Indonesia's Compute-Power Bargain: From Template to Contract Status

Rupiah Stability Watch · 2026-10-22

The premise

In October 2018, this organization published "The Compute-Power Bargain: Contract Architecture for Private Load Financing Public Grid Resilience in Indonesia." That piece established a template: five measurable metrics a contract must deliver (diesel litres avoided, outage minutes avoided, grid capex privately funded, storage duration matching Indonesian failure modes, FX exposure disclosed & hedged), three contract archetypes (nuclear uprate subscription, enhanced geothermal offtake pull-forward, storage-duration anchor for BESS+hybrid solar), and the verification infrastructure needed.

The question six months later is simple: has Indonesia moved from template to contract?

The answer matters because the external-balance channel is binary. Either private compute load finances measurable public resilience — diesel displaced, outage minutes cut, grid capex privately funded, storage duration matched to Indonesian failure modes, FX hedged — or it becomes an enclave drain: imported equipment, fuel-linked power, dollar-denominated contracts, and an AI-bloc risk premium that the rupiah absorbs.

What the evidence supports

Two signed PPAs, one record-setting. BDx Data Centers secured a 1.2 GW power portfolio with PLN across Indonesia in June 2026 — the single largest data-center power commitment in Southeast Asia to date. PT PLN Batam and DayOne signed a 511 MVA (~450 MW) PPA in April 2026, Indonesia's largest data-center PPA to date, for a hyperscale campus in Batam. Both are real, signed, and large.

A 1.3 GW investment pipeline has been declared. At a KADIN business breakfast on 10 July 2026, Coordinating Minister for Economic Affairs Airlangga Hartarto stated that global investors had expressed interest in a combined 1.3 GW of data-centre capacity, against roughly 580 MW currently operating. The pipeline is valued at Rp270–360 trillion (USD 17–23 billion). Nvidia and Telkom are named as anchor investors; Arm Ltd is engaged for digital talent development.

PLN is building a BESS pipeline, but not anchored to compute. PLN has 21 solar+BESS projects (513 MWp total) across seven provinces entering execution in 2026–2028. The stated purpose is fuel displacement in the power sector. No public document ties any of these projects to a data-center offtake anchor or to a storage-duration commitment matched to Indonesian grid failure modes.

Pertamina Geothermal Energy is studying a Kamojang pilot. PGE has a data-centre pilot project in the Kamojang geothermal working area (West Java) in its final study stages as of August 2026. It is explicitly a study; no offtake agreement, no pull-forward commitment, no commercial operation date.

ThorCon is in pre-licensing. The US molten-salt reactor developer has submitted documents to Indonesia's nuclear regulator (BAPETEN) and is investigating a site at Kelasa Island (Bangka). As of February 2026, it remains in licensing consultation. No construction licence, no power purchase agreement, no uprate subscription model.

A hybrid regulatory framework exists, but without grid-resilience clauses. Indonesia permits private power procurement for data centres (PLN wheeling, direct PPA with IPPs, on-site generation). PP 33/2026 (effective Jan 2027) governs personal data cross-border flows. The PwC 2026 Power Survey highlights bankable PPAs with "appropriate risk allocation" as an industry priority. No regulation or ministry decree requires or incentivises grid-resilience clauses — diesel avoidance, outage-minute reduction, privately funded grid capex, storage duration matching — in data-centre power contracts.

No FX hedging structures on data-centre power procurement. The RMB clearing-bank launch for Indonesia–China transactions (October 2026) and Bank Indonesia's Vastra framework for foreign portfolio hedging are market-plumbing tests. Neither is linked to data-centre power contracts. The BDx and DayOne PPAs are reported in USD terms with no disclosed rupiah or RMB settlement or hedging mechanism.

What the evidence does not support

The 13 GW figure from the template piece is not a current commitment. The October 2018 publication cited 13 GW as a demand projection. Airlangga's July 2026 statement puts the pipeline at 1.3 GW — one-tenth of that figure — against 580 MW operational. The gap between projection and pipeline is the gap between a template and a contract.

No nuclear uprate subscription exists. ThorCon is in licensing; no Indonesian nuclear plant is operating, uprating, or offering a subscription product to a data-centre offtaker.

No enhanced geothermal offtake pull-forward exists. PGE's Kamojang pilot is a study. Star Energy and other geothermal IPPs have not announced compute offtake pull-forwards.

No storage-duration anchor for BESS+hybrid solar exists. PLN's 21-project BESS pipeline (513 MWp) is a utility programme for fuel displacement. No project has a data-centre anchor committing to a storage duration matched to Indonesian failure modes (typically 4–8 hours for Java–Bali, longer for outer islands).

No contract discloses the five metrics. Neither the BDx 1.2 GW portfolio nor the DayOne 511 MVA PPA discloses: diesel litres avoided per year, outage minutes avoided per year, grid capex privately funded, storage duration committed, or FX exposure and hedge ratio.

The least-harm path

The least-harm path is to treat the two signed PPAs as the floor — proof that large private load can be contracted in Indonesia — and require the next contracts to carry the resilience metrics the template named.

For the Ministry of Energy (ESDM) and BKPMM: attach a grid-resilience rider to future data-centre power approvals. The rider need not be complex: a schedule declaring (a) baseline diesel generation displaced (litres/year), (b) SAIDI/SAIFI improvement committed (minutes/year), (c) grid capex the offtaker funds (substation, line upgrade, BESS), (d) storage duration contracted (hours) and the failure mode it covers, (e) FX denomination and hedge ratio. Make it a condition of the IUPTL or the BKPMM recommendation letter.

For PLN: convert one of the 21 solar+BESS projects into a storage-duration anchor. Offer a 6-hour or 8-hour BESS product with a data-centre offtake that pays a capacity premium for the duration guarantee. Publish the diesel litres and outage minutes that project avoids.

For PGE and geothermal IPPs: move the Kamojang pilot from study to a 50 MW offtake pull-forward with a 15-year tenor. Publish the diesel avoidance and the rupiah-denominated tariff structure.

For BI and the RMB clearing bank: design a data-centre power settlement window — rupiah or RMB — for contracts that carry the resilience rider. The Vastra framework can extend to corporate hedging of rupiah-denominated PPAs.

For the verification layer: this organization will track six observable signposts for the next six months (see watchlist below). Each signpost is a binary yes/no. The absence of a yes is the signal.

Watchlist — six signposts for the next six months

1. Nuclear uprate subscription: ThorCon or another SMR developer signs a term sheet with a data-centre offtaker for an uprate subscription (MW-year) with a rupiah-denominated capacity payment.
2. Geothermal offtake pull-forward: PGE or Star Energy announces a signed offtake agreement with a data-centre operator pulling forward a geothermal expansion phase.
3. Storage-duration anchor: PLN or an IPP announces a BESS+hybrid solar project with a data-centre anchor committing to a contracted storage duration (≥6 hours) matched to a published Indonesian failure mode.
4. Grid-resilience rider in a regulation or approval: ESDM, BKPMM, or BI issues a circular/letter requiring or incentivising the five metrics in data-centre power contracts.
5. FX-hedged PPA: A data-centre PPA discloses rupiah or RMB denomination and a hedge ratio (or natural hedge via local currency revenue).
6. Verification infrastructure live: A public dashboard (PLN, ESDM, or independent) publishes the five metrics for signed data-centre PPAs, updated quarterly.

What I'm uncertain about

The BDx 1.2 GW portfolio's internal structure. The announcement describes a "strategic partnership" and "power portfolio" across multiple sites. It is unclear how many distinct PPAs underlie the 1.2 GW, their tenors, tariff structures, or whether any site includes on-site generation or storage. BDx has not published the individual contracts.

Whether the DayOne 511 MVA PPA includes any on-site resilience. DayOne's Batam campus may include backup generation or BESS. The public materials do not say. If it does, the diesel avoidance and outage-minute metrics could be calculated — but they are not in the public domain.

The actual operational capacity vs. the 580 MW baseline. Industry trackers (DCByte, Structure Research) place Indonesia's operational data-centre capacity between 300–600 MW as of mid-2025. The 580 MW figure from Airlangga's office is plausible but unverified against a single registry.

Whether the RMB clearing bank will accept corporate PPA flows. The clearing bank is designed for trade settlement. Whether a rupiah- or RMB-denominated data-centre PPA qualifies for the settlement window is an operational question BI and the clearing bank have not answered publicly.

The AI-bloc risk premium. If Indonesia's data-centre contracts remain dollar-denominated, fuel-linked, and without grid-resilience clauses, the rupiah absorbs the risk premium when global AI capital reprices. Quantifying that premium requires a counterfactual (resilience contracts) that does not yet exist. The absence is the current data point.