls for domestic self-sufficiency. 

For a mid-sized economy, he argued, full sovereignty is simply unaffordable. He pointed to the roughly US$100 billion that leading AI labs are raising, against a Malaysian GDP of around US$400 billion. “For a country, even at a national level, that’s a huge amount of money,” he said. “So you have to pick your spots.” Malaysia’s data centre boom, in his reading, is one such spot, a way to stay relevant to the global supply chain rather than slip into being a customer country others can switch off. 

“As long as we can maintain that important role, Malaysia will always be relevant,” he said. He also pushed the definition beyond control: sovereignty can mean access, he noted, citing Canada’s plans to give citizens public compute as a kind of “GPU as a service.” That framing tracks closely with where much of the serious analysis has landed. 

Boston Consulting Group, whose thinking Sidhu drew on from his summit panel, has argued that for most countries outright AI sovereignty is an illusion, and that resilience, secure, diversified access to AI rather than home-grown self-sufficiency, is the more practical goal.

Where do tower companies fit? Not yet at the centre, Sidhu suggested, and here he was notably more sober than the industry’s usual pitch. Connectivity matters, he argued, because AI will increasingly run on the connected devices people carry, which makes coverage foundational. 

But he was direct about the limits of the fashionable idea that AI computing will soon move out of big data centres and onto the base of the tower, so-called edge computing. What the industry now calls an “edge” data centre already runs to one or ten megawatts of power: modest for a data centre, enormous for a tower site. “I don’t think we have a single site that has one megawatt,” he said. 

For AI inference—the running of trained models to answer real-world queries—to move onto towers at scale, either site power would have to grow dramatically, or the hardware would have to shrink to fit. He put it at perhaps a three-year question, and not one driven by demand today: “I don’t see people beating down my door saying, please put the stuff [AI compute] in my site. I’m happy to do it.” That candour is more useful than the alternative, given how eagerly some tower operators elsewhere have bolted themselves onto the AI story.

The sovereign question carries a sharper edge for EdgePoint than for most, because its own ownership is about to change hands. Its majority sponsor, the US-based digital infrastructure investor DigitalBridge, is being acquired by Japan’s SoftBank in a deal valuing DigitalBridge at around US$4 billion. Shareholders approved the transaction in April, and it is expected to close in the second half of this year, pending final regulatory clearances. SoftBank has framed the deal around AI infrastructure and is a central backer of Stargate, the vast US data centre venture involving OpenAI and Oracle, meaning an operator of Malaysian, Indonesian and Philippine towers will soon sit within the ownership orbit of one of the world’s most aggressive AI investors.

Sidhu was careful not to over-read it. With the deal not yet closed, he said it was too early to know whether a new direction would come from the top. “We’d like to see what happens in the market,” he said. “Market forces will be the number one determinant of how we operate, and where the investment priorities change as well.”

The deal does, though, sharpen a question beneath the whole sovereign AI debate: can infrastructure a nation calls strategic really be sovereign if it is owned from abroad? Sidhu answers that, for physical infrastructure, ownership matters far less than people assume. The pools of capital big enough to fund towers and data centres at scale, he argued, sit with a handful of global asset managers and sovereign wealth funds, the Blackstones and KKRs, alongside state funds such as Saudi Arabia’s PIF and Norway’s sovereign fund. 

“It would be silly to ignore that the capital exists, because you need that level of capital,” he said. “It will be very difficult to replace global capital with local or ASEAN-related capital. I just don’t think there’s enough.”

What genuinely anchors infrastructure to national interest, in his view, is not the shareholder’s nationality but the fact that the asset cannot leave. “Digital infrastructure has one huge advantage that software or virtual investments don’t have,” he said. “You can’t take the tower back with you to America.” The tower stays in the ground, bound by local licences, regulation and the government’s permission to operate. In Malaysia, EdgePoint holds Network Facilities Provider and Network Services Provider licences, which carry local ownership conditions, and like all operators it pays into the Universal Service Provision fund that subsidises coverage in unprofitable rural areas. 

Those obligations, he contends, are the real ties that bind. “Everybody is incentivised to make it work,” he said. “They need someone to come in, but guess what, the tower stays in the ground.” The argument has a blind spot, though. A tower may be immovable, but the decisions about where to invest, how fast to build, and which markets to prioritise are made wherever the capital sits, and those can shift with an owner’s strategy, as EdgePoint’s own change of hands may yet show. Physical permanence is not the same as local control.

It is, unavoidably, a self-interested argument from the founder of a foreign-backed towerco. But it is a coherent one, and it lands close to

where the more sober voices in the sovereign AI debate have arrived: that resilience and access, not outright ownership, are the achievable goals for most nations. 

For a business measured in decades, the question of who holds the equity today may matter far less than whether they intend to stay.

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.