Singapore’s Physical AI Push: Punggol Becomes a Real-World Robot Testbed

Singapore has spent years wiring intelligence into the city. Now it wants that intelligence to grow wheels, legs, sensors, and a job to do.

Later in 2026, Punggol Digital District will become Singapore’s first mixed-use public testbed where robots from multiple operators can work alongside people at scale. The initiative, led by the Infocomm Media Development Authority (IMDA), JTC, and the Singapore Institute of Technology (SIT), will put AI-powered machines to work on food and parcel deliveries, cleaning, security patrols, and other everyday tasks.

For Singapore, the experiment is about far more than seeing whether robots can navigate a pavement without bumping into someone carrying kopi. Punggol could help answer a much bigger question: What happens when artificial intelligence leaves the chatbot window and becomes part of a city's physical infrastructure?

Punggol becomes Singapore’s physical AI laboratory

The project brings together eight industry partners and several government agencies in what Singapore is calling a living testbed for robotics and embodied AI.

Certis, DHL, Grab, and QuikBot are among the initial companies expected to help design and test commercial robot services in the district. According to details of the physical AI project released by IMDA, food and parcel delivery robots could supplement existing delivery operations, while other machines could perform cleaning and security patrols, including outside normal operating hours or in areas that are difficult for workers to access.

That makes Punggol different from the carefully controlled robotics demonstrations that have become common at technology conferences.

As eWeek recently explored in its guide to what companies should know before deploying humanoid robots, the difficult part of robotics is not producing an impressive demonstration. The real test is whether machines can perform narrow, useful tasks repeatedly while dealing with people, changing environments, safety requirements, data, and operational systems.

Punggol is designed to expose robots to exactly that kind of messiness.

The 50-hectare Punggol Digital District combines businesses, SIT's campus, retail, dining, and community spaces. JTC expects the district to support about 28,000 jobs and 12,000 students, meaning robots deployed there will not be operating behind factory doors. They will have to function in a place where workers, students, visitors, deliveries, buildings, security systems, and public paths collide every day.

The city itself becomes part of the robot

For physical AI to become useful, the robot cannot be the only thing that is smart. A delivery machine needs to know where it is going, but it may also need access through a gantry, permission to use a lift, information about obstacles, and a way to communicate with surrounding systems.

Punggol already has infrastructure designed to make the district itself more connected.

At its center is an Open Digital Platform that JTC describes as an integrated operating system, providing businesses and students with access to real-time district data and using data-driven insights to optimize operations across Punggol Digital District. That architecture hints at where physical AI may be heading.

Instead of making every robot independently solve every problem it encounters, buildings and cities could eventually become active participants. Digital infrastructure could provide machines with information about their surroundings, while fleet systems could coordinate multiple robots operating within the same environment.

In that sense, Singapore is not simply testing smarter robots. It is experimenting with what happens when the environment around them becomes increasingly machine-readable too.

Singapore is also testing the rules robots will need

Software is only one part of the problem. A robot that moves through a public area also needs rules governing where it can travel, how it behaves around pedestrians, and what happens when something goes wrong.

Singapore is addressing that problem directly at Punggol.

As part of the project, Singapore's Land Transport Authority is introducing a precinct-level exemption framework under the Active Mobility Act. Instead of requiring autonomous mobile robot operators to seek separate exemptions for each trial, approved operators will have greater flexibility to experiment in Punggol while remaining subject to safety conditions.

That regulatory layer may prove just as important as advances in robot hardware.

AI companies have become very good at demonstrating what machines can do. Governments and businesses still have to decide what those machines should be allowed to do when they share space with humans.

Punggol gives Singapore a contained environment in which it can develop both sides of that equation at once.

Physical AI fits a much bigger Singapore strategy

The robotics push is arriving as Singapore aggressively expands its position in the wider AI economy.

At the 2026 Asia Tech x Singapore summit, the government unveiled new initiatives with OpenAI, Nvidia, Google, and other technology companies, emphasizing the move of AI from experimentation to real-world deployment.

OpenAI alone has committed more than S$300 million to Singapore and plans to establish its first Applied AI Lab outside the United States there. As eWeek reported when the Singapore lab was announced, the initiative is intended to help organizations adapt AI systems for practical business and government applications rather than focusing solely on developing new foundation models.

Robotics gives that strategy a physical counterpart.

Nvidia has been pushing heavily into "physical AI," including models, simulation tools, computing platforms, and reference systems intended to help machines perceive and act in the real world.

Singapore already has a foot in that ecosystem through homegrown robotics company Sharpa. The company worked with Nvidia and Chinese robotics maker Unitree on a humanoid research platform unveiled earlier this year, a collaboration eWeek covered as part of the growing physical AI race.

Sharpa is also working directly with JTC. In April, the Singapore-based company announced plans with JTC to deploy its North autonomous robots in Punggol, alongside an innovation lab in the district.

Singapore therefore has something more valuable than another robotics showroom: a potential pipeline connecting research, local companies, global AI platforms, government agencies, universities, and public deployment.

Robots arrive as Singapore's workforce gets older

There is also a distinctly Singaporean reason to pay attention to what happens in Punggol.

The country is becoming a super-aged society. In March, Manpower Minister Tan See Leng said Singapore would reach that demographic milestone in 2026, even as AI transforms the nature of work.

The shift is already visible in the workforce. Singapore's resident labour force participation rate fell from 70.5% in 2021 to 67.9% in 2025, with population aging cited as a major factor. That gives robotics a different economic context here than it might have in a country facing plentiful labour.

Cleaning, security, logistics, healthcare, construction, and other physically demanding sectors all have jobs where automation could supplement a constrained workforce. The Ministry of Manpower has specifically identified cleaning, security, and food services as areas where technology can reduce physically demanding work for older employees.

The government is careful, however, not to frame AI adoption simply as a replacement for workers. In February, MOM said Singapore's labour market remained resilient despite concerns about AI-related job losses. The government has also set aside more than S$400 million to support workforce transformation, including helping companies redesign jobs, reskill workers, and adopt AI-powered technologies.

That tension will follow the robots into Punggol.

A machine handling overnight security patrols or repetitive deliveries might ease manpower constraints. But every successful deployment also forces companies and workers to reconsider which tasks belong to humans, which can be automated, and what new skills workers need when their colleague is a mobile AI system.

The robot race is moving closer to everyday life

Across Asia, the development of humanoid and autonomous robots has accelerated rapidly.

The competition stretches from industrial automation to increasingly sophisticated humanoids. Nvidia's collaboration with Unitree and Singapore-based Sharpa is one example of how companies across the region are combining AI computing, robotics hardware, and embodied intelligence, a trend eWeek examined through the companies' humanoid robotics collaboration.

But technical capability alone will not determine the winners.

Robots also need places where companies can learn how to deploy them reliably, integrate them with existing infrastructure, satisfy regulators, protect the public, and determine whether the economics make sense. That is the more interesting bet Singapore is making at Punggol.

The district does not need to produce the world's flashiest robot. It could prove more valuable if it produces something much less glamorous: a repeatable playbook for getting robots out of laboratories and into ordinary life.

If that experiment works, Singapore may be showing everyone else how to build a city where machines actually know how to live.

Also read: For another sign of Singapore’s widening AI ambitions, see how the country is deepening cooperation with Australia on AI safety, cybersecurity, and digital infrastructure.