
Indonesia's data centre market has moved from megawatts to gigawatts. Operators are announcing AI campuses on a scale the country has not seen before, banks are lending record sums to build them, and the state electricity company says the demand it sees today is only the beginning. For enterprises planning where their cloud and AI workloads will run, the useful question is not how large the announcements are, but when the capacity actually arrives and what will decide that.
What is being built
Two projects show the scale. In Batam, Firmus Technologies and the Singapore-based developer DayOne are building a 360 MW AI campus with Nvidia, planned to house up to 170,000 GPUs delivered across 2027 and 2028, with operations scheduled to begin in the first quarter of 2027. The facility is designed as a multi-tenant site for AI-native customers rather than for a single hyperscaler, according to reporting from June.
In West Java, BDx Data Centers is developing CGK4 in Jatiluhur, a campus planned at 640 MW across six buildings, with a first 120 MW building due in phases from early 2027. BDx has secured 845 MVA of grid power for the site from PLN and more than 1.2 GVA across its Indonesian campuses, which the operator described in June as the largest power commitment to a data centre company in the country. Princeton Digital Group is building a separate 120 MW campus near Jakarta with a first phase due in the fourth quarter of 2026.
Power is now the constraint
In late September, PLN reported that it serves 159 data centre customers with a connected capacity of 2,038 MVA and annual consumption of 2,378 GWh. Ninety-nine per cent of that capacity sits on the Java-Madura-Bali grid, with West Java as the main hub. PLN's President Director, Darmawan Prasodjo, said the growth in demand shows how important energy infrastructure has become to the digital economy, and the utility's enterprise sales head, Dini Sulistyawati, said the company believes this is just the beginning. PLN says it now wants to coordinate with developers from the earliest planning stage, with realistic schedules and demand projections.
Investors say the same thing from the other side of the table. Speaking at the APAC Finance Forum this week, Oki Ramadhana, Chief Executive of the Indonesia Investment Authority, said reliable power will determine how fast Indonesia's data centre market can scale, and that power availability is overtaking valuation as a consideration in the authority's data centre investments, with returns still the first test.
Capacity is tight today, and permits can stop a project
The build-out is answering a shortage. At the end of 2025, Jakarta had about 322 MW of operational capacity, with 186 MW under construction and 901 MW planned, according to an industry analysis published on 30 September. The same analysis cited Danantara's chief technology officer as saying customers currently wait six to twelve months for capacity.
Timelines are not fixed. On 2 October it was reported that a BDx data centre project had been halted over missing permits. The details of that case may change, but the lesson for anyone depending on new capacity is clear: announced dates are plans, and permits, grid connections and substations can move them.
What this means for enterprises
Most organisations will never lease a hall in one of these campuses. They will feel the build-out indirectly, through the hyperscalers and local providers that take space in them, through the price and availability of AI-grade capacity, and through the growing possibility of keeping sensitive workloads inside Indonesia. A few consequences follow.
- Plan capacity earlier than feels necessary. If local AI or colocation capacity is part of your roadmap for 2027, the conversations with providers belong in 2026, because the waiting lists are already months long.
- Separate AI-grade needs from general cloud. Training and heavy inference need dense, liquid-cooled racks of the kind the new campuses are built for. Everyday applications do not. Mixing the two in one request inflates cost and delays both.
- Treat data location as a design decision. More capacity inside Indonesia makes it easier to keep personal and regulated data in the country, which the new data protection rules reward. Decide which data must stay, which may leave, and build the architecture around that.
- Ask providers about power, not just price. A provider's grid allocation, substation schedule and backup arrangements now say more about delivery dates than a brochure does.
- Keep commitments reversible. Prefer contracts and architectures that let you move workloads between regions and providers, so a delayed campus or a cheaper option elsewhere is a change of plan rather than a crisis.
- Measure efficiency now. Electricity is the scarce input. Workloads that are right-sized, scheduled sensibly and observable will be cheaper to run wherever they end up.
A build-out that is also a dependency
The concentration of 99 per cent of data centre load on one grid is an opportunity and a risk at once. It is where the fibre, the customers and the skills are, and it is also where any power shortfall would be felt first. Operators are already responding: BDx is pursuing hydroelectric supply for its Jatiluhur campus, and PLN is building a new 150 kV substation for it. Enterprises can respond too, by keeping a second region in their architecture and by knowing exactly which systems would be affected if one site slowed down.
What to watch next
- Whether the Batam and Jatiluhur campuses meet their early 2027 opening dates, and at what phase sizes.
- The outcome of the halted BDx project and what it signals about permitting for the pipeline behind it.
- PLN's next data centre figures, which will show whether connected capacity is keeping pace with announcements.
- Grid investment outside Java, where PLN says Sumatra, Kalimantan and Sulawesi are emerging as expansion areas.
Sources
- Power and returns test Indonesia's AI data centre boom: APAC Finance Forum 2026 · The Tech Capital(opens in a new tab)
- PLN: Indonesia Data Centers' Power Consumption Reaches 2,378 GWh · Rancak Media(opens in a new tab)
- Indonesia's AI Build-Out Runs on Power. BDx Has Locked In 845MVA · Global Data Center Hub(opens in a new tab)
- Nvidia-backed Firmus plans 170,000-GPU Batam AI data centre · Tech Wire Asia(opens in a new tab)
- AI Startup Firmus Enters Indonesia with Nvidia-Backed Data Centre · Fintech News Indonesia(opens in a new tab)
- BDx Secures Indonesia's Largest Data Center Power Commitment With 1.2 GW PLN Agreement · DC Pulse(opens in a new tab)
This briefing was prepared by the Indotek editorial desk with AI assistance, from the public sources listed above. It is general information, not legal or professional advice.
Cloud & Data
How Indotek can help
Deciding what runs where, and keeping that decision reversible, is cloud and data architecture work. That is what Indotek's Cloud & Data practice does.
- Cloud architecture and migration: we assess each workload and choose its path, from re-host to re-architect, so AI-grade needs and everyday applications are planned separately.
- Security and residency by design: we let access, encryption and data-location requirements shape the architecture up front, particularly for regulated workloads.
- Platform engineering: infrastructure as code, observability and cost controls, so workloads can move between regions and providers without being rebuilt.
- Cost as an architectural property: we design for unit economics from the start, which matters most where electricity is the scarce input.


