Cognex has agreed to acquire RealSense for about $500 million in cash, adding 3D depth-sensing and robotic perception technology to its industrial machine-vision portfolio. The transaction is expected to close in the fourth quarter of 2026, subject to customary closing conditions.

Cognex plans to fund the acquisition using cash and investments already on its balance sheet.

RealSense was established within Intel in 2014 and became an independent company in 2025. The business is expected to generate between $80 million and $90 million in revenue this year, according to Cognex.

Before the transaction closes, RealSense will separate its facial authentication product line into an independent company. Cognex will acquire the remaining depth-camera and robotic-perception business.

From fixed inspection to robotic perception

Cognex said its existing machine-vision systems are used in controlled industrial applications to identify and inspect products, take precise measurements, and guide fixed-arm robots. RealSense extends those capabilities into applications where robots use depth information to recognise their surroundings, judge distance, avoid obstacles, and navigate changing environments.

“We enable machines to identify, inspect, measure and guide while RealSense enables them to perceive and navigate their surroundings,” Cognex CEO Matt Moschner said.

The Association for Advancing Automation describes adaptive robotics as combining machine vision, sensors, AI, and real-time computing so robots can respond to changing conditions without extensive reprogramming. Conventional industrial robots typically operate in structured environments where movements are programmed and parts arrive at expected positions.

Cognex said its outlook for the robotic-perception market is based largely on vision-guided fixed-arm robots and autonomous mobile robots (AMRs), where adoption is more established. The company considers quadrupeds and humanoid robots longer-term opportunities rather than necessary to its core growth expectations for the acquisition.

RealSense’s industrial technology includes cameras that generate depth information to help machines determine the position and distance of objects in their surroundings. Its systems support applications including robot guidance, part recognition, measurement, tracking, and safety monitoring.

The company’s D555 depth camera connects to industrial Ethernet networks using Power over Ethernet (PoE), allowing power and data to travel through the same connection. The camera also includes onboard vision processing rather than operating solely as an image sensor connected to a separate computing system.

Its RealSense Vision SoC V5 combines stereo disparity processing and motion estimation with a vision digital signal processor and image signal processor. The D555 also supports native on-camera people detection.

RealSense says image processing and inference can take place on cameras based on its V5 architecture. The D555 can also publish data directly into ROS 2 networks over Ethernet, allowing depth information to feed into robotics software without first passing through a USB-connected host.

The D585 also uses a V5 system-on-chip with a depth engine, image signal processor, vision DSP, and quad-core Arm processor. It offers depth output of up to 90 frames per second, while RealSense lists on-camera person detection as a beta feature at launch.

This allows depth calculation and some vision workloads to be handled on the camera before information is passed into the wider robot software and control system.

Existing RealSense deployments provide examples of how that architecture is being used on production floors. French robotics company Inbolt mounts RealSense D435 depth cameras on industrial robots as part of a system that continuously determines the location of components while a robot is operating.

According to a RealSense case study, the Inbolt system can adjust when bins move or conveyors remain in motion instead of relying only on a previously determined component position. RealSense reports perception-to-motion latency below 80 milliseconds, with the technology deployed across more than 70 factories.

According to RealSense, the system continuously re-localises components while the robot is operating, allowing its motion to be updated when the position of a part changes. This differs from applications where a robot executes a predefined movement based on a fixed or previously measured position.

“RealSense is one of the few 3D cameras that delivers both the point cloud quality we require,” Inbolt COO Albane Dersy said in the case study.

According to a RealSense case study, Italian industrial automation company GEFIT has integrated RealSense depth cameras into its vision platform. The system uses depth data to provide robots with information about objects and people in their workspace for applications including bin picking, human detection, and movement in shared areas.

Bringing more vision processing to the factory edge

Cognex already processes machine-vision workloads on embedded hardware at the edge. Its In-Sight 3900, introduced in May this year, combines image capture, AI processing, and inspection software in a system that can operate without an external PC.

The system uses a dedicated AI processor for local inspection workloads and supports Ethernet connections to PLCs and robots. Cognex said its processing architecture enables deterministic, real-time inspection on production lines.

RealSense’s V5-based cameras perform a different set of workloads on-device, including depth processing and other vision functions used by robotic systems. Cognex described RealSense’s wider technology platform as spanning imaging, custom silicon, 3D vision algorithms, embedded software, and developer tools.

According to Cognex, RealSense has shipped more than one million units. Its software development kit has recorded more than 1.2 million downloads and has approximately 20,000 active developers across four major platforms.

The SDK provides tools for integrating RealSense depth data into robotics and computer-vision applications, including support for ROS and ROS 2.

Cognex estimates the robotic-perception market at about $600 million and expects it to grow by more than 25% annually to approximately $1.6 billion by 2030. The estimates come from Cognex and are not presented as independent market forecasts.

The company expects RealSense to generate between $80 million and $90 million in revenue in 2026. Cognex said the business has approximately 160 employees and more than 70 patents and patent applications.

Cognex has also discussed combining the two companies’ software and processing technologies after the acquisition closes. During its acquisition call, management said newer RealSense devices have enough onboard processing capacity to run vision functions directly on the device.

Moschner said possible integrations include running Cognex-trained AI models and vision tools on RealSense hardware, as well as applying RealSense’s 3D depth-sensing technology to Cognex’s existing machine-vision applications. Cognex has not announced these integrations as commercial products.

See also: LG brings factory AI closer to production data with EXAONE

Want to learn more about the IoT from industry leaders? Check out IoT Tech Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including AI & Big Data Expo and the Cyber Security Expo. Click here for more information.

IoT News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.