I was watching the live MWC25 Las Vegas keynotes this week when a fireside chat on delivery robotics piqued my interest.
Ali Kashani, CEO of Serve Robotics, said, "We want to be a shared platform. We want to be an infrastructure that everybody can use." That vision captures exactly where delivery robotics is heading, and where it already stands today.
I remember the first time I saw an autonomous delivery robot moving slowly down a sidewalk in Zaragoza, Spain, in 2022. It looked almost surreal, a piece of the future quietly navigating a human world I had only seen in China and Dubai before.
Goggo Network, the first firm in Spain to receive a logistics licence for autonomous robots, owned it. One of the first cities in Europe and the first in Spain to start an autonomous logistics operation was Zaragoza. The project was unfortunately shut down because of financial limitations.
Since then, I have followed the journey of these small but mighty machines. Today, autonomous delivery robots are real and currently deployed around the world.
In this piece, I want to look at how far delivery robotics has come, who is leading the way in real-world deployment, what challenges still exist, and what intelligent automation really means when it starts delivering at scale.
From Concept to Roaming City Streets
The history of delivery robotics started in laboratories and research facilities. Early robots worked in strictly regulated settings, such as test tracks or college campuses. Their initial greatest challenges were public trust, safety, and navigation.
However, these robots become more like the actual world thanks to developments in sensors, AI algorithms, edge computing, and lightweight design.
Now we see them navigating sidewalks, stopping at crosswalks, and even interacting politely with humans who cross their path. Designing machines for human environments is one of the most complex engineering challenges.
"If you have conviction that this is the future, the biggest hurdle is acceptance by people" - Ali Kashani
At MWC25 Las Vegas, Ali Kashani shared how Serve Robotics has scaled from 1,000 to nearly 2,000 deployed robots across U.S. cities. He said, "If you have conviction that this is the future, the biggest hurdle is acceptance by people."
Robots can be perfectly functional in theory, but they only succeed if people trust and accept them.
Serve's model focuses on the first and last mile of delivery. Their compact, autonomous robots move safely in pedestrian spaces and aim to provide shared infrastructure that other services can use.
The goal is not to replace human workers but to complement them by handling repetitive, low-margin deliveries that slow down logistics networks.
Businesses Enabling Self-Driving Delivery
Serve Robotics recently announced a new generation of smarter, safer, and more energy-efficient robots that operate autonomously in busy cities. They are also working to open their platform to multiple partners, allowing various businesses to share robotic delivery infrastructure.
Serve is a component of an exponentially growing ecosystem.
Starship Technologies (USA and EU) works on residential neighbourhoods and college campuses. It prioritises safety, dependability, and neighbourhood inclusion because it delivers more than 9 million packages annually. Ahti Heinla and Janus Friis, co-founders of Skype, launched it.
Nuro (USA) operates small autonomous vehicles designed for road delivery, and unlike sidewalk robots, they use regular roads to deliver groceries and retail products. Nuro has partnered with major brands like Kroger and Walmart to bring automation to suburban neighbourhoods. Co-founded by Jiajun Zhu and Dave Ferguson.
Kiwibot (Latin America and USA) deploys small robots to distribute food in urban areas. The business now operates in Colombia, the US, and several Asian countries. Their approach focuses on offering affordable, eco-friendly solutions and merging with nearby eateries.
Cartken (Asia and USA) builds autonomous sidewalk robots that specialise in short-distance deliveries. Their robots are modular, adaptable, and designed to operate in both indoor and outdoor environments. They have gained attention for their partnerships with universities and tech campuses. Founded by ex-Google engineers
Each company is solving different problems, but all are working toward a common vision of more efficient and scalable delivery systems powered by automation.
Intelligent Automation
When people think about automation, they often picture software running quietly in the background. But what happens when intelligence becomes physical? When AI steps off the screen and starts moving through our cities, the world begins to change.
Ali Kashani described this as "physical AI." During his MWC25 shares, he explained that we are entering what he calls the second machine age. Engines and machinery powered the first one. The next one, powered by AI and robotics, will bring intelligence into physical form.
Physical AI is tangible and accountable, it interacts with humans and must therefore earn trust through action, not promise
He also shared that he prefers working on hardware now because, as he put it, "AI looks really scary today." His reasoning is that physical AI is tangible and accountable. It interacts with humans and must therefore earn trust through action, not promise.
Serve Robotics, for example, has learned to design its robots around people. They do not build in isolation; they involve communities and teams at every stage. Kashani said, "We go to where people embrace us." That simple sentence captures the essence of successful innovation. Technology only thrives when people are ready to adopt it.
The Dreaded Challenges Ahead
Despite the rapid progress, let's talk about the pending challenges. Regulations for sidewalk and street robots are still inconsistent across countries. Cities must decide who owns the sidewalks, how robots share space with pedestrians, and who is responsible if something goes wrong.
Governments need to find ways to balance innovation with public safety.
Then there is the human side of the equation. Acceptance, as Kashani mentioned, is one of the greatest hurdles. People need to feel safe and comfortable sharing space with robots. Public perception can change quickly, and one high-profile incident could slow adoption.
Weather, uneven sidewalks, pets, children, and human unpredictability all test the limits of AI navigation
Economics is another challenge. Building, maintaining, and operating autonomous robots requires investment. While costs are coming down, achieving profitability depends on scale and consistent utilisation.
And finally, technology still faces the complexity of real-world environments. Weather, uneven sidewalks, pets, children, and human unpredictability all test the limits of AI navigation.
Quantum Innovation in Robotics
We are witnessing the birth of an entirely new layer of intelligent infrastructure. The leap from digital AI to physical AI is happening as we speak.
Robotics will augment people. Intelligent automation should complement human skills, not compete with them
The impact could go far beyond food or parcel delivery. Imagine healthcare robots transporting medical samples or disaster-response robots navigating dangerous zones.
Robotics will augment people. Intelligent automation should complement human skills, not compete with them.
The Future of Physical AI
As we move further into this second machine age, I believe robots will become part of everyday life. Neighbourhoods will have shared robotic fleets. Cities will include robotic infrastructure in their urban planning. Businesses will rely on robots to move products seamlessly between warehouses, stores, homes, and more that we have yet to envision.
The companies leading today, such as Serve, Starship, Nuro, Kiwibot, and Cartken, are setting the foundation for that future.
Their work reminds us that innovation takes time, and progress is often quieter than headlines suggest.
Now is the moment to engage. The physical AI revolution is unfolding, and as we enter the last quarter of 2025, we will take this conversation into 2026.
Get ready, physical AI is going to be a major trend for next year.
