SaaS: Customer or Hostage?

What a 1.2 rating on Trustpilot and $55 billion market cap have in common

Mar 09, 2026

The views in this article are my own and do not represent those of any past, present, or future employer. If you are a future employer reading this, hi, I am usually much quieter.

In the middle of a conversation a few days ago, going through some company names, Autodesk ADSK 0.00%↑ floated up. I know this company not because I invest in it but because I was a user of it back in school, sooner or later they sat us in front of AutoCAD for a semester of technical drawing. We all had a more or less common opinion about it, not the best if you asked.

That was over fifteen years ago and I wanted to check whether anything had changed. Looking at the numbers seems solid, revenue is $7.2 billion now, free cash flow $2.4 billion. Shareholders must be happy… but what about customers? I checked Gartner Peer Insights, as I always do when looking into a SaaS name, which is one of the few review sites where verified buyers actually write reviews that matter. Autodesk gets a 4.3 there. Not too bad. Then I thought about Cloudflare, one that engineers love and I already wrote about that here, the kind of company people recommend at meetups without being asked. Cloudflare gets a 4.7 on Gartner and first place in its categories. The scores are close but the comments feel different. We live in an era of data aggregation and we often and I am the first guilty tend not to scratch more than the global number so I read many comments and Cloudflare users talk about choosing something. Autodesk users talk about depending on something.

“Your balls in a bag”

Between the years 2015 and 2017 Autodesk killed perpetual licenses for every major product: AutoCAD, Revit, Maya, Inventor, 3ds Max all gone. If you had paid a few thousand dollars for a copy and used it for ten years, that option was gone and you well… eh… now have to pay the subscription.

The author of the CAD Nauseam blog described it like this: “You’re asking me to give you my balls in a bag in the hope that you won’t squeeze too hard”. Everybody has a poet inside them.

Prices went up about 5.7% a year after the switch, and licenses where a team could share a pool of seats got eliminated and renewal discounts were stripped in January 2026, and three-year auto-renewals started getting pushed without an obvious way to opt out. So basically when your subscription dies the software stops working, cloud projects go view-only, and your files sit in .dwg and .rvt, proprietary formats that no competitor fully reads. Ten years of templates and custom building components and drafting standards, all locked inside a format controlled by the company you are trying to leave. It does not sound pleasant to be honest.

On Autodesk’s own forums, across different products and across years, paying customers keep arriving at the same word and I am going to quote verbatim since my Substack’s earnings does not cover a legal team and put the sources at the end of this article:

A Fusion 360 user in 2020: ”Your software is crappy and your holding my files hostage”.

On Practical Machinist in 2022: Their cloud system takes your files hostage and makes it very difficult to keep copies of your own data”

An Inventor user in July 2025: ”I fundamentally can’t, and will not accept a software licensing scheme that de-facto holds my work as a hostage to get leverage over me”

Nobody told these people what to write, they just found the word on their own because it matches what they experience when they try to stop paying. So what are their options? Keep paying.

Revit, the workhorse and jewel in the crown of Autodesk, coordinates buildings across structural, mechanical, electrical, and architectural disciplines in ways nothing else matches at commercial scale. Every consultant expects .rvt files, and students learn Revit in school on free licenses Autodesk hands out to every architecture and engineering program. I saw that firsthand but to be honest Microsoft, Oracle and Sun Microsystems did the same in my own school. Yes, Sun Microsystems. Yes I am that old but hey I ended up working once on Sun Solaris so call it destiny.

But going back to the point, switching from Revit means rebuilding a decade of templates and component libraries while live projects keep running, and for most firms that cost is measured in years.

One core

I keep pulling the thread and this is getting worse. Big 2026. The hottest word of the last 4 years probably is GPU. Companies are fighting over them, Nvidia surpassing Apple, Jensen printing trillions… and Revit runs mostly on a single CPU core. Yes you read it well and no it was not a typo. What that means in practice is that an architect buys a $3,000 workstation with a shiny 16 processor cores and a powerful graphics card, opens a large hospital model, switches a view, and Revit uses one core while the other fifteen sit there doing nothing and the graphics card barely wakes up.

Every other kind of heavy software moved past this a long time ago. Game engines and video editors and AI tools all spread their work across every core and GPU, and chip manufacturers responded by building processors with more cores rather than faster individual ones because the physics make it nearly impossible to push a single core much further. Hardware keeps going one direction and Revit is stuck going the other, needing one fast core that the industry has mostly stopped optimizing for because it does not make any sense.

Seventeen major firms including Zaha Hadid Architects asked Autodesk for a fundamental rewrite in 2020. Five years later there is nothing on any public roadmap or nothing that I could find. Autodesk is building a cloud platform called Forma that could move heavy computation to servers eventually, but it is early and nobody is designing the next shopping mall in it.

When the Hostages Escape

Unity U 0.00%↑ proved that stored resentment moves fast once there is somewhere to go.

A few years ago about 70% of top mobile games ran on Unity and developers used it and respected the engine. I tried it myself, it was a very cool thing to develop on. Then Unity imposed a retroactive per-install fee on games that had already shipped and within weeks Godot, a free open-source engine, went from side project to serious alternative. Publishers started screening Unity games out of funding and the stock dropped from $200 to $20.

Switching cost for game engines took between one and two years, which is painful but people got through it when game developers were angry enough. Autodesk’s switching cost is years because of the file formats and the accumulated templates and the consultant ecosystem built entirely around .rvt. So the same kind of blowup has never happened to Autodesk.

What about the future?

Autodesk has not faced a real competitive test since the subscription switch a few years ago because the lock-in makes it nearly impossible for firms to run one.

What happens when AI tools can route the electrical system of a building or test every possible site configuration or generate structural layouts in minutes? That work is already happening in startups today. Right now those tools plug into Revit as helpers, but each one absorbs a piece of the design hours that make expensive fees feel worth it. If the valuable thinking keeps migrating into AI plugins, at what point is the architect paying for a documentation shell?

Autodesk knows this is coming and they are building their own AI and it seems promising. At least what you can see on their website. They showed foundation models that generate 3D geometry from prompts, they are rolling out an AI assistant across their products. The question that I would ask if I had skin in the game is whether a company that has not managed to make its flagship product use more than one CPU core in twenty years can move fast enough in AI to stay ahead of startups that were built from scratch to do exactly this. What happens when Anthropic decides to enter that game? Maybe ask IBM IBM 0.00%↑ when Anthropic mentioned Cobol. Just saying.

Also I think underestimating open source can be a mistake. What would happen when open-source alternatives get good enough? Godot just proved in gaming that a free project can go from curiosity to real competitor in under two years once a wave of angry professionals shows up and starts contributing code. FreeCAD shipped 1.0 last year with twenty million downloads. If frustrated users start contributing to open-source BIM the way frustrated game developers contributed to Godot, how fast could that gap close? What about the AI leverage that now many people will have?

What happens if someone builds a tool that converts .rvt family libraries to an open format without destroying the parametric relationships? Remember that frontier models are very good in reverse engineering things and translating, so the thing that keeps switching costs measured in years is the accumulated template library. If that library becomes portable probably switching costs can be measured in months and not years.

Just an interesting fact, Singapore already requires open-format building submissions for all new projects by October 2026. If a bigger construction market follows, let’s say US, Japan or any big European country with actual enforcement, what happens to a moat that depends on nobody being able to read .rvt files without paying Autodesk?

And if every serious computing workload in 2026 runs on multiple cores and GPUs while Revit sits on one core with no public plan to change, what happens when AI agents need to run complex building simulations in parallel and Revit physically cannot keep up with the tools being built on top of it?

The stock trades at almost 30x EBITDA. That price assumes none of these questions get answered the wrong way.

So I answer myself my original question, AutoCAD was slow and in my opinion overpriced already back then and the interface was something everybody complained about. Fifteen years later the complaints are almost the same except now they include the word hostage and the subscription costs ten times what it used to but everybody still uses it.

For now.

This is not investment advice. I am not a financial advisor, I am not a stock analyst, I am a software engineer who used AutoCAD once in 2010 and apparently never got over it. The only advice I am qualified to give is to always keep a local copy of your files. And to back them up. Now go back up your files, you never know.

This post is public so feel free to share it.