Given the lack of natural resources and a population of just 6.04 million as at June 2024, Singapore has long relied on strategic human capital planning to punch above its weight. According to Singapore’s Ministry of Trade and Industry, overall labour productivity (measured by real value-added per actual hour worked) rose 3.7% year-on-year in the first quarter of 2025. This contributed to a 3.9% expansion of Singapore’s economy in the same period.
However, that strategy is increasingly being challenged. The city-state’s total fertility rate remained at a historic low of 0.97 in 2024, and by 2030, nearly one in four citizens will be 65 or older.
Immigration has traditionally served as a pressure valve, with foreign workers making up about a third of the workforce. But even that lever is losing effectiveness as Singapore moves into new growth areas, such as artificial intelligence (AI), biomedical sciences and sustainability, where global talent is scarce.
With an ageing workforce and talent shortages in critical sectors, humanoids could play an essential role in narrowing the gap between Singapore’s growth ambitions and its labour constraints. Designed to move, sense, and interact with their surroundings, those human-like robots are increasingly seen as a transformative tool for boosting productivity in labour-intensive sectors.
“Singapore has been looking at using robotics in sectors that are important to its economy since the formation of the National Robotics Programme (NRP) in 2016. Today, we’re increasingly seeing the aviation, sea port, manufacturing, healthcare and built environment sectors beginning to explore the use of humanoids to raise productivity and address other industry challenges,” says Tung Meng Fai, NRP’s executive director, at a July 24 panel discussion hosted by Microsoft Research Asia in Singapore.
Last year, the NRP received an additional $60 million in funding to scale adoption of market-ready advanced robotics in sectors such as healthcare, facilities management, manufacturing and logistics. This builds on the initial $450 million invested by the National Research Foundation, which supported the development of a robotics middleware framework and the training of over 700 research scientists, engineers and professionals in Singapore.
See also: AI agents expected to drive revenue by Apac CFOs, adoption rising in Southeast Asia
In parallel, Singapore is laying down complementary infrastructure such as AI compute capacity, 5G connectivity and interoperability frameworks. These are critical enablers for real-time robot deployment in dense urban environments.
The endgame is not just automation, but true human-robot collaboration. That vision is taking shape in the Punggol Digital District (PDD), a 50ha smart precinct expected to be completed in 2026. PDD runs on an Open Digital Platform (ODP), which allows all compatible robots and digital systems to interact and be centrally managed.
Through the ODP, estate managers can track robot locations, receive system alerts, and resolve operational issues in real time. Energy use, temperature, and movement data are visualised on dashboards and overlaid on a 3D digital twin (or virtual replica), turning PDD into a living command centre for robotic urban operations.
See also: Traders are fleeing stocks feared to be under threat from AI
The economics of a humanoid workforce
Morgan Stanley Research estimates that 2 million workers, which is nearly half of Singapore’s workforce, could eventually benefit from humanoids or robotics. “We see potential for humanoids/robotics to supplement 52% of Singapore’s workforce that is more manual labour-related. From an industry perspective, we believe that manufacturing and construction possess the highest humanoid optionality [or the ability to be replaced by humanoids],” wrote the analysts in their report titled Singapore at 60: Unlocking Wealth Creation.
That shift is already underway. Two autonomous robots are currently being trialled at Changi Airport’s Terminal 4, where they patrol public areas using cameras, 3D light detection, sound navigation, and other systems that enable them to localise themselves, navigate safely indoors, and avoid collisions. Called Gibson, they are developed by the Home Team Science and Technology Agency in collaboration with the A*STAR Institute for Infocomm Research. Additionally, they can be summoned by the airport police via a mobile app and be ridden to enable faster response to incidents.
Humanoids are edging toward economic viability, too. With median non-PMET wages in Singapore at around $3,000 a month, a humanoid robot costing US$150,000 ($192,922) on average to deploy could break even in five to six years, or sooner if it is used across multiple shifts. Costs are expected to fall further as scale builds. “Assuming US$50,000 and a 10-year useful life, we estimate a cumulative cost saving of US$222,000 deploying a humanoid versus an average non-PMET worker in Singapore,” note the Morgan Stanley Research analysts.
While there are concerns about job losses, the analysts believe that Singapore workers will be able to easily adapt or be reskilled to roles augmented by robotics or AI, given that 43% hold at least a degree qualification. Moreover, the government has been rolling out several initiatives, such as SkillsFuture, to prepare the workforce for the advent of new technologies, including humanoids.
Smarter foundations needed
According to the International Federation of Robotics, Singapore ranks second globally in robot density. The number of robots per 10,000 workers rose from 488 in 2016 to 770 in 2023.
To stay ahead of the latest tech trends, click here for DigitalEdge Section
Ashley Llorens, corporate vice-president and managing director of Microsoft Research Accelerator, expects humanoids to gain greater traction in enterprise use than traditional robot arms or autonomous robots that can only execute specific tasks.
He explains: “If you have very predictable operations, you’ll benefit from a production line made up of highly specialised robots. But if that workflow changes, it’s a painful process [to replace or retrain those robots] and every minute of downtime costs money. [Humanoids are better for operations that evolve as they are] general-purpose systems.”
Increasing adoption also calls for significant advancements in some fundamental aspects of humanoids.
One is the need for more capable multimodal foundation models. “Previously, a foundation model focused on a specific task. Now, we are working to develop a multimodal foundation model for robots like humanoids, in which one model can handle multiple tasks,” says Lin Shao, assistant professor at the National University of Singapore, at the same July 24 panel discussion.
Llorens agrees. He says: “[A humanoid] needs to perceive and understand the environment before it can plan and carry out actions. Multimodal foundation models currently enable robots to achieve the former, but significant development is still needed in the planning and action-taking phases. This will require models possessing enhanced reasoning capabilities and the capacity for seamless, intuitive human collaboration.”
Chong Jiayi, chief technology officer of Singapore-based dConstruct Robotics, believes humanoids are still far from ready for mass, real-world deployment. “[Humanoids] are complex as they need to receive multiple inputs (such as visuals, touch and acoustics) and some of the corresponding actions, like walking stably or grabbing things with the right pressure, are still challenging for them. More so, to do it reliably and safely at scale,” he explains.
Even selecting the right AI approach for each task can be difficult. “For instance, a humanoid can solve a vision- or perception-based problem in various ways. One is to apply a much heavier foundational model. The other is to train a much smaller neural network with a more efficient architecture. Or maybe I won’t even train a model — I could just apply a classical computer vision approach and write the algorithm directly. None of these answers are wrong; it’s about picking the right AI tool for the right job,” he adds. A neural network is a type of machine learning model inspired by the human brain that can be trained to recognise patterns and make decisions.
Access to diverse and high-quality data is also critical for humanoids, says Shao. “Real-world data from human experts [is rich and] doesn’t suffer from the domain gaps that often exist in simulation. But simulation data can be generated quickly, at scale and lower costs. We need a system that can bridge both types of data and digest them [to gain more holistic insights and outcomes] instead of wasting any piece of information.”
Bubble or not?
Despite being a nascent area, humanoids are here to stay. Morgan Stanley Research estimates that there could be more than 1 billion humanoids in use globally by 2050, with 90% used for industrial and commercial purposes.
“[Humanoids are] definitely a bubble, but it doesn’t mean it’s something bad. Every tech bubble has wiped out companies, but also laid the foundations for long-term value. For instance, many companies went bankrupt in the dot-com bubble, but the Internet survived and brought tremendous value over a much longer time frame. I believe we’ll see the same with machine learning, or what people now call AI,” says Chong.
He continues: “This deep learning-based approach solves fundamental problems and offers transformative architectures. It’s becoming another essential toolset across domains. What’s interesting about this is that we have to reformulate complex challenges into differentiable problems that can be optimised. That shift in thinking unlocks solutions in fields well beyond conventional AI applications.”
Shao echoes that sentiment. “Three years ago, most of us would’ve said ChatGPT wouldn’t work. But AI doesn’t evolve linearly, and neither does the world. You’ll make a mistake if you extrapolate based only on the present. From a research standpoint, it’s not just about what’s deployable today — it’s about exploring what’s possible tomorrow.”
For Tung, the long view is what matters. “We recognise a lot of hardware innovation is coming from China and many robotics companies in Singapore have adapted those systems for local use cases. It’s unrealistic to assume we’ll replicate China’s robotics ecosystem [which is more mature and comprehensive]. But that doesn’t mean Singapore lacks ambition. We’re thinking carefully about where we can punch above our weight and how we can synergise with different ecosystems around the world.”
“[Humanoids] is still a nascent field, which allows Singapore to invest deliberately in targeted areas that will contribute meaningfully to global progress,” he says.
