DCI Indonesia

1. 01Build the hall, then sell the megawatt

   DCI runs about 128 MW of IT load across nine buildings in Cibitung, Karawang, Jakarta and Surabaya, and it was the first operator in Southeast Asia to hold the Uptime Institute’s Tier IV Gold certification for operational sustainability. Colocation, renting secure powered space, was Rp2.36tn of FY2025 revenue, roughly 92.8% of the total.

   EconomicsThe unit of this business is a megawatt of IT load, so start there: 128 MW produced Rp2.54tn of revenue, which is about Rp19.8bn of revenue per MW per year, and at a 49.7% operating margin roughly Rp9.9bn of operating profit per MW. Almost everything else about DCII follows from two questions, how many megawatts it has and how full they are.

2. 02Fill it with hyperscalers, and pass the power bill through

   Hyperscale clients account for roughly 65% to 75% of recurring revenue in 2025 and 2026, and their contracts pass the cost of power through to the tenant rather than leaving it with DCI.

   EconomicsElectricity is the largest variable cost a data centre has, and passing it through converts it into a neutral item. That single contractual feature is why the EBITDA margin stayed inside a 60.8% to 67.1% band across five years that included a global energy-price shock. The price of that stability is concentration: a small number of very large tenants decide the revenue line.

3. 03Accept the margin dip while the new hall fills

   Gross margin ran 54.6%, 57.4%, 59.1%, 58.3%, 54.1% and operating margin 48.0%, 50.4%, 52.4%, 53.7%, 49.7%, so both peaked before FY2025 and then slipped. JK6, at 36 MW, opened in June 2025.

   EconomicsA data hall costs from the day it is energised but bills only as tenants move in, so opening a 36 MW building mid-year mechanically dilutes margin for a while. This is the same pattern MIKA shows when it opens a hospital, and it means a margin dip during a build is expected rather than alarming. What would be alarming is a dip that does not reverse as the hall fills.

4. 04Spend more than you earn on the next building

   Capex went Rp0.32tn (FY2023) → Rp0.88tn → Rp1.72tn, which is 68% of revenue and 6.1× depreciation, against record operating cash flow of Rp1.54tn. Free cash flow was therefore −Rp0.18tn, debt rose Rp1.06tn → Rp1.88tn, and a USD 1bn credit facility from BCA stands behind the next phase.

   EconomicsNegative free cash flow here is a decision, not a symptom. Note the five-year totals before judging it: Rp4.22tn of operating cash flow against Rp3.73tn of capex, so the build has been largely self-funded and debt actually FELL from Rp1.39tn to Rp1.06tn before the FY2025 acceleration. The question that follows is the only one that matters, and it is the next step.

5. 05Check whether the new capital earns more than it costs

   Return on invested capital went 13.0%, 16.5%, 20.5%, 22.5%, 19.1%. The engine puts the weighted cost of that capital at 17.61%, built on a levered beta of 1.55 taken from Damodaran’s Software (Internet) row, and the engine’s own note on that row says a data-centre landlord belongs nearer Information Services, at 0.74. Substituting 0.74 gives a cost of capital of 12.18%.

   EconomicsGrowth is worth having only when the return on the incremental capital beats the cost of that capital. At 12.18%, a 19.1% ROIC is roughly 6.9 percentage points of genuine value creation, and it has been positive and rising since FY2021. That is exactly the test BREN fails on the same page structure (6.4% against 7.98%) and DCII passes comfortably. Learn to run these two numbers side by side before you form any view on price, because they separate a business that deserves to grow from one that merely wants to.

Cost structurePower, then depreciation, then very little else. Cost of revenue was Rp1.17tn against Rp2.54tn of revenue, a 54.1% gross margin, and the dominant input inside it is electricity, which is contractually passed through for hyperscale tenants. Depreciation is Rp0.28tn, about 11% of revenue, and it climbs with every building energised. Interest is remarkably cheap for a capital-hungry business: Rp0.09tn on Rp1.88tn of debt, a 4.76% cost of debt, which is why interest cover improved to 14.1× even as borrowings rose 77% in one year. Working capital runs tight by design, with a current ratio of 0.95 and receivables at 88 days, having spiked to 127 days in FY2024 around new-capacity handovers.

Cash cycleContracted colocation revenue arrives steadily, converts to cash well (operating cash flow exceeded net income in all five years), and then goes straight back into concrete and transformers. In FY2025 the loop ran Rp1.54tn in and Rp1.72tn out, so the shortfall was funded on the balance sheet: debt Rp1.06tn → Rp1.88tn plus a USD 1bn facility for what comes next. The important feature of this cycle is that it is capacity-constrained rather than demand-constrained. Fitch, affirming AA-(idn), expects installed capacity to move from about 128 MW toward more than 850 MW within two to three years, which means the binding limit on revenue is how fast buildings can be energised and financed, not whether tenants exist.

Unit economicsWork in megawatts and the whole company becomes legible. Today: 128 MW, Rp19.8bn of revenue per MW, Rp9.9bn of operating profit per MW. Apply the same revenue per MW to the 850 MW capacity path the rating agency describes and you get roughly Rp16.9tn of revenue, about 6.6× FY2025, which is arithmetic on a disclosed pipeline rather than a forecast, and it deliberately assumes no improvement in price or utilisation. Now put that against the price, because this is where the discipline lies: at 297× EV/EBITDA, EBITDA has to rise about 11.9× merely to reach 25×, a rich multiple for infrastructure anywhere, and at the delivered 40% pace that takes roughly seven years. The business can be excellent and the price can still require perfection.

High and cash-backed, with one line that needs checking outside this page. The good part is unambiguous: operating cash flow exceeded net income in every year of the window (1.66×, 1.83×, 1.44×, 1.05×, 1.54×), there are no one-off items to strip out, and the growth is independently corroborated rather than self-reported, since the FY2025 revenue of Rp2.54tn and its 40.1% increase match the company’s own disclosure and the capacity additions behind it are separately dated (JK6, 36 MW, June 2025). Two things deserve a reader’s attention. Receivables are lumpy around handovers: days outstanding ran 94, 69, 74, then 127 in FY2024 before recovering to 88, which is worth watching but has not cost cash. More important is tax. The effective rate has been 20.9%, 17.0%, 12.1%, 10.9% and 14.7%, persistently below Indonesia’s 22% statutory rate, and nothing in the ratio set explains why. If that rate normalised to 22%, net income would be roughly 9% lower, so do not capitalise the current rate into a valuation before reading the tax note in the accounts.

Checked: no material one-offs found in the reviewed window.

Cash conversionStrong, and the negative free cash flow is not a contradiction. Operating cash flow grew 3.6× over the window, Rp0.43tn → Rp1.54tn, comfortably ahead of net income each year. Free cash flow was −Rp0.06tn, +Rp0.36tn, +Rp0.43tn, −Rp0.04tn, −Rp0.18tn, so it swings with the build schedule rather than with profitability, and cumulatively across five years operating cash flow of Rp4.22tn still exceeded capex of Rp3.73tn. The FY2025 deficit is the cost of a 36 MW building, not a deterioration in collection.

The purest builder in the roster, and it is provable rather than asserted. DCII has never paid a dividend, and the balance sheet confirms it: FY2021 owners’ equity of Rp1.21tn plus the Rp2.68tn of profit earned across FY2022 to FY2025 comes to Rp3.89tn, against Rp3.90tn actually reported, so essentially every rupiah earned stayed inside the company. That capital went into capacity at 6.1× depreciation, and unlike a builder that merely grows, this one clears the bar that makes growth worth having: ROIC of 19.1% against a cost of capital nearer 12.18% once the beta is drawn from the right sector row. Compare BRIS among the banks, which also retains almost everything and also earns well above its cost of capital, and contrast BREN, which retains and builds but earns below it.

DeploymentFive years of cash, in order. Into capacity: capex of Rp0.50tn, Rp0.31tn, Rp0.32tn, Rp0.88tn and Rp1.72tn, Rp3.73tn cumulatively, with 70% of it in the last two years. Funded by: Rp4.22tn of cumulative operating cash flow, so the programme was self-funded until FY2025, when debt was raised from Rp1.06tn to Rp1.88tn. Out to shareholders: nothing, in any year. Standing ready: a USD 1bn BCA facility for the next phase. Note what the debt path actually shows, because it is easy to misread a jump: borrowings FELL from Rp1.39tn to Rp1.06tn across FY2021 to FY2024 while capacity grew, and only rose when the build accelerated.

Returns trendROIC 13.0%, 16.5%, 20.5%, 22.5%, 19.1%, and the FY2025 dip is the newly opened 36 MW building sitting in the capital base before it is full. Against the engine’s 17.61% cost of capital the spread looks thin; against the 12.18% that the correct sector beta produces it is roughly 6.9 points, and rising over the window. So the business verdict is clear and positive. The price verdict is the opposite, and the honest reading requires naming three limits of the models on this page. First, the reverse DCF returns nothing at all for DCII: base-year free cash flow is negative, so there is no implied-growth rate to compare against delivered growth, and that tool simply does not apply to a company in a heavy build. Second, the forward DCF prints Rp327 a share on its defaults, which should not be read as a fair value: it discounts at 17.61% using a Software (Internet) beta the engine itself flags as the wrong row for a data-centre landlord, and it holds capex at 46.9% of revenue in perpetuity while growth fades to 2.5%, an incoherent pairing that means paying forever for growth you never receive. Correct the beta and it is about Rp1,000; put capex at a steady state equal to depreciation and it is about Rp4,650; do both and it is roughly Rp7,900. Even that last figure is about 4% of the Rp191,651 price. Third, the relative check offers no rescue and must not be quoted as if it did: the printed industry median of 156.66× EV/EBITDA is the midpoint of exactly two members, DCII at 296.5× and EMTK at 16.8×, with GOTO and BUKA suppressed as loss-makers, so it is a two-name artefact rather than a benchmark. What survives all three limits is the arithmetic: at 297× EV/EBITDA the market is asking for about seven consecutive years of 40% EBITDA growth just to arrive at a rich infrastructure multiple, and the free float is 18.55% with four holders owning roughly 78%, so a thin market is setting that price.