The €1 million metal Ender 3
The actual productivity of metal 3D printing systems is an uncomfortable truth that nobody in the industry wants to talk about
I read a great article over the weekend. Sam Davies from TCT Magazine decided to answer a simple question: are industrial 3D printing machines too expensive?
At first glance, it’s a rather trivial question. Of course they are. Show me someone who thinks paying €1 million for a machine is cheap.
But that’s not really the point. The article raises a much more interesting issue - one that aligns perfectly with something I’ve been patiently arguing in my own writing for quite some time.
Two of Davies’ interviewees: Joe Calmese from ADDMAN and Christina Perla from Makelab, shared their thoughts on machine productivity.
Because that’s really the biggest obstacle holding back metal AM adoption: Productivity. Repeatability. Reliability.
Poor reliability in metal AM is one of the industry’s dirty little secrets. People either whisper about it or don’t talk about it at all. Everybody knows. Nobody wants to say it out loud in front of a prospective customer who’s just taken out their wallet.
If I’m being honest, metal AM today is roughly where low-cost desktop FFF was between 2018 and 2022.
The Ender 3 era.
Well, that’s quite expensive for what you actually get.
Fortunately, the solution is already visible on the horizon. And it will come from exactly the same place it did before.
Software.
The article
Sam Davies set himself an ambitious goal: to break down, line by line, exactly what makes up the price of an industrial AM machine. About half of the CEOs he approached politely declined to answer - which, in itself, is an answer.
One of them, however, agreed. Avi Reichental.
As some regular readers may have noticed, I’m not exactly what you’d call a fan of Mr. Reichental. But credit where credit is due - he was remarkably open. Maybe that’s because he’s reached a point in his career where he simply doesn’t care anymore. I don’t know. Either way, this time I genuinely appreciate his transparency.
According to Reichental, roughly 20-25% of the price reflects amortized R&D, 30-35% covers components and manufacturing, and another 15-20% goes toward sales and distribution. The rest is spent on material development, hidden costs, and profit, which, in reality, quickly disappears into investors’ pockets.
His conclusion is refreshingly honest: the price of the machine reflects “absorbing risk and complexity.” At today’s production volumes, nobody can significantly lower costs because there simply isn’t an application base large enough to generate economies of scale.
Others are singing the same gospel.
Rich Garrity from Stratasys and Sam O’Leary from Nikon SLM both argue that what really matters is the cost per part over the machine’s entire lifecycle. Cutting costs isn’t an option because reliability would suffer. In their view, an AM machine costs roughly what you’d expect to pay for a high-end CNC machine tool.
And you know what? They’re right.
The argument is logical, measurable, and difficult to dispute.
However, Filomeno Martina, former CEO of WAAM3D, probably summarized it best:
The industry needs a better economic language, not just better machines.
And yeah, it could have ended there as another fairly standard industry article...
…but Davies expanded his list of interviewees to include actual users.
And that’s where things get really interesting!
Joe Calmese from ADDMAN - a company operating a serious fleet of metal AM systems, states it bluntly:
For us, hardware cost isn’t the real issue.
Christina Perla from Makelab adds an observation that should send chills down every machine manufacturer’s spine. Every company she visits starts by talking about machine problems - not machine prices.
It’s much more about effectiveness and throughput.
Jacob Rindler from The Exploration Company is even less diplomatic:
The machines are too expensive - and I’ll keep saying it until my last breath.
Machine vendors want to discuss purchase price. The people who actually run these machines want to discuss downtime.
To illustrate the point, Davies quotes this brilliant analogy (I quote it in full because it’s simply too good):
In the 1980s, UK sitcom The Blackadder was cancelled after its first series, largely because there was a greater focus on production value than actual comedy. As Ben Elton, the co-writer of the show’s subsequent three series, jokes in his 2026 memoir, it looked a million dollars, but it cost two.
The same logic applies here.
If a $1 million machine only delivers 50% utilization, it might as well cost $2 million, because the customer will ultimately need a second one just to achieve the required throughput.
Sixty-five percent
ADDMAN guarantees a maximum machine utilization rate of 65%. They often exceed that number, but above it, they make no promises.
Calmese says their metal AM systems experience roughly 30% more downtime than CNC machines, and describes CNC machines as bricks: they sit there, they run continuously, and they only require scheduled maintenance.
Well, yeah... Sixty-five percent.
A machine costing €1-2 million, sold to aerospace or defense manufacturers, is only confident about performing as intended for roughly two-thirds of the time.
The remaining third is spent doing something every owner of a cheap desktop printer from a decade ago remembers all too well:
Breaking down.
Being repaired.
Producing failed builds.
This leads to a conclusion Davies never explicitly states, but one that follows directly from the numbers.
Today’s €1 million metal AM machine has roughly the same reliability profile as a $200 desktop printer from 2018.
It’s a metal Ender 3.
The only difference is that an Ender 3 cost about as much as a decent bicycle.
For AM veterans, that’s hardly surprising. For a new customer who’s just signed a seven-figure leasing agreement, it’s a punch straight to the solar plexus.
And the purchase price is only the beginning. As Davies writes:
The point-of-purchase price is not just a challenge for smaller operations, but also for larger corporations just getting started, and even experienced users looking to unlock a new application. Even when the machine advancements do pertain to the business case – such as the productivity improvement a larger laser configuration can provide – that only serves to help existing applications. Because of the cost hike a machine with multiple lasers will demand, the business case for a new application often doesn’t make sense.
In plain English:
Even when manufacturers genuinely improve their machines - for example by adding more lasers to increase productivity, it only benefits applications that already exist.
For new applications, the business case still falls apart because multi-laser systems become prohibitively expensive.
The machine gets bigger. The investment threshold rises with it. The cycle repeats itself.
Machine utilization - not purchase price - is what ultimately drives metal AM adoption. Cheaper hardware doesn’t move the needle much.
Calmese estimates that reducing machine price by 10% lowers part cost by only around 3%. Increasing utilization from 65% to 85%, on the other hand, would fundamentally change the economics of the entire industry.
The real question is: where will that improvement come from?
Well… There’s no shocking plot twist here.
The answer, once again, is software.
Hardware alone is not enough
Let’s go back to 2022–2023, to what completely transformed low-cost desktop FFF.
That revolution wasn’t driven by hardware. It was driven by software. Firmware. Automatic calibration. Slicers capable of selecting parameters automatically. Integrated ecosystems of validated print profiles and ready-to-print models.
Suddenly, a $500 printer became repeatable and predictable. A cheap machine became something you could actually trust.
That’s exactly the transformation metal AM has yet to experience. Because all the fundamental 3D printing technologies have already been invented. Hardware can be copied.
What’s left is process preparation, validation, process monitoring and post-processing.
And every one of those areas is fundamentally a software problem.
Yes folks, hardware alone is no longer enough.
The manufacturing process stops being a collection of settings stored inside an engineer’s head and becomes a formal, programmable description. Something that can be version-controlled, compared, validated, and automated.
That’s what moves reliability from 65% toward 85% without adding a single extra laser.
It’s the software-defined manufacturing, if you like catchy names.
Companies like Dyndrite are pushing metal AM in exactly that direction. And that’s where my cautious optimism comes from.
Real progress in metal AM is happening. Quietly. Away from the spotlight.
It’s happening inside the code that simply makes machines work.
Metal AM is still waiting for its own 2022 moment. From where I stand, it’s no longer a question of „if”, only „when”.