Grab’s acquisition of robotics company Infermove is a strategic response to a stubborn delivery problem: customers expect speed and low fees while labour, fuel and regulatory costs continue to pressure margins.
Rather than buying a finished robot from a vendor, Grab is bringing the development loop inside the company. That choice gives the Southeast Asian platform greater control over where automation operates, how quickly it improves and which economic trade-offs it makes.
Real streets teach what simulation misses
Infermove builds robots for unstructured environments. Its technology can learn from real-world movement data, including information generated by non-motorised delivery vehicles. For Grab, that means training around how people actually navigate pavements, crossings and crowded drop-off points.
Simulations are useful, but dense Southeast Asian cities are defined by exceptions. Weather changes, street layouts differ and human movement is difficult to predict. A system trained only on an idealised environment may perform well until the first unusual curb, obstruction or busy entrance.
Grab already operates millions of deliveries and generates extensive mobility data. Connecting that information directly to robot development can shorten the feedback loop between a failure in the street and an engineering improvement. It also keeps sensitive operational data closer to the company.
The acquisition is therefore as much about learning capacity as machines. Infermove will continue operating independently, with its founder reporting to Grab Chief Technology Officer Suthen Thomas. That structure suggests an effort to preserve execution speed while connecting the team to Grab’s scale.
Robots take segments, not entire routes
Grab is not presenting automation as a wholesale replacement for riders. Scooters and bicycles remain effective in complex urban networks where humans can adapt quickly. Robots are more likely to handle selected first- or last-mile segments that are short, repetitive and comparatively structured.
That approach could help smooth peak demand, reduce delays and add capacity when labour is scarce. The economic question is whether the robot lowers cost per delivery without creating new failure points, supervision burdens or customer friction.
Live performance matters more than broad market forecasts. A vehicle may work technically but still struggle with local rules, tropical weather, building access or public acceptance. Operating across multiple countries compounds the challenge because infrastructure and regulation vary widely.
Selective deployment makes those variables manageable. Grab can begin where route conditions and task boundaries are clearest, compare performance with human delivery and expand only when the economics hold.
Owning robotics tightens operational control
Large platforms increasingly treat AI as part of core operations rather than an extra software layer. In logistics, that shift reaches into the physical world, where development is costlier and mistakes are more visible.
Internal ownership lets Grab align robotics with its own regional footprint instead of waiting for a supplier whose priorities may differ. It can choose deployment pace, integrate robots with dispatch systems and use its data to optimise the complete delivery chain.
There is also a strategic dependency argument. A third-party robotics provider controls product direction and pricing. An acquired team gives Grab more influence over both, although it also transfers the expense and risk of development onto Grab.
The Infermove deal is not primarily a bet that pavement robots will replace the delivery workforce. It is a bet that physical AI has matured enough to improve carefully chosen parts of a large logistics network.
That distinction keeps the strategy grounded. Grab does not need a robot capable of every journey to create value. It needs systems that handle specific segments reliably, learn from the streets they serve and lower operating cost without degrading service.
By pulling robotics closer to its data and operations, Grab can test that proposition on its own terms. If it succeeds, the advantage will come not from owning a futuristic machine, but from making automation an ordinary, dependable part of how millions of deliveries move.
