By the Numbers: Singapore's War on Duplicate Images Is Bigger Than You Think

Singapore's digital storage bill is quietly ballooning. Across government data centres, e-commerce warehouses and news archives, duplicate image files — identical or near-identical photographs stored multiple times across different servers — now account for an estimated 18 to 22 percent of total unstructured data held by large enterprises, according to findings published by the Infocomm Media Development Authority in its 2025 Digital Infrastructure Review. That single category of redundancy is costing organisations here tens of millions of dollars annually in avoidable cloud storage fees.

The timing matters. Singapore is midway through its Smart Nation 2.0 roadmap, which commits the government to slashing its own data centre energy consumption by 10 percent before 2028. Duplicate images are a deceptively large contributor to that footprint. A single high-resolution product photograph uploaded to an e-commerce platform can be automatically resized, compressed and cached into six or more derivative files, each stored separately. Multiply that across a catalogue of 500,000 listings and the redundancy compounds fast.

Where the Problem Lives

The issue is most acute in three sectors: retail tech, public sector archives and media. Lazada's Singapore logistics hub in Jurong, which processes product imagery for Southeast Asian markets, has publicly acknowledged migrating to a deduplication-first cloud architecture, though the company has not disclosed specific storage savings figures. The National Archives of Singapore, housed on Canning Rise, manages a digitised collection that spans more than two million records; its 2024 annual report noted that a deduplication pass conducted on its photographic holdings between January and June 2024 identified approximately 340,000 redundant image files, freeing up roughly 4.7 terabytes of storage.

At Mapletree Business City in Pasir Panjang, several multinational tenants running regional creative and marketing operations have begun mandating perceptual hashing — a technique that identifies visually similar images even when file names and metadata differ — as a standard step in their digital asset management workflows. Perceptual hashing can detect duplicates that a simple file-size or checksum comparison would miss, which is critical when images have been slightly cropped or re-exported at different resolutions.

The monetary stakes are concrete. Hyperscale cloud storage in Singapore — where data sovereignty rules push many companies to keep locally compliant copies — runs at roughly SGD 0.023 per gigabyte per month on standard tiers as of mid-2026, based on publicly listed rates from major providers operating out of the Tuas and Jurong data centre corridors. An organisation holding 500 terabytes of image data with a 20 percent duplication rate is paying for 100 terabytes it does not need. At current local rates, that works out to approximately SGD 27,600 a month, or more than SGD 330,000 a year, in pure waste before factoring in egress costs or backup storage.

What Organisations Are Doing About It

The IMDA's advisory framework, updated in March 2026, now recommends that public agencies run deduplication audits at least once every 12 months and adopt content-addressable storage systems for new image repositories. Several statutory boards have already begun compliance reviews ahead of the government's FY2027 budget cycle, which is expected to tie data efficiency benchmarks to ICT procurement approvals.

For private enterprises, the practical pathway is more incremental. Tools built on perceptual hashing algorithms — several of which are already integrated into platforms like Adobe Experience Manager and open-source alternatives maintained by communities based partly in Singapore's developer scene at spaces like Blk 71 in one-north — can automate duplicate flagging without requiring a full system overhaul. The typical enterprise deduplication project runs three to six months for initial audit and remediation, according to publicly available project timelines from government-linked technology integrators.

Companies that delay face a compounding problem. As generative AI tools produce more image variants at scale — a trend accelerating across Singapore's marketing and retail sectors — the duplication rate is expected to climb. Organisations that build deduplication into their intake pipelines now, rather than treating it as a cleanup exercise later, will find both their storage invoices and their carbon reporting in noticeably better shape when Smart Nation 2.0 milestones come due in 2028.