SaaS CAD Suffers With Other Victims of AI
Opinion by Ralph Grabowski
Ryan Vlastelica over at Bloomberg reports that financial analysts are the most bearish on Adobe since 2013 [source]. Some analysts have downgraded the stock as the price of ADBE shares has fallen 20% since the start of the year (at time of writing), and now stands at less than half its peak price. Not just Adobe.
Adobe’s share price falling since January 1
(Bearish means people are reluctant to buy a stock, which then pushes down its price due to lowered demand. Downgrade means the analysts think the price will not being going up soon-ish. Going up is one way we make money from our mutual funds. The other way is from dividends paid out by corporations.)
The reason is the threat from AI -- and not the AI that we keep hearing about, the one that’s apparently super-charging businesses, according to CEOs. It’s the other AI, the one that’s creating slop. AI creates slop in areas like text, art, videos, music, and software programing.
Slop is badly made output. Slop is acceptable for lazy people wanting quickly-written essays, fast illustrations for articles, any kind of background music to videos.
And so the stock market thinks that Adobe is in danger. People looking for customized illustrations no longer need to learn how to use PhotoShop or Illustrator. Instead, people ask AI to make something for them, without needing to buy and learn software that previously did the job for them.
SaaS Got the Sads
Not just Adobe. Nils Jacobsen talks about the falling share prices of SaaS firms [source]:
SAP down 17% so far this year
Microsoft down 17%
Workday down 22%
Oracle down 24%
Salesforce down 25%
ServiceNow down 28%
DocuSign down 30%
Many software firms changed their revenue model over the last decade from permanent licenses to SaaS [software as a service], meaning we pay non-stop to use the software. We stop paying, the software stops working, holding us at ransom. Revenue perfected! from the vendor point of view.
But now Wall Street trading desks are using the template that ‘AI eats software’. If the AI threat becomes reality (I doubt it will), then Jacobsen muses that “If generative image and design models deliver in seconds what teams of graphic designers used to do, how many Photoshop licenses does a company need from Adobe then?”
“If the business model actually shifts from Per-Seat to Per-Outcome (meaning payment based on results, rather than number of users),” adds Jacobsen, “it could change the entire revenue logic of the industry.”
Right now, the Per-Outcome payment model for requests from LLM-based AI software is free. Adobe’s monthly fee for Photoshop is $23; Autodesk charges $260 monthly for AutoCAD.
Is CAD in Danger?
Were the future of Adobe as diminished as a print newspaper, might the same affliction impact CAD software firms? I say not, due to this difference.
Graphics produced by PhotoShop/Illustrator need only be “good enough,” suitable for whatever pleases the eye. If the trees in the background are off by a few inches, no big deal. Hence, slop is fine.
CAD, on the other hand, is all about precision. Close enough is not good enough. A milling machine cannot receive instructions to place holes wrong by a few inches. Here, slop can kill.
Programming is also about precision. Slop is not acceptable in code, as Microsoft appears to be finding out [source].
Online program adding snow effect to 3D house (image source Autorender.ai)
CAD firms big and small are playing around with AI, but much of it isn’t for doing design. It makes suggestions, like finding a command name and generating good-enough renderings. Even there, I am skeptical that it’s AI working in the background, and not just a bunch of if-then programming subroutines.
Nevertheless, share prices also are lower for American SaaS CAD vendors, such as PTC (down 12% on the year) and Autodesk (down 17% ). European ones are not as badly affected: Dassault (down 2%) and Siemens (up 6%).
I think that share prices of SaaS CAD firms are being hit by a reluctance to buy because analysts are unable to distinguish between the accuracy required in the CAD/CAM world, and the lack of it elsewhere.
LLM-based AI is a bubble. This I firmly believe. OpenAI won’t be able to go public and won’t find any other way to pay off its monster debt, given its non-existent profits. Once this industry crashes NFT-style, then SaaS firms will be back.



LLM-based AI is effective and probably here to stay until an even better technology stands on its shoulders. Having spent a life in CAD (Autodesk & MCS) and more recently AR/VR (Apple), I can tell you that I can do more with an AI assistant in a shorter time than I could with a small team of mid-level programmers. That's not saying that I'm happy about the reduced demand for young engineers, it worries me, but it's is real and here to stay.
Nemetschek stocks are down by 41% in the last 1-year span.