November 2013. I’m sitting in my office at Autodesk and working on Autodesk Labs projects – which I’d started several years earlier – with an iPad in front of me. I open a browser and launch Autodesk Inventor. Full Autodesk Inventor. Not some stripped-down mobile viewer, but actual Autodesk Inventor with all its features.
Here are Autodesk Inventor and Maya running in Firefox
Autodesk Inventor Running in FireFox on a Mac Video
Then I grab my MacBook and open Revit. Revit has never run on a Mac. It can’t. It’s Windows-only and always has been. But there it was, running perfectly in Mozilla Firefox browser.
I switch back to the iPad and open 3ds Max & Maya. Same thing -full application, full performance, running on hardware that should choke trying to open a web page, let alone render 3D scenes. I ran a rendering of an animation that was complex underwater Nemo’esque coral reef environment and, on my desktop, took 42 minutes per frame to render which takes days for a 30 second sequence, but in this cloud-based config each frame rendered in under 1 second and the entire 30 second sequence was done in under 10 minutes. My mind was blown! I thought it was an error and the file didn’t actually render, but it did.
Here are screen shots of Autodesk 3ds Max and Revit in my Firefox browser.
It all worked. Not “demo well enough to fool people” worked. Actually worked. Production – ready worked.
My old post from 2013: Running Autodesk Applications in My Web Browser
And then Autodesk decided not to ship it after it graduated from the Autodesk Labs.
The Setup
Four companies announced a partnership on November 5, 2013: Mozilla, OTOY, Amazon Web Services, and NVIDIA. The pitch was simple: run high-performance desktop applications in any browser, on any device, with no plugins.
The tech stack:
- NVIDIA’s new GRID GPUs for graphics horsepower in the cloud
- Amazon’s EC2 G2 instances to host the applications
- OTOY’s ORBX.js codec to stream the graphics
- Mozilla’s asm.js and WebGL to make it all work in browsers
And Autodesk provided the killer apps to prove it: AutoCAD, Inventor, Revit, 3ds Max, and Maya. All running from AWS servers, streaming to browsers.
Jeff Kowalski, our CTO at the time, put out a press release saying “It’s no longer a requirement to run sophisticated 3D design applications on a powerful workstation. Now all you need is a simple browser and an Internet connection.”
He wasn’t just saying it. We actually did it.
What We Built
Let me be clear about what we accomplished because people still don’t believe me when I tell them.
This wasn’t AutoCAD “lite” or some web viewer. This was the full desktop applications. The same executable that shipped on shipping products. Running in the cloud, streamed to browsers.
I ran Inventor on an iPad. Complex assemblies. Full feature set. It performed like it was running on a $5,000 workstation.
I ran Revit on a MacBook. Architects had been begging for Revit on Mac since Revit was created. We’d always said no because porting would be too expensive. Well, here it was. Running in Mozilla Firefox. Because the Mac wasn’t running anything – the Windows application was running on an AWS server and streaming to the browser.
Same with 3ds Max. Same with Maya. These are applications that normally require serious hardware—discrete GPUs, multiple cores, 16+ GB RAM. And they were running smoothly on devices that had no business running them.
Because none of the compute was local. It was all happening in Amazon’s data centers on NVIDIA GRID GPUs. Your device was just a screen. Imagine just swapping to another screen if you dropped your device, easy recovery just get to a browser and sign in and boom you were back up and running.
Why This Mattered
The business case was obvious. At least it was obvious to me.
No more hardware requirements. Companies wouldn’t need to buy expensive workstations. A $300 Chromebook could access the same computing power as a $5,000 workstation. Students could do professional work on whatever hardware they had.
IT costs disappear or greatly reduced. No more managing driver updates across hundreds of machines. No more “works on my machine” problems. Everyone’s running the same hardware configuration in the cloud. Known, tested, optimized.
True bring-your-own-device. Employees could use whatever they wanted. Mac users could finally run Windows-only apps. People could work from home on personal laptops without sacrificing performance.
Simplified development. Instead of supporting millions of hardware combinations – different GPUs, different drivers, different CPUs – we’d be developing for one known cloud configuration. Better performance, fewer bugs, faster development in theory.
Platform problems solved. Revit on Mac? Done. Inventor on Linux? Done. Professional tools on tablets? Done. The platform became irrelevant.
For customers: lower costs, instant access anywhere, no installation headaches, automatic updates, real mobility.
For Autodesk: predictable infrastructure, simplified support, optimization for known hardware instead of the combinatorial explosion of user configurations.
Better product. Better business model. Obvious win.
What Actually Happened
OTOY made their AMIs free. Over 500 AWS customers subscribed in the first few weeks. Thousands of hours of streams were running. The tech press covered it. Customers were excited.
And internally at Autodesk? Crickets.
The technology preview in Autodesk Labs just… faded. No product launch. No roadmap. No commitment. Just quietly stopped talking about it.
Why Autodesk Killed It
I speculate why. I watched it happen.
Problem 1: Nobody knew how to price it.
In 2013, Autodesk still sold perpetual licenses. You bought AutoCAD, you owned it. Well, sort of. You paid for upgrades and maintenance, but the core model was “buy once, own forever.”
How do you charge for software that runs in the cloud? By the hour? By the project? Per user? And if customers don’t need expensive workstations anymore, do they pay the same license fee? More because we’re handling infrastructure? Less because they’re renting instead of owning?
Nobody had answers. And nobody wanted to be the person who figured it out by accidentally cannibalizing our existing revenue.
Problem 2: Internal politics.
We had teams already working on “cloud CAD.” AutoCAD WS was the official web/mobile product. Those teams had been working for years on a simplified, mobile-first experience.
The OTOY thing made all that work obsolete. Why would anyone use AutoCAD WS – a deliberately limited web viewer- when they could run full AutoCAD in a browser?
The possible political answer: protect the existing cloud teams by making sure the full-app streaming never became a real product.
Problem 3: It worked too well.
If the tech had been janky, we could have positioned it as a supplementary tool. “Use this for quick reviews, but you’ll still need your desktop for real work.”
But it wasn’t janky. It was good. Really good. I could do serious design work on an iPad or basic level laptop. Complex models performed smoothly. This wasn’t a supplement to desktop workstations – it was a replacement.
And I’m sure that terrified the business side.
Problem 4: Timing.
This was 2013. Autodesk didn’t announce subscription – only until 2015. Didn’t make it mandatory until 2016’ish. They were still making millions/billions from perpetual licenses.
Launching a browser product that could replace desktop sales could have felt too risky. What if it cannibalized the core business without generating enough new revenue?
The safe play: add cloud features incrementally while protecting desktop revenue.
That’s exactly what they did.
What We Got Instead
AutoCAD Web. A limited browser experience good for viewing and light editing. Useful, but deliberately limited so it wouldn’t compete with desktop.
Mobile apps that are fine for viewing but can’t handle real work.
A subscription model that customers were excited about because we took away ownership without giving them anything genuinely better.
Meanwhile, the tech we demonstrated in 2013? NVIDIA GRID became standard for cloud gaming. Browser apps became the norm. WebAssembly made the performance even better. Remote work during COVID proved device- independent access actually matters.
And Autodesk slowly, cautiously added cloud features while competitors started plotting to eat their lunch.
The Pattern
This keeps happening at Autodesk.
John Walker built an SF Bay area online marketplace before eBay. Ted Nelson worked on web browsers before the web. Autodesk Chaos explored AI before machine learning (more on this later in the series and the tie in to ants according to John Walker in a discussion with me).
Autodesk had browser native professional apps in 2013. Years before the industry accepted it was possible.
The pattern: we see the future early. We build it. We prove it works. And then they don’t ship it because the business model isn’t ready.
What Could Have Been (IMHO)
If Autodesk had committed in 2013:
- Revit would basically run on Mac – it just would work.
- Hardware wouldn’t matter. Students and customers anywhere could access professional tools.
- IT departments wouldn’t spend weeks setting up workstation fleets and updating them.
- Development would be simpler for Autodesk. One configuration. One deployment. Instant updates.
- Instead, we got incremental cloud features and subscription model customers weren’t that excited about.
The Real Lesson
- Big companies can’t disrupt themselves. At least not planned.
- The desktop business was printing money. Why risk it on an unproven cloud experiment?
- Answer: because if you don’t, someone else will.
But that’s easy to say now. In 2013, the safe play was protecting existing revenue. And Autodesk took the safe play.
We had it working. We had the future.
Autodesk just wasn’t ready to ship it.
It was probably the largest missed opportunity for technology advancement for customers and Autodesk benefit in my 25 years at Autodesk (I left in 2023). I still think of it today as that missed opportunity. Even in 2026 it would still be further than we have now. I would love to deploy access for 500 users to a super powered design application by the cloud and NVIDIA and AWS from some basic hardware at their work desk.
