CloudNC just raised $20 million to do the boring, expensive part of manufacturing faster

CloudNC has picked up a $20 million Series B extension, four years after its last major raise, bringing total funding to $128 million. The U.K. startup is still working on the same problem it picked nearly a decade ago: taking some of the slowest, most judgment-heavy work in CNC machining and turning it into software.

That matters because CNC shops don’t spend most of their time cutting metal. They spend a lot of time deciding how to cut it. Tool choice, approach direction, feeds and speeds, workholding, order of operations. Each of those decisions affects cycle time, scrap risk, and whether a job makes sense at all.

CloudNC says its CAM Assist product can handle a large chunk of that first-pass planning inside existing CAM systems like Autodesk Fusion and Mastercam. An operator still reviews the output. The software drafts toolpaths, suggests tools, and generates the code needed to run the machine. For shops short on experienced programmers, that’s the kind of automation that actually matters.

The bottleneck is before the spindle turns

People outside machining often assume the hard part is the machine itself. It isn’t.

A CNC machine only becomes useful after someone translates a part design into a machining strategy. That’s where the expertise lives. A good machinist knows the same part can be made several ways, but only some of those ways make sense for a given machine, material, tolerance target, and production run.

Traditional CAM software helps, but it still expects a human to make most of the decisions. CloudNC is trying to compress that planning loop. CAM Assist looks at the part and proposes a path an experienced programmer might choose, rather than just helping the user click through familiar operations.

That matters. A lot of manufacturing software mainly cuts clicks. CloudNC is trying to cut down the amount of expert reasoning needed for a first pass. That’s a harder problem, and it’s why the company has spent years on it.

Saville says more than 1,000 machine shops worldwide now use CAM Assist. That’s a decent signal for a tool in a conservative industry, especially one that plugs into workflows people already know. Manufacturers hate rip-and-replace software. If your product sits between Autodesk or Mastercam and a shop floor, it has to earn trust quickly.

Why this kind of AI is different from the usual pitch

CAM Assist sits in a more interesting category than the generic “AI for manufacturing” line that shows up everywhere.

It’s not trying to predict equipment failure from sensor noise. It’s not a chat interface bolted onto a factory dashboard. It’s a constrained optimization problem wrapped around an old industrial process. That makes it more practical, and more unforgiving.

In machining, bad suggestions are expensive. Pick the wrong cutter or the wrong cutting speed and you can waste time, burn through tools, or ruin the part. There’s no room for a model that sounds confident and guesses wrong. The software has to work within hard physical limits, actual tolerances, and a long tail of edge cases.

That’s probably why CloudNC describes CAM Assist as an assistant rather than an autonomous system. The user reviews and approves the output. That human-in-the-loop step isn’t a weakness. It’s the product. Shops will accept automation when it shortens setup time and keeps judgment in the loop. They won’t accept black-box code that spits out bad toolpaths and walks away.

There is a trade-off. The more the product depends on expert review, the less it standardizes work across a shop. But that can still be worth it if the goal is throughput. Most shops don’t need a robot machinist. They need fewer delays before the machines start running.

Where CloudNC is headed next

CloudNC isn’t stopping at toolpath generation. The company says a new product, Quote Agent, is set to launch next month. It targets another painful part of shop operations: estimating cost and risk fast enough to decide whether to take a job.

That’s a sensible extension. Quoting is where manufacturing software starts to touch revenue directly. If a shop can estimate setup time, machining complexity, and rework risk more quickly, it can respond to customers faster and avoid wasting time on bad-fit work. For small and midsize shops, that can mean the difference between a profitable schedule and a full calendar of low-margin jobs.

It also shows where CloudNC sees the bottleneck: not just machining, but the front end of manufacturing decisions. One product helps generate the plan. The next helps decide whether the plan is worth executing.

That’s a more believable strategy than trying to automate the whole factory at once. Quoting and CAM programming are adjacent, repetitive, and expensive enough to justify software. They’re also the kind of work where experienced staff spend a lot of time on pattern recognition and arithmetic that machines can support pretty well.

The funding says something about the market

The new round was led by Nimble Ventures, with participation from Calculus Venture Capital, Entrepreneurs First, and LM Capital, Lockheed Martin’s venture arm.

That mix is telling. This isn’t just startup money chasing a vague industrial AI story. Defense and advanced manufacturing have both been under pressure to do more with less, and both care about domestic supply chains. Saville pointed to reshoring trends in the U.S. and the shortage of skilled manufacturing labor. Those aren’t new problems, but they’re harder to ignore now.

Machine shops can buy more equipment. They can’t instantly hire more senior CAM programmers.

That shortage is one reason software like this gets attention. A strong programmer can squeeze more throughput out of the same machines, and that matters when capital equipment is expensive and lead times are long. If CAM Assist can cut programming time in a meaningful way, it turns human expertise into something closer to shared infrastructure.

The catch is adoption. Manufacturing software sells slowly for good reasons. Shops care about reliability, edge cases, and integration more than polished demos. A model that works on clean examples but falls apart on awkward geometry or weird fixturing will get ignored fast. CloudNC’s job isn’t proving that AI can generate plausible machining strategies. It’s proving those strategies hold up across the mess of real production work.

What developers and technical teams should notice

There’s a lesson here for anyone building software that touches physical systems.

The highest-value automation usually starts with the ugly, repetitive part of expert work, not the flashy part. In CNC, that’s setup and first-pass programming. In other domains, it’s triage, quoting, planning, or review. The pattern is the same: find the subproblem where a human is doing expensive reasoning over structured constraints, then put software beside them.

But the implementation has to respect the domain.

In machining, that means integrating with established CAM tools instead of replacing them. It means producing machine-readable outputs that fit existing workflows. It means keeping the human approval step. And it means being explainable enough that an operator can see why a toolpath was suggested.

That last part matters more than vendors usually admit. If the software can’t show its work, shops won’t trust it when a part gets expensive or the geometry gets weird. Nobody wants to debug a bad tool collision by reading model output.

It also means the bar for product quality is higher than in software-only automation. A bad recommendation here has physical consequences. Broken tooling, wasted material, machine downtime. Those costs show up immediately.

CloudNC still has a hard climb ahead

The company says it has spent years on the problem, including running its own factory. That’s believable, and probably necessary. Industrial software usually improves when the team has real process data and can watch failure modes in the wild. Synthetic demos don’t teach you much about how machinists work under pressure.

Still, the goal is ambitious. There are many ways to machine a part, and the search space is huge. CloudNC is trying to collapse that complexity into something useful enough to save time without pretending the physics are optional. That’s a real product challenge, not a branding exercise.

If CAM Assist and Quote Agent keep working as advertised, CloudNC could become one of the few manufacturing software companies that chips away at labor and throughput instead of just adding another dashboard.

The harder part now is scale. The shops are out there. The question is how quickly CloudNC can turn a useful assistant into a standard step in the workflow.