Your smartphone feels clean – no exhaust pipe, no smokestack. Yet streaming an hour of video uses about 0.037 kWh of electricity, equivalent to running a ceiling fan for about three to four hours.
When scaled across billions of global searches, streams and artificial intelligence (AI) interactions each day, the energy demand behind routine digital activity is substantial.
“The environmental impact of our digital consumption is far more significant than most people realise,” said Dr Lawrence Wee, director of business and ecosystems at Singapore’s Infocomm Media Development Authority (IMDA). As part of IMDA’s BizTech Group, his team focuses on advancing the country’s digital innovation and sustainability across the tech ecosystem.
“We are seeing a digital paradox: the very technologies that offer solutions for climate change are simultaneously driving up energy consumption.”
Studies over the past several years have placed the Information and Communications Technology (ICT)’s share of global greenhouse gas emissions between 1.8 and 2.8 per cent, with a more recent analysis estimating it at around 3.4 per cent. Data centres – facilities that house servers and keep the internet, cloud services and digital applications running – are among the most energy-intensive components of this ecosystem because they require continuous power and cooling.
Singapore is no exception to this trend. The city-state hosts one of the region’s densest clusters of data centres, with about 1.4 gigawatts (GW) of computing capacity from 70 data centres, which power everything from e-payments to food-delivery apps.
By regional standards, this capacity is unusually concentrated. Unlike neighbouring countries such as Indonesia, Malaysia or Thailand, where data centres are spread across vast land areas and multiple power grids, Singapore compresses a similar scale of digital infrastructure into just 734 square kilometres, intensifying pressure on electricity supply, land use and carbon limits.
“As a small island nation, climate change poses an existential threat,” Wee said. “Digital sustainability must therefore be a national priority.”
Dr Lawrence Wee, director of Business and Ecosystems at IMDA says digital sustainability must be a national priority for Singapore. Image: Mark Stoop on Unsplash
In fact, the city-state has been working to improve energy efficiency in its digital infrastructure for more than a decade, driven by structural constraints around land and power.
Throughout this period, IMDA has played a central role in shaping Singapore’s approach, working with industry to move energy efficiency from a best-practice goal to a regulatory and operational norm.
One early milestone was its collaboration with industry to establish one of the world’s first energy-efficiency standards for data centres operating in tropical climates, encouraging facilities to run at higher operating temperatures while maintaining performance. The framework helped set expectations for how digital infrastructure could grow within Singapore’s physical and energy limits.
These hardware-oriented measures helped stabilise energy use, but the rapid expansion of cloud services and AI soon highlighted the need for a more holistic approach.
“Optimising physical infrastructure alone is no longer enough,” Wee said. “The energy consumed by hardware ultimately translates to powering the applications and AI running on it.”
This prompted IMDA to develop a digital sustainability framework covering government operations, the ICT sector and the wider economy. It supports sustainable IT procurement, promotes energy-efficient development practices and provides tools to help businesses assess the energy impact of their software.
A key part of this effort is the growing focus on green software, a concept still unfamiliar to many users but increasingly relevant to organisations managing large digital workloads.
Designing software with energy in mind
Navveen Balani, managing director for tech sustainability innovation at Accenture, describes green software by analogy: an appliance that performs the same task using less electricity.
A key challenge is that measuring the environmental impact of software is not straightforward and can take many forms. Companies today rely on a mix of approaches, ranging from high-level dashboards provided by cloud service providers to more detailed metrics at the application or AI-model level.
Software efficiency can be improved in several ways, such as by reducing unnecessary data movement, writing code that executes tasks more efficiently, selecting smaller AI models where appropriate or scheduling certain workloads for times or locations where cleaner energy is available. These adjustments reduce the energy required to deliver the same output.
Balani said many organisations still lack visibility into how their applications consume electricity or generate emissions.
“The moment they start measuring, they see where the real inefficiencies are,” he said.
Some workloads may run continuously without necessity, and some AI models may be larger than required for the intended function.
The efficiency gains translate directly to business benefits. Research has shown that adopting green computing practices can lower energy usage by 17 per cent on average, with some cases achieving reductions as high as 90 per cent. These improvements cut electricity costs while maintaining – or even improving – application performance.
AI systems add further complexity. Developers sometimes adopt large, general-purpose models by default, even when smaller ones could achieve comparable results. Emerging multi-step “agentic” workflows can multiply energy use if not designed with efficiency in mind.
Navveen Balani, managing director for tech sustainability innovation at Accenture, said many organisations still lack visibility into how their applications consume electricity or generate emissions. Image: Samuel Sianipar on Unsplash
Tracking what software consumes
Understanding these patterns has become central to IMDA’s Green Software Trials, which bring together cloud providers, system integrators and enterprise users to study how coding choices, workload management and AI model selection affect energy use and performance.
The trials involve technology firms and digital application developers applying software carbon reduction techniques to real-world applications. Accenture, NCS, and FastCo have put these testing methods such as resource redistribution, application modernisation, AI optimisation and computational offload on their live systems.
Early results showed that such techniques can deliver cost and energy efficiency without compromising performance, generating data that will inform industry guidelines on green software development.
To identify where emissions are coming from and which changes will have the greatest impact, it is increasingly important to have consistency and comparability across these methods.
This led to an increased in the adoption of the Software Carbon Intensity (SCI) specification developed by the Green Software Foundation, which was recently published as an ISO standard. The SCI provides a common methodology for calculating the carbon footprint of software systems, offering a shared reference point that can complement existing dashboards and tools.
According to Wee, clearer measurement is desirable. It allows organisations to better understand the environmental impact of their software and identify where optimisation efforts, such as code choices or computing workloads, can deliver the greatest reductions.
He noted that measurement challenges and limited internal expertise remain key reasons some firms have not yet acted, and that trials with industry partners help develop skills in emissions-aware engineering.
Governments, businesses reassess AI-led growth
As attention shifts from hardware efficiency to software and AI, governments are increasingly reassessing how fast-expanding digital ambitions align with energy and climate constraints.
Marie Teo, senior advisor at the Tony Blair Institute for Global Change, said the energy profile of AI systems is becoming a strategic consideration for policymakers and business owners, not just a technical one.
AI can support economic growth and public-sector innovation, but sustained expansion depends on the availability of low-carbon power.
“Digital sustainability is not just about reducing environmental impact,” Teo said. “It’s about ensuring governments can fully seize the AI opportunity without being constrained by energy and emissions bottlenecks.”
Panel discussion at Digital Sustainability Forum, ATX Summit. Image: IMDA
The same tension is emerging for businesses scaling AI across operations from data analytics and automation to customer-facing services, where rising compute demand can quickly translate into higher energy use and costs.
Against this backdrop, Teo pointed to Singapore as an early mover. She said the city-state was among the first to introduce energy-efficiency standards for tropical data centres and is now extending that thinking across the digital value chain, including software and AI.
That more holistic approach, she added, offers a reference point for other economies facing similar land, energy or carbon constraints.
Singapore as an early mover
Building on that approach, Singapore is moving beyond policy signalling to practical implementation, positioning itself as an early mover in addressing the questions Teo highlighted – how AI-driven growth can advance within tight energy and climate constraints.
A core pillar of this shift is IMDA’s Green Software Trials, which examine how efficiency techniques perform in live systems. The trials test approaches such as optimising software architecture, right-sizing AI models and reducing unnecessary data movement, generating evidence on how energy use can be reduced without compromising performance.
Funding support under the Advanced Digital Solutions programme complements this work by helping companies adopt sustainability-focused digital tools and scale deployment across operations.
Beyond domestic implementation, Singapore is also contributing to the development of global standards in sustainable computing through its participation in the Green Software Foundation, enabling knowledge exchange and shaping emerging best practices.
To support adoption, IMDA is working across industry and academia to validate applications, build local capabilities and address skills gaps that remain a barrier to wider deployment.
Wee said these efforts support Singapore’s ambition to position itself as a “Green Computing Hub,” linking digital innovation with environmental stewardship.
Green Software Foundation (GSF), Global Summit Singapore. Image: IMDA
The urgency of this transition is growing. Digital activity is expected to continue rising as AI becomes embedded across industries, and energy demand is projected to increase even with improvements in chip efficiency.
Teo said countries that align AI development with clean energy expansion may gain an advantage in attracting investment from global technology providers.
For Singapore, the ongoing challenge is balancing digital competitiveness with climate ambitions. Wee said the goal is to ensure innovation “is fundamentally sustainable,” highlighting the role of software efficiency and greener AI systems in future growth.
The electricity behind every stream, search and AI query is increasingly part of Singapore’s broader sustainability conversation. The stakes are rising as surveys show Singaporeans are among the world’s heaviest users of digital services, with some of Asia’s highest streaming uptake and one of the fastest rates of enterprise adoption of AI tools.
As more of daily life moves online, software design and digital infrastructure are expected to play an increasingly important role in meeting Singapore’s long-term environmental goals, as the city-state seeks to cap emissions at around 60 million tonnes by 2030 and reach net zero by 2050.
“Singapore is driving the conversation, regionally and globally, on how to ensure that technology is not just fast and scalable, but fundamentally sustainable,” Wee said.
