TL;DR: Build an AI-powered cloud CAD tool with push-button aluminum CNC fabrication - like Cursor for Onshape, with SendCutSend-style monetization. Lots of hard technical problems but potentially a very large, very sticky business if you can pull it off.

Today I talked to someone who is a manufacturing manager at Apple, and was at Tesla when Elon was pushing for maximally-automated manufacturing. He was interested in building a fully-automated aluminum CNC facility, similar to what Hadrian is doing. The way he described things to me seemed plausible, and he's much more knowledgable about the specific challenges than I am.

I've been workshopping my AI-powered CAD idea. Most people realize that there's a big opportunity for AI-powered CAD software, but there are two big weaknesses with the software-first business model:

1. High switching costs
2. Small TAM

Why software-only CAD is a tough business

The first point is clear to anyone who has spent significant time in SolidWorks, Fusion 360, or Onshape. Once a team has a decade of part libraries, assemblies, plugs, scripts, drawings, and workflows built around a specific CAD platform—not to mention internal verification processes—switching is extremely painful. Even if a new tool is 10x better for ideation, interoperability becomes the bottleneck.

The most plausible GTM for a new CAD tool has always been bottom-up: win students, hobbyists, and early-stage startups, then slowly expand upward. This is how Onshape grew, basically. But the enterprise market is structurally resistant to change, since there isn't usually some acute moment where an engineering team will say that they urgently need a new CAD tool. So vendors compete on custom enterprise integrations, PDM/PLM compliance, and long-tail usability features, which increases R&D cost for potential competitors and narrows the number of buyers. There are a lot of parallels with how Cursor approached the coding market which might be interesting to draw from but unfortunately there is no VSCode equivalent in CAD land, and any disruptor would probably just need to rebuild something themselves.

The second weakness is TAM. Dassault Systèmes, the parent company of SolidWorks and CATIA, reported $6.72B in 2024 revenue across software, subscriptions, and services. That’s materially smaller than Adobe’s $21.51B in the same period. A Figma-style winner could take a large share of that amount, but still ends up competing for a pie that’s not especially large compared to other creative-software sectors.

Vertical integration changes the economics

If you add manufacturing to the stack, everything changes.

Right now, getting machined aluminum for small batch production is a frustrating experience for most companies who aren't Tesla. Lead times are unpredictable, quoting is slow and opaque, file formats and tolerances often get mishandled, and the price/quality tradeoff is painful. SendCutSend is the de-facto choice for hobbyists and prosumers for sheet metal, which can get you pretty far for a lot of random projects, and they’re exploring aluminum CNC, but the quoting workflow still remains pretty manual (as deduced after experiencing several major manufacturing errors when using SendCutSend). The broader ecosystem of OSH Cut, Fictiv, Protolabs, Xometry and others span the spectrum of marketplaces with varying levels of in-house production, but so far I am not aware of a manufacturer that tightly integrate design, simulation, manufacturability analysis, instant quote, and automated machining - probably because it is a really hard technical thing to do.

It would be cool if someone did this though. The dream here is essentially to automate mechanical prototyping - basically, a single mechanical engineer should be able to churn out a very complex design in a day or two, and have the part at their door within a week. This would be pretty revolutionary. But besides that, a vertically integrated AI CAD + CNC facility could solve several structural business problems at once.

Switching cost invert

Instead of asking companies to switch CAD tools, you give them the fastest and cheapest way to get CNC parts. The manufacturing side becomes the wedge. Once they’re started designing within your service, it becomes progressively easier to add more features like AI automation, DFM feedback, and eventually fully-automated workflows.

TAM expansion

Manufacturing is vastly larger than CAD. CNC machining alone is tens of billions globally. Even capturing a small portion of low-volume CNC spend dwarfs typical SaaS subscription revenue. The CAD tool becomes a customer acquisition and retention mechanism for a high-revenue downstream service. It's much easier to picture what revenue growth in manufacturing looks like compared to revenue growth with CAD subscriptions, particularly in a post-AI world.

Structural defensibility

If your quoting, toolpathing, fixturing, and machining systems are fully automated, your operating costs scale sublinearly compared to traditional job shops. Combining that with AI-assisted design creates a workflow moat that’s extremely hard to replicate without owning the whole stack.

Fat monetization layer

Manufacturing margins, especially with fully automated CAM and machining, offer a far larger monetization surface than SaaS licenses. This mirrors SendCutSend’s business: relatively simple software for design upload, with revenue driven by fabrication.

Competitors

My prior post lists some companies which are building AI-powered CAD tools. I'm currently not aware of anyone who is looking to combine this with some vertically-integrated manufacturing play.

Some companies which currently provide mail-order custom parts (not including Chinese companies like JLCPCB and PCBWay):

Conclusion

Personally, I really like this idea. It seems like a fun, impactful thing to build, and the revenue and growth potential seems pretty big.