A week before EOS is due to unveil a new polymer powder bed fusion platform at Formnext 2026, one of its best-known customers has already published benchmark numbers. According to VoxelMatters, Materialise measured 44% higher productivity on the unreleased machine than on the previous EOS platform. On a production part for GKN Aerospace, build time per part fell from 19 minutes 53 seconds to 13 minutes 51 seconds. EOS plans to show the system on November 17.
Formnext launches usually come with a spec sheet and some marketing language. Getting an early throughput figure from a service bureau, on a real customer part, is less common. That makes the claim worth a close look, both for what it shows and for what one benchmark can't tell you.
Reading the Number Correctly
The two headline figures describe the same result in different ways. "44% higher productivity" is a throughput number: more parts per hour. The per-part build time is the inverse. Converting to seconds, 19:53 is 1,193 seconds and 13:51 is 831. Divide one by the other and you get about 1.44, which matches the 44% throughput claim. Read as time saved, the same result is about a 30% cut in machine time per part: roughly six minutes off every cable twist wheel.
The difference matters when you set your expectations. A 44% throughput gain doesn't mean a build finishes in 56% of the time. It means a build finishes in about 70% of the time, and a machine running all the time turns out about 1.44 parts for every one part it used to make. For a service bureau billing machine hours, that higher output per hour is the figure that counts, so it's no surprise Materialise led with it.
The test part was a cable twist wheel made for GKN Aerospace. A real aerospace production part is a better benchmark than a showroom demo, but it is still one part. Powder bed fusion throughput depends heavily on how parts are nested in the build volume, how much area each layer has to scan, and how evenly that area is spread across the bed. A packed build of small parts and a build of a few tall, sparse parts can behave very differently on the same machine. The reported figure applies to this part, on this machine, at Materialise.
Where the Speed Comes From
The details reported so far point to three changes: more lasers, software that balances their workload, and a new recoater. Only the first is a raw hardware gain in exposure capacity.
Three lasers. The platform uses a three-laser architecture. Adding lasers is the obvious way to speed up powder bed fusion, because more lasers mean more exposure capacity per layer. But more lasers only help if each one has useful work to do.
Intelligent utilization and dynamic load balancing. The platform is described as using intelligent utilization with dynamic load balancing, which VoxelMatters says "maximizes output per build by equally balancing the exposure workload." The likely effect is that the controller shares scan work among the lasers so none of them sits idle while another finishes its part of the layer. On a multi-laser machine with fixed zones, a crowded region of the bed can hold up the whole layer while the other lasers wait. Moving work between lasers as each layer requires is how a three-laser system gets close to three lasers' worth of output, rather than being limited by its busiest zone. The exact algorithm hasn't been described in the material available so far.
An advanced roller recoater. The third change affects the whole workflow, not just the build. Materialise reports that the new roller recoater produces a softer powder cake, which makes depowdering easier. Post-processing labour is a real part of the cost per part, and a hard cake around delicate features costs time. A machine that builds 44% faster but still needs the same cleanup at the bench only solves half the problem. A softer cake works on the other half.
Open Parameters, Not Just Presets
The platform reportedly lets users run validated parameter sets or open parameters. That choice matters more than it might seem. Validated sets suit production users who want repeatable, qualified output and don't want to own the process development. Open parameters suit bureaus, research groups and advanced users who want to tune exposure, layer strategy or materials themselves. A company like Materialise, which runs many applications across many customers, has an obvious interest in that freedom. The fact that the machine offers both suggests EOS is aiming at serial production and process development at the same time.
Materialise's comments came from Giovanni Vleminckx, the company's process engineering manager, who spoke to VoxelMatters about the testing. He told VoxelMatters that what stood out was "not one isolated feature, but the overall performance of the system."
What It Means for Makers
Few FilamentFeed readers will buy an industrial polymer PBF system. Plenty will order sintered nylon parts from service bureaus, or design parts that end up made that way. For them, the economics run through companies like Materialise. If a bureau's machines turn out more parts per hour and need less labour to clean, that can eventually show up as lower prices, shorter lead times or more capacity for small, odd jobs that bureaus usually push to the back of the queue. Nothing guarantees it will, but that is the path by which a gain like this reaches makers.
Be cautious about the details. This is one benchmark, on one part, run by a major EOS customer on a machine that hasn't been publicly launched. As of this writing, no primary EOS or Materialise documentation was available to confirm the test setup, the build layout or which earlier platform was used as the baseline. Independent results across a range of geometries and nesting densities will be more useful than any one launch-week number.
The broader trend is clear, though. Industrial polymer PBF vendors are competing less on spot size and build volume and more on orchestration: keeping several lasers fully busy, reducing post-processing, and letting expert users take control of the process. Formnext 2026 will be crowded with launches across the industry, from large industrial systems to companies such as FELIX Printers showing their own lineups. November 17 should show how much of the reported 44% reflects the platform in general and how much reflects one well-chosen part.
Bottom Line
Materialise's figures, a 44% throughput gain and per-part time cut from 19:53 to 13:51, are internally consistent and come from a real production part. They point to a machine whose throughput gain rests on its three lasers and load balancing, with a softer powder cake that eases depowdering. Treat them as an informed early report rather than a spec, and wait for EOS's own launch details and wider testing before planning around them.
