An AI that writes CNC programs just raised $20m, with Lockheed Martin's venture arm in the round
CloudNC raised $20 million on 8 September 2026, and Lockheed Martin's venture arm put money in. The London company sells AI that writes CNC machining strategy — the decisions a programmer makes before any metal is cut — and the round takes its lifetime total to $128 million.
It is a Series B extension, not a Series C. CEO Theo Saville told TechCrunch the last major raise was four years ago. Four years between rounds, and the new money arrives as an extension of the old one. That is a specific shape, and it is worth reading before the shape the press release wants you to see.
Who is in the round
Nimble Ventures led. Calculus Venture Capital, Entrepreneurs First and LM Capital — Lockheed Martin's venture fund — are the named participants. Saville founded CloudNC in 2015 with Chris Emery, now chief science officer. The company has 80 employees, says more than 1,000 machine shops use its software, and puts 80% of its customer base in the United States.
The Lockheed name is the one that will get quoted. A defence prime's venture arm backing software that programs machining is a signal about where the aerospace supply chain thinks its bottleneck is, and defence work is exactly the kind that combines low volumes, hard materials and long programming times. No contract was announced. No Lockheed shop is named as a user. The investment is the fact; anything past it is inference.
What the software actually does
CAM Assist plugs into an existing CAM package rather than replacing it. Saville's description to TechCrunch: it "selects suitable tools, approach directions, and cutting feeds and speeds, and then drafts the code required to tell the CNC machine what to do." The programmer then reviews, edits and approves.
The framing matters more than the mechanism. CloudNC's argument is that a machine shop's constraint is not spindle hours, it is the hours before the spindle starts — how the part is held, which tools are picked, what order the operations run in. That work has always lived in one experienced person's head, and there are fewer of those people every year.
“Traditional CAM systems are powerful, but in most workflows, they are still tools that allow programmers to manually specify how a part should be machined.” — Theo Saville, CloudNC, to TechCrunch, 8 September 2026
On CloudNC's own site the product lists integrations with Autodesk Fusion, Mastercam, Siemens NX, GibbsCAM, SolidCAM and Creo, and states support for 3-axis and 3+2 axis machines. That last line is the real specification. Simultaneous five-axis is not on it.
The numbers the coverage carries, and the ones it does not
Here is everything with a figure attached, and where each one comes from.
| Item | Figure | Stated by |
|---|---|---|
| Round size | $20m, Series B extension | TechCrunch, 8 Sep 2026 |
| Lifetime raised | $128m | TechCrunch |
| Gap since last major raise | 4 years | Saville, to TechCrunch |
| Founded | 2015 | TechCrunch |
| Employees | 80 | TechCrunch |
| Machine shops using CAM Assist | “more than 1,000” | Saville |
| Share of customers in the US | 80% | TechCrunch |
| CAM packages integrated | 6 | cloudnc.com |
| Axis coverage | 3-axis and 3+2 axis | cloudnc.com |
| Programming time saved | “up to 80%” (vendor marketing) | cloudnc.com; same figure three years ago |
| Quote Agent launch | “next month” | Saville, 8 Sep 2026 |
| Valuation | not disclosed | — |
| Revenue or ARR | not disclosed | — |
| Cycle-time or scrap improvement | not stated anywhere | — |
| Named customer | none in the coverage | — |
The bottom four rows are the story for anyone deciding whether to trial this. A funding announcement is not a benchmark. The one performance figure in circulation — "up to 80%" — is the company's own, it describes programming time and not cycle time, and it has sat on CloudNC's material since the demonstration video the company published three years ago. A marketing claim that has not moved in three years is a marketing claim, not a measurement.
Quote Agent is the more interesting product
The second product got two sentences in the coverage and deserves more. Quote Agent estimates the cost, risk and delivery time of an incoming job so a shop can accept or decline quickly. Saville put it plainly: "Machine shops need to quote faster, program faster, and deliver more with the people and machines they already have." He said it launches next month. CloudNC's own navigation already carries a "New: Quote Agent" label, which is a little ahead of that.
Quoting is where a machine shop actually loses money, and it loses it twice. It loses it by quoting slowly, so the enquiry goes elsewhere. It loses it by quoting wrong, and finding out eleven weeks later on a part it has already cut. Software can plausibly fix the first. The second is harder than it looks, because a quote is not only arithmetic on cycle time and material. It is a judgement about whether this particular customer's drawing is finished, whether the tolerance stack is achievable on the machine that is free, and whether the delivery date survives the one operation that goes outside for heat treatment.
What it changes for a buyer of machined parts
If your supplier adopts this, the visible change is response time. A quote that took three days comes back in an afternoon. That is worth something, and it is not free.
The thing to watch is what happens to the estimate underneath. An automatically generated first-pass strategy is a first pass. Somebody still signs it off, and the value of the signature depends entirely on whether the shop treats the AI output as a draft to be corrected or as a number to be sent. Ask the question directly: who reviews the generated strategy, and does the quote go out before or after that review. A shop with a good answer has made itself faster. A shop without one has made itself faster at being wrong.
The 3+2 limit is the second thing to check. If your parts need simultaneous five-axis work — impellers, blisks, complex aerospace geometry — this tooling does not currently cover them, and any promised turnaround improvement applies to the rest of the job list, not to yours.
Vesprr quotes machining work, so the honest position is that we have skin in this. Faster quoting is good. A quote is still a commitment, and a commitment generated in minutes carries exactly the same liability as one that took three days. Talk to us about how your quotes are produced before you judge them on how quickly they arrive.
Sources
- Dominic-Madori Davis, CloudNC raises $20M to automate manufacturing's most pressing bottlenecks, TechCrunch, 8 September 2026 — the round, the investors, the company figures and every Saville quotation above.
- CloudNC lands $20m to expand its AI software for precision manufacturing, Tech.eu, 9 September 2026.
- cloudnc.com — CAM integrations, 3-axis and 3+2 axis coverage, the Cutting Parameters engine, and the “up to 80%” claim.
- CAM Assist Demo, CloudNC — the company's own demonstration and the earliest appearance of the 80% figure we found.
- Increase #CNC Programming Productivity With Artificial Intelligence | AI Webinar CloudNC CAM Assist, MLC CAD Systems — reseller-run webinar on the workflow.
- Introducing CAM Assist 2.0 for Mastercam, CloudNC — the current release for Mastercam 2025 and 2026.