Xaba, a startup building synthetic brains for industrial robots with zero code, announced it has secured a $6 million seed investment led by Hitachi Ventures.
The Toronto startup said the extension to its seed round will accelerate the deployment of AI-powered robotics and cognitive industrial control systems.
Hitachi Ventures led the round using funds from its new $400 million fund, with participation from Hazelview Ventures, BDC, Exposition Ventures, and Impact Venture Capital.
Xaba is pioneering the application of industrial artificial intelligence (AI) to remake manufacturing processes. Its flagship product, xCognition, empowers industrial robots and collaborative robots (cobots) with AI-driven cognition and awareness, enabling them to autonomously generate programs and execute complex tasks such as welding, drilling, assembling, and additive manufacturing.
By integrating real-time intelligence into automation, Xaba’s solutions significantly reduce deployment costs and enhance the quality, consistency, and flexibility of manufacturing operations. The company said it is taking aim at a $9 trillion opportunity. The shortage of skilled robotics programmers and control engineers creates even more challenges for companies to scale automation effectively.
Massimiliano Moruzzi, CEO of Xaba, said in an interview with GamesBeat that industrial automation remains highly inefficient, relying on outdated controllers, rigid programming, and extensive manual intervention. Programming and deploying industrial robots alone cost the industry $7 billion annually, with 80% of automation costs stemming from manually developing logic for industrial controllers.
“Our vision is to disrupt the giants. What we’re developing is what I call a synthetic brain for information, or cognitive control,” said Moruzzi.
Similar to what Open AI is doing for natural language commands for AI, Moruzzi said that Xaba is remaking factory language so that it can enable better automation, with the result being not only better robots for industrial purposes but also better human supervision and human support.
Xaba’s generative industrial AI equips machines with cognitive intelligence, allowing them to autonomously adapt, optimize, and execute tasks with precision. At its core is xCognition, which acts as a kind of “Open AI for Industrial Automation” — fully automating industrial robotics and machine programming for any task while automatically generating both part-programs and all the programmable logic controller (PLC) machine logic required to bring any machine to life. In essence, this is automation driven by self-programming robots that can easily transition from “text to action,” Moruzzi said.
“Traditional robotics systems require extensive programming, constant human supervision, and struggle with real-world variability, in geometry, process parameters, materials, and actual production KPIs,” said Moruzzi. “We’re redefining automation by enabling robots and machines to self-optimize and execute complex tasks with minimal programming. The result is a dramatic reduction in waste and up to a 10x reduction in costs.”
Meet Xaba: The autonomous AI control system for industrial automation
With Xaba, manufacturers can simply describe automation goals, production KPIs, or operational tasks in human-readable text or functional specifications. From there, xCognition and PLCfy autonomously generate the required code, enabling robots and production lines to operate independently with real-time adaptability.
Digital twins are meant to perfect designs of factories before they are built in the physical world. But Moruzzi said that the concept ought to be renamed “automated reality.” he said that industrial managers need to have a machine that can synthesize the experience of a human and transfer that to a robot.
“At Xaba, we are developing foundational AI for automation, which means that, for example, my synthetic brain captures the physical, electromechanical model of the machine. Why am I doing this? Because experience is not inside to the encyclopedia that Open AI is accessing in order to transform that text into the actual. What I’m doing is technology called data ontology. Data ontology is going to be the next big wave in AI in order to transform weak AI to strong AI.
What is data ontology?
He said data ontology is his own segment within neuroscience data.
“It has the capacity to do what at the moment only the human can do, which is called abduction. Abduction means that the brain is capable to formulating scenarios. To augment the task that you were about to do, that you learned, or to do a new task based on the experience that you have assimilated before,” he said. “I’m leveraging legacy data from the factory now. What I did for the last few years is capturing knowledge from legacy that comes from operator machines. We’re using this to do the same disruption that Open AI did for creating an email or summarizing a book or, in my case, automating something.”
The result is faster deployment, minimized downtime, and smarter, more resilient automation across industries through:
Physics-Informed Machine Learning Model: Acting as a true digital twin, it accurately replicates real-world environments, adapting to different machines and motion platforms for precise, real-time optimization.
Robotics & PLC AI Code Generation: Proprietary AI models autonomously generate both robotic programs and PLC code by understanding operational workflows and machine logic. This reduces deployment time by up to 80% and eliminates manual coding.
Real-Time Process Learning Module: Powered by Data Ontology and Graph Neural Networks (GNNs), this module captures, maps, and understands complex relationships between machines, sensors, and processes. It ensures dynamic adaptation and continuous optimization.
Cognitive Control Framework: A universal AI platform that integrates seamlessly with any robotic system, CNC machine, or PLC controller, supporting both legacy and modern equipment.
“Industry 4.0 promised intelligent, autonomous factories—but it’s often been stuck in pilot purgatory, held back by rigid, code-heavy systems and legacy infrastructure,” said Gayathri Radhakrishnan, a partner at Hitachi Ventures, in a statement. “Xaba breaks that deadlock. By giving industrial machines the ability to self-learn and self-program through generative AI, Xaba is turning the vision of smart manufacturing into a scalable, reality today.”
Xaba’s AI is already transforming aerospace, automotive, and high-precision manufacturing by eliminating costly rework and manual adjustments in areas such as automotive manufacturing.
Xaba’s AI optimizes aluminum casting and forging, enabling robots to precisely machine metallic castings while adjusting for machining tolerances—dramatically reducing assembly costs, rework, and production time.
It’s also doing large-scale robotic drilling. Manufacturers have achieved 10 times faster production rates while significantly lowering capital expenditures, Moruzzi said. Unlike traditional systems requiring rigid programming and manual adjustments, Xaba’s AI allows robots to seamlessly reconfigure different parts and processes without costly downtime.
Xaba is also doing robotic welding. Xaba’s AI automates MIG (Metal Inert Gas) welding and TIG (Tungsten Inert Gas) laser welding, and laser welding, ensuring consistent, high-quality output across production lines while accelerating production timelines.
And Xaba is handling large-scale 3D printing. Xaba’s xTrude system optimizes Fused Deposition Modeling (FDM), preventing delamination, collapse, and distortion. It allows manufacturers to fine-tune print parameters in real-time, improving reliability and reducing material waste.
“Max and his team have created a bold new blueprint for the future of robotics and industrial automation,” said Marco Andriano, CEO Fives Cinetic Corp, in a statement. “Together, with Xaba’s xCognition, we’re delivering intelligent systems that transform decades of inefficiency into agile, scalable manufacturing environments—finally solving the most persistent programming and production challenges the industry faces today.”
The company has 24 people working for it. The team includes AI scientists, mathematicians and mechatronics experts. They’re running an AI applied automation lab.
Moruzzi believes that large language models (LLMs) are not the right technology for automation because the foundational model behind the LLM is based fundamentally on a set of weight factors. It’s like using an encyclopedia to answer a question about whether a robot should turn left or right. On top of that, LLMs are prone to hallucinations, which is bad in industrial settings.
That means scale that may or may not synthesize the semantics in the way that gives you the right response, he said. Moruzzi designed his AI to be completely different, building a system that can create synthetic data on its own.
“Your brain is not an LLM,” he said.
Moruzzi said his company is entering production in the coming months. He noted there are only about 4.4 million industrial robots in the field now. That’s virtually nothing, considering how many humans there are. And the reason, he said, is that their brains — little more than industrial controllers — aren’t good enough. They’re like “empty boxes,” he said.
“That’s why I’m building a cognitive brain,” he said. “That is the way to talk with the physical world.”