For years, the machine vision industry has measured progress in megapixels. Every major trade show brought another sensor with a higher resolution, another camera boasting faster frame rates, or another manufacturer claiming the sharpest images on the market. It was an easy story to tell because bigger numbers usually translated into better performance.
But something has changed.
The latest announcements coming from camera manufacturers suggest the industry’s focus is shifting. Resolution is still important, of course, but increasingly it isn’t the camera sensor that’s limiting system performance, it’s the amount of data that has to move around the inspection system once the image has been captured. That’s why we’re suddenly hearing far more about 25GigE, 50GigE and now 100GigE Ethernet cameras. On paper it looks like another specification war. In reality, it marks a significant change in how machine vision systems are being designed.
The question is no longer who can build the fastest camera. It’s who can move huge volumes of image data quickly enough to satisfy the demands of AI-powered inspection.
The driving force behind all of this isn’t ethernet, it’s the rapid evolution of image sensors themselves. Sony’s Pregius S family continues to increase both resolution and speed, while manufacturers such as Gpixel and onsemi are pushing their own sensor technologies into applications that generate extraordinary amounts of data. Every new generation of sensor places more pressure on the rest of the imaging chain.
That’s becoming particularly apparent in industries such as semiconductor manufacturing, battery production and electronics inspection, where multiple high-resolution cameras may be inspecting products simultaneously. In those environments, bandwidth has become just as valuable as image quality. Capturing the image is only half the challenge; processing it in real time is where the real engineering begins.
Several manufacturers have recognised this shift and are beginning to position themselves accordingly.
Allied Vision has attracted considerable attention with its expansion into 100GigE cameras, signalling that ultra-high-bandwidth imaging is moving beyond specialist scientific applications and into mainstream industrial automation. Rather than treating 100GigE as a niche offering, the company appears to be building it into a broader product strategy, suggesting it sees genuine market demand emerging over the next few years.
Interestingly, Allied Vision isn’t entering an empty field. Emergent Vision Technologies has been advocating high-bandwidth Ethernet for some time, building much of its reputation around 25GigE, 50GigE and 100GigE imaging solutions. In many respects, Emergent has been ahead of the market, waiting for the rest of the industry, and perhaps more importantly, end users, to catch up.
Baumer has chosen a slightly different path. While it also embraces faster networking and the latest GigE Vision standards, its recent developments around Canon RF lens integration demonstrate that bandwidth alone isn’t the whole story. Modern optics, electronic lens control and flexible camera platforms all contribute to creating a more capable imaging system. It’s a reminder that innovation rarely happens in isolation; advances in sensors, optics, networking and software are all moving forward together.
Elsewhere, the competitive picture becomes much more nuanced. Basler continues to focus on building a comprehensive ecosystem around its cameras, placing significant emphasis on software integration and ease of deployment. Teledyne DALSA remains a dominant force in high-end industrial inspection, particularly where CoaXPress continues to offer advantages in deterministic performance. IDS Imaging is investing heavily in embedded vision and AI, while companies including Lucid Vision Labs, JAI, Cognex, Keyence and Omron continue to develop solutions that reflect the needs of their own customer bases rather than chasing headline interface speeds.
That diversity of strategy is healthy. It also reflects the reality that not every inspection problem requires 100GigE. A simple barcode reading application has very different requirements from an eight-camera battery inspection system analysing gigabytes of image data every second. There has never been a single technology that suits every application, and that isn’t about to change.
Which raises an obvious question. If 100GigE is so compelling, why isn’t everyone adopting it?
The answer is that the camera is only one component of a much larger system. High-bandwidth imaging requires compatible network infrastructure, specialist network interface cards, high-performance GPUs and software capable of making efficient use of technologies such as RDMA. Integrators also need confidence that these systems can be deployed reliably on the factory floor, not just demonstrated in a laboratory.
That is precisely why the arrival of GigE Vision 3.0 feels so significant. By introducing support for Remote Direct Memory Access (RDMA), the standard allows image data to move far more efficiently through the system, reducing processor overhead and helping GPUs receive image data with minimal latency. It isn’t the kind of announcement that grabs headlines in the same way as a new camera launch, but it may prove to be one of the most important developments of the year.
Looking ahead, the direction of travel seems fairly clear. Over the next two or three years we’ll almost certainly see more manufacturers introducing 25GigE and 100GigE products as sensor technology continues to advance. AI will demand ever more data, GPUs will become an increasingly common part of machine vision architectures, and the traditional boundaries between cameras, computing platforms and software developers will continue to blur.
Our industry has always enjoyed a specification race. We compare megapixels, frame rates, dynamic range and, now, ethernet bandwidth. There’s nothing wrong with that, innovation is what keeps machine vision moving forward.
But history also tells us that customers rarely buy specifications. They buy solutions that improve productivity, reduce costs and make manufacturing more efficient. The companies that ultimately come out on top won’t simply be those producing the fastest cameras. They’ll be the ones that make these next-generation technologies practical to deploy, easy to integrate and reliable enough for everyday industrial use. Gigabit Ethernet once seemed excessive for many applications. Today it’s almost taken for granted. It wouldn’t be surprising if, in a few years’ time, we look back at today’s discussion around 100GigE in exactly the same way.
The race has certainly begun.
The real competition, however, isn’t about bandwidth. It’s about making the next generation of machine vision simpler, faster and more accessible than ever before.
