Is Telkom’s Infranexia and NeutraDC Spin-Off Safe From the Commodity Trap?

Telkom is making the textbook move to escape commodity economics. Whether it works comes down to a handful of decisions you can watch — and to a set of precedents more sobering than the strategy’s champions admit.

There is an easy version of the Telkom story, and it is wrong in an instructive way.

The easy version goes like this. Telkom made its money for two decades as a connectivity utility, watched the value in telecom migrate over the top to messaging apps, streaming platforms and cloud providers, and never learned the lesson. Now it is spinning its fiber into a wholesale company called Infranexia and building hyperscale data centers under NeutraDC. Same company, same instincts, same commodity destiny — a bigger pipe and, now, a bigger farm. The DNA hasn’t changed, so the outcome won’t either.

The problem with that story is not that it’s too harsh. It’s that it’s unfalsifiable, and unfalsifiable criticism is useless to anyone actually trying to decide whether these businesses will create or destroy value. If monetizing fiber proves you’re “still a pipe,” building data centers proves you’re “now a dumb farm,” and unbundling proves you’re “rearranging the deck chairs,” then no possible action by Telkom could ever count as evidence against the thesis. A claim that survives all evidence isn’t analysis. It’s a mood.

So let me write the version that can be wrong, and is therefore worth reading. The question is not whether Telkom has the soul of a utility. It probably does. The question is under what specific, observable conditions each of these businesses becomes a commoditized landlord, and under what conditions it becomes something with a moat. Those conditions are knowable. Most of them are already being decided.

Separation is the cure, not the disease

Start by giving the move its due, because the reflexive dismissal gets the history backwards.

The value destruction that hollowed out telecom operators was never caused by “being infrastructure.” It was caused by integration — by operators who kept the retail customer, bore the full weight of the capital expenditure, and then let others capture the margin, the brand and the pricing power on top of their networks. The mistake was category confusion: running utility economics while wearing a growth-tech costume and being valued as neither. Towers are infrastructure too, and tower companies trade at premium multiples precisely because they are boring, contracted and shared. Being a pipe was never the sin. Pretending a pipe was a growth story was.

Seen that way, structural separation into a pure fiber company is the recognized response to the mistake, not a fresh instance of it. Telstra carved out InfraCo. Telecom Italia sold its fixed network. Telecom New Zealand demerged Chorus. The Czech group PPF built CETIN. The logic is consistent wherever it’s tried: put the long-duration, capital-heavy, stable-cashflow assets into their own vehicle so they can be financed and valued on infrastructure terms, and free the rest of the group to be valued on its own.

The numbers behind Infranexia fit that intent rather than empire-building. The first phase, signed on 18 December 2025, moved about half of Telkom’s wholesale fiber assets — a book value near Rp 35.8 trillion (roughly US$2.2 billion) — into PT Telkom Infrastruktur Indonesia, with the strategy-and-portfolio director, Seno Soemadji, describing a total heading toward Rp 90 trillion once the second phase completes in the second half of 2026, and an eventual asset value the company frames near Rp 150 trillion (about US$9 billion).[1][2][3] The footprint is genuinely large — roughly 179,000 kilometers of fiber across some 501 cities.[1] Telkom keeps 99.9 percent.[2] It sits inside a broader restructuring the company calls TLKM 30, whose third pillar is explicitly about unlocking the value of its infrastructure portfolio, and it carries the imprimatur of Danantara, the state investment holding now sitting above the SOEs.

None of that is the behavior of a company that hasn’t noticed the pipe problem. On paper, it is a company doing something about it. Which is why the interesting failure modes are subtle — and why the precedents deserve a harder look than the strategy’s champions give them.

But did the precedents actually work?

Invoking that lineage only reassures if those carve-outs created durable shareholder value. Sort them by asset type and outcome, and a pattern emerges that should temper the optimism, not feed it.

The towers crystallized real premiums that then compressed. Telstra sold 49 percent of its tower company, later Amplitel, in 2021 at a valuation of about A$5.9 billion — 28 times EV/EBITDA — kept control, and returned roughly half the proceeds to shareholders.[5] The model working as advertised. Mitratel, Telkom’s own towers, priced its 2021 IPO at a mid-teens tower multiple against the roughly five times the market was assigning the blended parent, so the crystallization worked at the moment of sale; but the stock has since drifted below its IPO price and the multiple has compressed toward the high single digits.[7] Cellnex, Europe’s purest tower compounder, ran to an all-time high near €62 in August 2021 and now trades at roughly half that, dragged down by rising rates and forced into asset sales.[6] Infrastructure multiples are real, but they are long-duration and rate-sensitive: a re-rating is a moment, not a ratchet, and it can reverse. (The point is visible even today in live comparables — American Tower trades near 18.5x EV/EBITDA while Mitratel, also a tower company, trades near 9.5x.[7][13] The multiple tracks growth, scale and rates far more than “which layer of the stack you are,” which is exactly why a tidy fiber-versus-towers-versus-data-center valuation ladder would mislead more than it informs.)

Fiber — Infranexia’s actual asset class — is more sobering still, and it splits three ways. Chorus, demerged from Telecom New Zealand in 2011, is the success case: a wholesale-only fiber network barred by law from selling to consumers, with EBITDA margins above 45 percent, a dividend yield near 6 percent, a five-year total shareholder return around 55 percent, and a market capitalization that has grown to approach that of the retail company it was split from.[8] But read the fine print on why it works — Chorus is a regulated near-monopoly, its returns set and protected by five-yearly regulatory resets. Take away the regulatory guarantee and you take away the thesis. Infranexia has no equivalent protected-return regime; it is a commercial wholesaler exposed to competition, so the precedent it most resembles is not the one that worked.

The two fiber precedents it resembles more are cautionary. Openreach is the textbook separation — functionally split from BT in 2006, legally separated within the group in 2017 — and yet the value stayed trapped: analysts (Barclays among them) have pegged Openreach’s standalone worth near £30 billion against a BT Group market capitalization of only around £21 billion, BT’s own chief executive has said publicly that the value isn’t reflected in the share price, and full monetization keeps being pushed out to after the fibre build completes around 2030.[9] Separation, in other words, did not force a re-rating; the conglomerate discount persisted for the better part of a decade, and rivals argue Openreach’s independence is cosmetic precisely because BT’s own retail arm remains its captive anchor tenant — the same captive-anchor structure Infranexia has with Telkomsel. And TIM is the distress case: Telecom Italia sold its fixed network, NetCo, into FiberCop and on to a KKR-led consortium in 2024 in a deal valued up to €22 billion — but that was debt-driven survival, not value maximization.[10] TIM offloaded the best asset it owned to cut about €14 billion of net debt, its equity stayed depressed, its largest shareholder opposed the sale, and the infrastructure value accrued to the buyers rather than to TIM’s shareholders. Worth noting for later: the network TIM kept access to is governed by a fifteen-year master service agreement priced at market rates, because the buyers insisted.[10]

The table below sorts the record. Place Infranexia against it and its position is uncomfortable: a fiber wholesaler, without the regulated-monopoly returns that made Chorus work, carrying a captive anchor tenant like the one that keeps Openreach’s value trapped inside BT, owned 99.9 percent by a parent that — if it ever needs to sell under pressure, as TIM did — would watch the value leave for the buyer. The strong precedents are towers, which have the site-by-site scarcity fiber lacks. The fiber precedents that match Infranexia’s profile are the ones where separation either didn’t unlock the value or handed it to someone else. That does not doom the strategy. It does put the burden of proof on Telkom, not on the comforting lineage.

The real indictment is about layers, not culture

Here is the version of the bearish case that survives scrutiny, and it has nothing to do with DNA. It’s about where value sits in the stack.

Fiber is the low-margin floor of the AI build-out — low relative to the compute and power above it. The economics of this cycle concentrate in the chips and in the power and tenant relationships around them, and the strand that connects it all captures the least. That said, the “floor” is not uniform: fiber acquires real scarcity value at specific choke points — subsea cable landing stations, major cloud-interconnection hubs, and the routes into dense AI campuses — where a particular path is genuinely hard to replicate. This is the difference between selling commodity bandwidth, which has no pricing power, and selling position, which does. It also happens to be where Batam’s subsea link and Infranexia’s better metro routes could matter. The structural point stands, but the escape route runs through scarce positions, not raw capacity.

Within the floor, fiber faces two traps that towers escape. A tower has genuine local-monopoly economics: permits, anchor tenancy and physical non-replicability, site by site. Fiber routes are bimodal. On dense, valuable metro and backbone routes they are contestable — a competitor can overbuild them, and competition on identical capacity collapses to price. On thin regional routes they are the opposite problem: uneconomic to replicate, but also uneconomic to fill. So a fiber wholesaler can end up with commodity margins where it competes and stranded capital where it doesn’t — unless the footprint has real non-replicability and the operator prices on scarcity rather than throughput.

This is the honest way to read the utilization figure Telkom keeps citing. Wholesale fiber utilization is reported at only about 40 percent, mostly consumed by Telkomsel, and management frames that spare capacity as headroom to grow into.[2][3] The more revealing way to read it is as a dilemma the company has not resolved in public. Telkom has, after all, already done the obvious thing to drive consumption through its own channel: in 2023 it folded IndiHome, its dominant consumer fixed-broadband business, into Telkomsel under its Fixed Mobile Convergence plan, routing the retail broadband base through the mobile champion to lift uptake.[4] If, after years of ownership and that convergence, the passive network still runs at only 40 percent, one of two things must be true. Either the unused sixty percent is not economically monetizable — thin regional and dark fiber laid to satisfy coverage obligations rather than to meet paying demand — in which case the “significant headroom” that underwrites the spin-off’s value is largely hollow. Or it is monetizable and management simply failed to sell it across years of ownership — in which case it is fair to ask why the same team captures it better merely because the asset now sits in a differently named box. The company cannot rest easily on either horn, and it will not disclose the denominator — how much of that 40 percent is Telkomsel backhaul, how much is fixed broadband, how much is idle dark strand — that would tell you which one bites.

Note, too, where the demand to fill it would have to come from. Lowering mobile tariffs would not do it: mobile traffic touches this network only as backhaul, a thin slice that rides on already-lit routes, not the idle regional fiber that drags the number down. And in a mobile-first market whose retail consumption has already been pulled inside Telkomsel, Infranexia’s external customers are not consumers at all; they are rival operators’ backhaul, tower companies, data centers, ISPs and enterprise — the very parties most reluctant to lease from a network their largest competitor owns. The utilization question and the neutrality question are therefore the same question wearing two hats: the 40 percent becomes 60 or 70 only if outsiders who distrust the landlord decide to move in.

But the single most likely way Telkom manufactures a commodity pipe out of Infranexia is neither of those. It is transfer pricing. Telkom owns both sides of the anchor relationship at 99.9 percent. Price the Telkomsel lease to flatter Telkomsel’s margins — cost-recovery rates, set internally, undisclosed — and Infranexia becomes a captive utility with thin earnings, no matter how good the fiber is. Price it high, and minority investors or an incoming strategic partner read it as circular related-party revenue and discount it anyway. The entire re-rating hinges on this variable being handled at arm’s length and being seen to be. It is telling that Telkom has, for now, favored a strategic partner over a public listing — a partner diligences anchor-lease terms far harder than a retail order book.[3] The TIM precedent is instructive here too: even that fully independent network sits on a market-priced, fifteen-year agreement, because the buyers demanded it.[10] Infranexia’s terms are not public. This is the thing to watch, and it is not yet visible.

The farm has the same disease — and one real antibody

Extend the logic upward and NeutraDC looks, at first, no safer. A hyperscale data center is, stripped to its economics, a landlord: it provides the shell, the power and the cooling, while the tenant brings the chips and captures the compute, the platform, the models and the applications. This is why Equinix and Digital Realty are structured and valued as real-estate investment trusts, not technology companies, and run adjusted-EBITDA margins around 50 percent — healthy, but landlord economics, not software economics.[14] And “data center” is not one business but several — retail colocation, wholesale colocation, single-tenant hyperscale build-to-suit, sovereign cloud, GPU-cloud — with very different economics; Batam sits at the hyperscale-wholesale end, the most capital-intensive and the most exposed to a handful of very large, very powerful tenants. On the stack logic, then, the data center faces the same charge as the pipe: value accretes above the box, while the operator collects rent on a depreciating, power-hungry, location-specific asset.

There is one disanalogy that keeps this from being a clean equivalence, and it is the crux. The scarce input the operator can actually own in this cycle is not compute — the GPU shortage is real, but that rent goes to Nvidia, not the building — and certainly not glass. It is power. Secured, affordable, ideally green power at scale, on permitted land, near connectivity, is genuinely constrained and slow to build. A data center can be a landlord that nonetheless earns toll-road economics — but only if it controls the scarce input beneath it. Resell someone else’s power to tenants who hold the demand, and it is squeezed from both sides: the rent leaks upstream to the generator and downstream to the workload.

This is where the Batam facts complicate the bearish case in Telkom’s favor. The hyperscale project at Kabil Industrial Estate — BTM-1, topped off on 30 October 2025 and targeted to go live in the first half of 2026 — is a joint venture between Telkom’s NeutraDC, Singtel’s data-center arm Nxera, and, crucially, Medco Power.[11] Medco is not purely an arm’s-length utility selling into the building; it is inside the equity, so a share of the power economics the “reselling someone else’s input” critique assumes leaks away is captured within the venture — though the campus also draws grid supply under a separate electricity-purchase agreement with PLN Batam, so the “owns its power” claim is partial, not total.[11] The design points toward the premium end: an initial IT load around 18 megawatts scalable toward 54, Uptime Tier III certification, liquid-cooling-ready high-density AI floors, and a stated push toward green energy.[11] It is marketed as carrier-neutral, with master agreements across seven connectivity partners, and wired into a planned Batam–Singapore subsea cable to cut latency to the Singapore cloud and exchange hubs.[11] (Total investment is reported near Rp 1.4 trillion over five years — a modest figure that implies this first campus is a foothold, not the whole bet.[11])

Then the geography, which is where the argument needs its sharpest correction. Batam sits minutes from Singapore’s digital core but inside Indonesian jurisdiction, so for workloads that need low latency to Singapore and Indonesian data residency — a combination regulation increasingly forces — it is genuinely advantaged. The trap is to mistake that for NeutraDC’s moat. It is Batam’s moat, and Batam has become a contested cluster: GDS, Oracle, BW Digital, Gaw Capital and others are building there too.[12] Sovereignty-plus-latency is now table stakes on the island, not a NeutraDC differentiator. What differentiates NeutraDC within Batam is a bundle — Nxera’s operating experience, Medco’s power inside the equity, Telkom’s connectivity, Danantara’s capital. That bundle is real, but a bundle is contestable in a way a true monopoly is not, and should be judged as such.

So the farm has an antibody the pipe does not. It does not follow that the antibody works.

The bull case, made as well as I can make it

An article that only prosecutes is easy to dismiss, so here is the strongest defense of these two bets, stated without hedging.

Telkom is not sliding backward into commodity economics; it is disaggregating so each layer can be valued on its own terms instead of buried inside a blended utility discount. Mitratel already showed domestically that the market will price Telkom infrastructure on infrastructure multiples once it is separated. The Batam data center is not a speculative land grab but a partnership with the region’s most experienced operator, the power producer inside the tent, sovereign capital behind it, and a location marrying Singapore-adjacent latency with Indonesian jurisdiction. Power scarcity, correctly owned, is a toll road, and NeutraDC has arranged to own a piece of it rather than rent all of it. The carrier-neutral posture and subsea interconnection suggest a company trying to sell position — the thing that carries margin — rather than raw capacity. On this reading, Telkom has looked squarely at the pipe problem and is executing the recognized escape from it, with better cards than most incumbents held.

That case is coherent, and parts of it are already supported by facts on the ground. Here is why it can still fail.

What will actually decide it

The outcome is contingent, and the contingencies are observable. Each item below is a test you can apply as evidence arrives — which is the point, because it replaces “we can’t know until it’s too late” with things you can in fact know early.

The anchor-lease terms. Is the Telkomsel lease priced at arm’s length and benchmarked against external wholesale rates, or set internally to protect Telkomsel’s margins? The tell is whether the terms are disclosed at all, and whether Infranexia’s standalone EBITDA margin reads like an infrastructure company’s or a captive cost center’s. This single variable dominates the fiber outcome — and it is exactly what kept Openreach’s value trapped inside BT.

Position versus bandwidth. Does Infranexia sell interconnection, dark fiber and cloud on-ramps to parties who value the specific route — or gigabits per second, won on being cheapest? The moment the pitch becomes “most capacity, best price,” the margin is gone.

Credible neutrality. Do competitors actually lease from Infranexia, or only Telkomsel? Rivals will not route their traffic through a network owned by their largest competitor unless the neutrality is real — governance firewalls, non-discriminatory pricing, plausibly a strategic minority partner with board rights. Without it, utilization never rises and the pipe stays captive.

Route discipline. Does Telkom monetize the genuinely non-replicable parts of the footprint and starve capital on contested metro routes, or overbuild where competition is already a knife fight to show growth? Route mileage is not value; building into commodity corridors manufactures commodity economics with your own money.

Power ownership, in practice. The Medco joint venture is promising, but how much power rent does the venture retain versus pass through, and is future capacity secured under favorable long-term green terms or bought at market from the grid? The moat is only as real as the contracts beneath it, and those are not public.

Tenant concentration. Does the Batam campus fill with a single hyperscale anchor that dictates spec and price and can walk at renewal, or with a diversified base of enterprise, sovereign and regional tenants who value the jurisdiction and the latency? The gap between build-to-suit for one tenant and a carrier-neutral sovereign hub is the gap between commodity and moat.

Narrative discipline. Does Telkom sell Infranexia as a stable, contracted, yield-plus-moderate-growth infrastructure asset — what infrastructure investors pay premium multiples for precisely because it is boring — or promise it will grow into a second Telkomsel? The latter re-imports the original mistake: a utility sold as a growth story, destined to disappoint. The current messaging leans toward the growth promise, and that is the instinct worth resisting.

Integration. Does the group keep fiber commercially wired to the data-center and interconnection ecosystem, so the glass becomes the fabric the cloud, edge and DC hang off — or sever the two for a cleaner carve-out and throw away the one thing that makes fiber more than a pipe? A tidy separation that cuts the commercial integration optimizes for accounting legibility at the cost of strategic value.

The same tests, as a scorecard a reader can mark to date:

The empty right-hand column is not a gap in the analysis; it is the analysis. Almost nothing that decides the outcome is yet public.

The verdict, contingent by design

The bleak-future thesis, in its “the DNA never changes” form, predicts that the textbook fix will fail without explaining why it fails in this case — and quietly assumes the very thing it should prove. The more defensible position is less satisfying and more useful.

Telkom is making the right structural move into two businesses whose economics are genuinely unforgiving. Fiber is the low-margin floor of the AI stack, and a data center is a landlord; those are the starting conditions, and they are not favorable. The precedents cut the same way. The carve-outs that created durable value were towers, which have the scarcity fiber lacks. The fiber case that worked — Chorus — did so as a regulated monopoly, a protection Infranexia does not have. The fiber cases that match Infranexia’s profile are Openreach, where separation left the value trapped behind a captive anchor for a decade, and TIM, where a forced sale sent the value to the buyer. Against that, NeutraDC has assembled real ingredients of a moat — power partly inside the equity, a Singapore-adjacent sovereign location, a carrier-neutral posture — with the qualifier that the location advantage is shared with a growing Batam cluster, so its edge rests on a contestable bundle rather than a monopoly. And the residual risk, even with flawless execution, is the one no strategy repeals: value in this cycle concentrates above both layers, and a perfectly run floor of the stack is still the floor.

So the danger is not that Telkom is trapped in the past. It is subtler and more specific: that it could execute a flawless carve-out and a well-built data center and still capture only a thin slice — unless the anchor-lease pricing, the neutrality, the power economics and the tenant mix all break the right way. Those are not questions of corporate soul. They are decisions, most of them being made now, all of them visible if you know where to look. Judge the company on those, as they land, and you will know years before the multiple tells you.