Serve Robotics launches Physical AI Services for other robot makers
By Anika SharmaAI & Innovations Editor, Peepals Tech
Representative image: bohacekmarek / Pixabay
Serve Robotics launched Physical AI Services on 5 November 2026, offering its robot-fleet connectivity, data and training tools to other robot makers.
What Happened
Serve Robotics, a US company that runs fleets of small sidewalk delivery robots, launched a new business called Physical AI Services on Thursday, 5 November 2026, offering other robot makers the software and data tools it built for its own fleet. "Physical AI" is the industry term for artificial intelligence that controls machines moving in the real world, such as robots and self-driving vehicles. The company announced the service alongside its third-quarter results and named Bedrock Robotics, which builds self-driving systems for construction machines, as one of its first customers.
Key Facts
- Launch: Physical AI Services, announced by Serve Robotics on 5 November 2026.
- Serve Link: connects robots working in the field to AI models running in data centres.
- Serve Data Curator: picks out the most useful clips from fleet camera footage for training AI models.
- Data catalogue: a library of data on sidewalk terrain and street infrastructure gathered by Serve's robots, offered for training foundation models, the large general-purpose AI models that other systems are built on.
- First customers: Bedrock Robotics is among the first, according to reports.
- Pricing: not disclosed.
- Results: Serve reported third-quarter revenue of $2.8 million, up 313% from a year earlier.
- Fleet: the company says it operates more than 2,000 robots in over 40 cities; these are company figures.
Why It Matters
Building robots that can move safely through streets is hard and expensive. Companies need ways to connect thousands of machines to powerful computers, collect huge amounts of video, and choose which pieces of that video are worth using to train their AI. Serve built these tools for its own delivery robots. By renting them to others, it is trying to turn an internal cost into a new source of income, much as Amazon once turned its own computing systems into the cloud business Amazon Web Services, a comparison made in coverage of the launch.
The move also shows where the value in robotics may be shifting. The machines themselves can be copied, but real-world data, collected over millions of trips on actual pavements, is scarce. A company that owns such data and the tools to sort it can sell it to others who are building robots for different jobs, from construction to warehouses.
For Indian companies building robots and autonomous machines, the model is worth watching: shared tools and data could lower the cost of building such systems. But the service announced is for Serve's own markets, and the company has not said it will offer it in India.
Serve's revenue is still small. Its claims about fleet size and the value of its data come from the company, and the business case depends on enough robot makers paying for tools they could, in theory, build themselves.
Peepals calculation: A 313% rise to $2.8 million means revenue a year earlier was about $0.68 million (2.8 ÷ 4.13 ≈ 0.68), so the company earned roughly four times as much in this quarter as in the same quarter of 2025. On the company's own figures, its fleet averages about 50 robots per city (2,000 ÷ 40 = 50). Our calculation from the reported figures.
| Product | What it does |
|---|---|
| Serve Link | Connects field robots to AI models in data centres |
| Serve Data Curator | Selects useful training footage from fleet operations |
| Data catalogue | Sidewalk terrain and infrastructure data for training foundation models |
Impact
Short-term: Robot makers such as Bedrock Robotics can use Serve's tools instead of building their own, and Serve gains a second line of business beyond deliveries.
Long-term: If more companies buy such services, robotics could start to look like cloud computing, where a few firms supply the common tools and data that many others build on.
Who is affected: Robotics and autonomous-vehicle companies; Serve's investors; and, indirectly, people in cities where delivery and service robots operate.
Key Takeaway
Serve Robotics launched Physical AI Services on 5 November 2026, selling the connectivity, data-sorting and training-data tools behind its delivery robots to other robot makers, with Bedrock Robotics among the first customers.
Questions and Answers
What is Serve Robotics' Physical AI Services?
A new business launched on 5 November 2026 that offers other companies the tools Serve built for its delivery robots: Serve Link for connecting robots to AI models, Serve Data Curator for choosing training footage, and a catalogue of sidewalk and infrastructure data.
What does "physical AI" mean?
It refers to artificial intelligence that controls machines acting in the physical world, such as robots, drones and self-driving vehicles.
Who are the first customers?
Bedrock Robotics, which builds autonomy systems for construction equipment, is among the first customers, according to reports.
How big is Serve Robotics?
It reported third-quarter 2026 revenue of $2.8 million, up 313% from a year earlier, and says it runs more than 2,000 robots in over 40 cities.
Disclaimer: Prepared by the Peepals newsroom from publicly available sources with AI assistance. Information is accurate to the best of our knowledge at the time of publication and may change. Images may be representative. Not professional advice. Report an error via our contact page.
By Anika Sharma, AI & Innovations EditorCovers AI & Innovation and more for Peepals Tech. All stories by Anika Sharma
Block layoffs: second round hits Cash App, Square operations teams
ServiceNow layoffs: third 2026 round hits India, UK, US staff
Get the day's stories in your inbox
Peepals Weekly: the week decoded, free. One email a week, unsubscribe in one click.
Anika Sharma is a pen name of the Peepals Global Editorial Team, which researches, fact-checks and publishes every story. The portrait is an illustration.
