Something big is happening in desktop CNC. I don’t think most people see the problem hiding inside it.
I work on the CAM workbench in FreeCAD. A lot of people have been sending me Facebook and Kickstarter links to a new wave of desktop milling machines. I’ve spent years thinking about how toolpaths get made and what they cost. So when I look at these machines, I don’t see what the Kickstarter videos show. I see a gap. This post is about that gap.
The trend
For more than a decade, desktop CNC meant three axes. A spindle moves in X, Y, and Z. It cuts what it can reach from above. Fine machines. Real limits. If you want to reach the other sides of the part, you need to rotate the part or the spindle.
Now watch the axis count climb.
Four-axis is going mainstream right now. The two biggest desktop CNC campaigns in Kickstarter history closed within weeks of each other in late 2025. The Makera Z1 raised $10.2 million. It’s four-axis ready and starts at $899. The Nestworks C500 raised $12.2 million with an optional simultaneous fourth axis. Three-axis machines still dominate the installed base, but four-axis is where the growth is.
Five-axis is the wave behind it. Pocket NC proved a desktop five-axis machine could exist back in 2015. The trade press called it “5-Axis 4 Grand” because the idea was absurd until it wasn’t. Then, in a single twelve-month window from mid-2025 to mid-2026, four desktop machines claiming true simultaneous five-axis launched or pre-launched: the InfiMaker K1, the IFMAKER M5, the Xhorse WM-100, and the TOPFAB TF500. Every one of them is priced between $5,200 and $8,600.
To understand how strange those numbers are, you need the old numbers. Entry into five-axis machining used to start around $150,000 and climb fast. When Haas released the UMC-500 at roughly $166,000, the industry considered it disruptively cheap. That was 2017. Nine years later, companies are promising the same capability, albeit at a much smaller scale, for the price of a nice laptop and a vacation.
Tormach deserves a mention here. Years ago they bet there were customers who wanted machines far cheaper than commercial iron but far better than hobby routers. They were right, and they built a real business in the $10,000 to $40,000 range. Part of how they hit those prices was building PathPilot on open-source machine control. Keep that thought. We’ll come back to it.
One honest caveat before we go on. The five-axis wave is, right now, a wave of campaigns. Almost none of these machines have shipped. Pocket NC is still the only desktop five-axis with a delivery track record. The trend is real. The machines are still mostly vapor.
The promise
The marketing for all of these machines makes the same promise: this will be as easy as 3D printing.
That’s not my paraphrase. “The first desktop CNC that’s as easy as 3D printing” was the literal tagline on the original Carvera campaign in 2021. The 2026 machines escalate it. One-click CAM. Toolpaths from a text prompt. “No CNC background needed.”
Each machine ships with its own “studio” software. The demo videos are genuinely impressive. Smooth five-axis motion. Beautiful surfaced parts. The target customer is an artist or a maker, not an engineer, and the message is clear: press start and walk away.
I want that future to be real. But there’s a chain of dependencies the videos skip over.
The catch
A machine can only cut what its toolpaths tell it to cut. Toolpaths come from CAM software. So the machine’s utility is limited by the CAM you can afford.
The bundled studio software is where this gets uncomfortable. Nobody outside these companies has verified what the applications can actually do, because the machines haven’t shipped. Look closely at what the demo videos actually show. It’s almost always 3D surfacing. Take a mesh, wave a ball endmill over every point of it, and a figurine emerges. Surfacing is beautiful, and it’s also the easiest toolpath to generate automatically. One generic operation. That’s why the “easy as 3D printing” pitch works — for that one kind of part.
Functional parts are a different problem. A bracket has flat mating faces, bores at dimension, tapped holes, and a tolerance stack. Machining it means choosing the right operations, ordering them correctly, selecting the right tools, and sequencing the setups. No text prompt gets you there. This is the work CAM software actually exists to do, and it’s the work the demo videos never show.
The vendors themselves hint at the ceiling. InfiMaker’s own product page tells experienced users they can import G-code from Fusion 360, NX, or PowerMill.
Read that again. The alternative to the likely-limited bundled software is professional CAM.
So what does professional CAM cost?
The economics
In the old world, this question didn’t matter much. If you just spent $250,000 on a five-axis machining center, then dropping another $1,000 a year for CAM was almost a rounding error. Less than half a percent of the machine. Shops still complained about how expensive CAM software is but really nobody blinked at the total cost of ownership.
The new world breaks that math.
The cheapest credible path to simultaneous five-axis toolpaths today is Autodesk Fusion plus its Manufacturing Extension: $680 a year for the base subscription, plus $1,465 a year for the extension. That’s $2,145 per year, every year. On a $5,000 to $10,000 machine, you’re paying 21% to 43% of the machine’s purchase price annually, forever, just to generate the toolpaths the machine was sold on.
And that’s the budget option. A professional seat of Mastercam with the Multiaxis add-on, or hyperMILL, or ESPRIT, runs $20,000 to $50,000, plus maintenance around 15% a year. Two to five times the cost of the machine itself.
Something tells me that the customers who are seriously considering a CNC at this price point are going to choke on those numbers.
There are two kinds of multi-axis work. Indexed machining, often called 3+2, rotates the part to a position, locks it, and then cuts like a normal three-axis job. Simultaneous machining moves all five axes at once while cutting. Indexed is genuinely cheap now. Base Fusion includes it. DeskProto sells a hobby license for about 250 euros. But simultaneous is the line the entire CAM industry draws its top price tier at. Every vendor gates it. And simultaneous is exactly what these Kickstarter videos show.
The trend isn’t reassuring. It’s moving the wrong direction. Fusion used to include simultaneous multi-axis in the base subscription. In 2022, Autodesk removed it and put it behind the paywall. No CAM vendor has moved down-market to meet these machines. Not one.
How this plays out if nobody pushes back
The machine vendors know all this. Their answer will probably follow a template that already exists.
CAM engines can be licensed wholesale. ModuleWorks, a company most machinists have never heard of, supplies the five-axis toolpath engine behind a majority of commercial CAM products, and they offer white-label packages to machine builders. Tormach shows the retail version of the model: they sell an OEM edition of SprutCAM with real simultaneous five-axis for $3,775, and it only posts code to Tormach machines. Expect the studio software on these new machines to be some mix of licensed engines and homegrown AI, wrapped in a friendly interface.
That can work. But buyers should go in with eyes open, because this model carries two risks.
The first is lock-in. Machine-locked postprocessors. Proprietary project formats. One vendor, IFMAKER, openly markets that they “eliminate post-processor headaches” by building everything in-house. That means eliminating portability. Your parts, your setups, and your workflow live inside one company’s software.
The second is the cloud. AI-driven CAM implies heavy computation, and heavy computation usually means their servers, not your computer. Your designs travel through their infrastructure. Your machine’s brain runs on a startup’s runway. These are Kickstarter companies. What happens to cloud CAM in year three if the company doesn’t make it?
If you’re considering one of these machines, here are the questions I would ask before backing:
- Can I use third-party CAM with this machine?
- Do you publish postprocessors? For which CAM packages?
- Does toolpath generation run on my computer or on your servers?
- When I outgrow the bundled software, what does simultaneous five-axis CAM cost me per year?
- If your company shuts down, does my machine still work?
If a vendor can’t answer these, that tells you something.
The missing piece: a FOSS solution
There is no open-source simultaneous five-axis toolpath generator. Not in FreeCAD. Not anywhere.
Control is mostly solved – LinuxCNC has executed simultaneous five-axis motion with tool center point control for years – but what’s missing is the software that decides where to move. Toolpath generation is the gap.
FreeCAD is climbing the same ladder as the machines. The upcoming 26.3 release ships our first multi-axis capability: indexed work and rotary surfacing, plus a machine library and machine-aware postprocessing. Simultaneous five-axis is the open frontier. I know, because it’s my frontier. This is the part of FreeCAD I work on.
Which brings me back to Tormach. They lowered their costs and earned trust by adopting open-source control instead of building a proprietary one. They should be applauded. The same logic now applies one layer up, at CAM. And CAM is the layer that determines what these machines can actually cut.
So here is my message to the teams behind these launches. Your machines’ ceiling is CAM. None of you can fund a Mastercam competitor alone. But a shared, open CAM engine makes every machine in this category more valuable, including yours. The buy-in is cheap. Publish your postprocessors. Publish your machine definitions. Put hardware in developers’ hands. Show up in the FreeCAD forum. You have a vested interest in this work, and right now, none of you are anywhere to be seen.
And if you’re a buyer: ask the five questions. Favor the vendors with open ecosystems. Makera has hired a community member who was writing postprocessors and community firmware for their machines. She wrote the FreeCAD postprocessor and verified it with community support. The Makera community, with no help from Makera, has expanded the firmware to include variable support, o-codes, and soon cutter compensation. Buyers will do the work when vendors simply allow it. Imagine what happens when vendors help.
The machines got cheap. Now the toolpaths have to.
Prices and campaign figures are as of August 2026. Machine prices are early-bird/launch prices from the linked campaigns; CAM prices are from the linked vendor pages and, where vendors publish nothing, from the linked Practical Machinist threads.