Launch · AI & Deep Tech

Addverb launches humanoid robot for factories

By Startup Enthusiast · 13 February 2026

What they do

Addverb makes robots for warehouses and factories

It launched a 6-foot humanoid robot after 14 months of R&D

The robot targets solar, battery, electronics and hazardous industries

Addverb launched a humanoid robot after 14 months of R&D and training.
The robot is 6 feet tall, weighs 80 kg, and can carry 15 kg payload.
It currently uses wheels for legs; a biped version is planned later.
The humanoid is designed for unstructured environments like factories and warehouses.
It uses multiple cameras, LIDAR sensors, and a high-end compute engine for physical AI.
About 80% of the robot is built in-house at Addverb's Noida facility.
The rest is outsourced to local partners.

India's robotics and industrial automation market is estimated at ~$2 Bn and growing at 15-16% CAGR.
Global interest in humanoid robotics is driven by AI advances and labour shortages in manufacturing hubs.
Vision-language-action (VLA) models enable robots to perform complex tasks autonomously.
A key challenge is the lack of high-quality training data for general-purpose robots.

Addverb started with grid-based and natural navigation robots for warehouses and factories.
It also makes quadruped robots (robot dogs) called Digital Mule for surveillance, research, logistics, and defence.
The quadruped robots have a payload of 5-30 kg and are used in uneven terrains and mountains.
Addverb earlier made collaborative robots (cobots) — six-axis robotic arms for pick-and-place, painting, welding, screwing, and bolting.
The company built a data factory at its Noida campus to generate millions of training episodes via teleoperation, simulation, and synthetic data generation.
Teleoperation involves a human operator controlling a master robot; the slave robot replicates actions autonomously, creating paired datasets.
Addverb uses smart glasses with multiple cameras to capture egocentric data (visual inputs mimicking the human operator view).
Simulations are created using Nvidia Omniverse and Google DeepMind's MuJoCo for synthetic training data.
The company also leverages open datasets from academic collaborations and Hugging Face.

Addverb plans to build ~100 humanoid units over the next year.
It will collaborate with 5-6 selected customers for product maturation.
Initial deployment targets: solar, battery, electronics manufacturing sectors, and hazardous jobs in nuclear and chemical industries.
Clearer cost structures are expected after Addverb transitions from proof-of-concept to mass production.