Policy / Project Profile
Digital Infrastructure Bill — Public Consultation
Licensing for data centres and cloud services — the first time Singapore brings the AI compute layer under hard-law regulation. Data centres with ≥10MW critical IT load and IaaS/PaaS cloud services earning ≥S$100m/year in Singapore must obtain a "major FDI licence"; data centres ≥3MW must hold a "DC licence" and meet mandatory power usage effectiveness (PUE) requirements. Consultation ran 1–22 July 2026.
- Category
- Sector Regulation
- Published / Updated
- 2026-07
- Issuing body
- Ministry of Digital Development and Information (MDDI) / Infocomm Media Development Authority (IMDA)
- Lead ministry
- MDDI
- Parliament links
- 1 records
- Timeline years
- 2026
Strategic Context
1 parliamentary record linked.
Appears in the timeline around 2026.
Detailed Notes
On 1 July 2026, the Ministry of Digital Development and Information (MDDI) and the Infocomm Media Development Authority (IMDA) opened public consultation on the draft Digital Infrastructure Bill (DIB), closing 22 July 2026 at 10am. Its significance for Singapore's AI agenda: it is the first time the compute layer itself becomes a licensed industry, rather than regulating only the models and applications running on top of it.
Two licences. The first is the *major FDI licence* (foundational digital infrastructure), covering two service types: (1) data centre facility services provided in a DC with a critical IT load (CIL) of ≥ 10 megawatts serving unrelated third parties — i.e. cloud and co-location DCs; and (2) cloud computing services generating ≥ S$100 million per year in Singapore revenue averaged over the preceding three years, falling within IaaS or PaaS (but not SaaS). Licensees must implement physical and cybersecurity measures, business continuity and disaster recovery plans, and notify IMDA of cybersecurity incidents and service delivery disruptions. The second is the *DC licence*, with a lower threshold — every DC operator with CIL ≥ 3 megawatts must apply — focused on energy efficiency and sustainability. Operators meeting both criteria need both licences; IMDA will streamline the joint application.
Energy efficiency moves from voluntary to mandatory. In assessing a DC licence application, IMDA will weigh energy and water efficiency alongside operating experience, and may consider the renewability of energy sources, greenhouse gas emissions from electricity generation, and the economic or strategic importance of the applicant's operations to Singapore. Licensed operators must meet facility-level power usage effectiveness (PUE) requirements; the Bill also leaves room for IMDA to set IT-equipment energy efficiency and facility-level water efficiency requirements in future. This hardens the Green DC Roadmap — the government states plainly that "voluntary measures alone cannot ensure consistent sustainability outcomes across the sector."
Interface with existing law. The 2024 Cybersecurity Act amendments already imposed cybersecurity requirements on major FDI services, but there was no statutory framework for broader operational resilience. The DIB fills exactly that gap, while making related amendments to the Cybersecurity Act 2018 and the Cybersecurity (Amendment) Act 2024 to align the definitions of "foundational digital infrastructure service" and "data centre facility service". On enforcement, IMDA gains powers to grant, renew, suspend and revoke licences, issue codes of practice and directions, impose financial penalties, and conduct enforcement and investigation.
Why this matters for AI watchers. Singapore's consistent approach is to avoid a horizontal AI act and instead bring AI under existing sector statutes. The DIB is that methodology applied at the compute layer: it says nothing about models or training, yet by licensing data centres and cloud services it pulls the security, resilience and energy footprint of AI infrastructure into regulatory view. It also answers repeated parliamentary questions on one-north AI park energy consumption and AI's energy footprint — where the government's answer has consistently been that the regulatory centre of gravity sits at the infrastructure layer, not the model layer.
Resources
More in This Category
2021-10
Artificial Intelligence in Healthcare Guidelines (AIHGle)
Ministry of Health (MOH) / Health Sciences Authority (HSA) / Synapxe
2007
Health Products Act — AI-Medical Device (AI-MD) Regulation
Health Sciences Authority (HSA)
2017-03
Road Traffic Act — Autonomous Vehicle Trials and Use
Land Transport Authority (LTA) / Ministry of Transport (MOT)
2024-10
Guidelines on Securing AI Systems
Cyber Security Agency (CSA)