CloudNC raises $20M extension to expand CAM Assist, an AI tool that drafts CNC programs for human review

A programmer opens a part file and, before any metal moves, has to settle how the thing gets clamped and which tools go into the carousel ^[9]. Chief executive Theo Saville's argument is that traditional CAM software is where that decision gets typed up rather than where it gets made: powerful, he says, but still a toolkit that lets a programmer manually specify how a part should be machined ^[8]. CAM Assist plugs into those systems, examples being Autodesk Fusion and Mastercam, picks tools, approach directions and cutting feeds and speeds, and drafts the machine code ^[5][6].

Then the programmer reviews, edits and approves it ^[7]. Saville is explicit that the human gate stays, describing the product as an expert assistant that automates the first-pass thinking and repetitive setup ^[19], with the aim of making skilled staff more productive rather than taking judgment out of the loop ^[20]. That framing is honest, and it also pins the entire value of the tool to one quantity: how much of the first draft survives review.

The announcement does not contain that quantity. It contains more than 1,000 machine shops using CAM Assist and 80% of the customer base sitting in the US ^[10][11], which is a breadth number rather than a depth one. Nothing in it addresses retention among those shops, or the share of drafted strategies that reach a spindle untouched. With 80 employees ^[12], CloudNC is carrying roughly 12 shops per head ^[2], so scaling adoption ^[22] is a go-to-market problem in its telling, and the raise is sized accordingly: $20m as a Series B extension, the first major round in four years, on top of the $108m the company had raised before it ^[1][3][1].

The more interesting line item is the roadmap. Quote Agent, due next month, scores the estimated cost and risk of a new job so a shop can accept or reject it faster ^[15], and it is funded by the same money that is meant to widen CAM Assist take-up ^[14]. Saville puts both needs in one sentence, saying shops have to quote faster, program faster and deliver more with the people and machines they already have ^[17], against US reshoring and a shortage of skilled workers ^[16]. A shop losing work because its quotes go out late does not win that work back by programming faster. Building the quoting product says CloudNC has heard that from customers.

Two axes decide whether either tool pays in a given shop. First, which resource is actually the queue: programmer hours, or spindle hours. Second, how repetitive the part mix is, because a shop cutting families of similar parts already has templated strategies and buys a smaller delta from an automated first pass. Programmer-constrained shops doing genuine one-offs are the cell where a drafted strategy converts directly into throughput. Spindle-constrained shops get a faster program and the same weekly output. The retrospective version of that test needs no sales call: of the last twenty programs the shop wrote, how many were strategic decisions and how many were setup typing.