Dear CAD vendors: Cloud Will Delight Your Customers. Building It Yourself Won't.

The business models of NVIDIA and Intel are vastly different. Understanding this difference is crucial for vendors of CAD and BIM software who want to offer their software as a service. In this article I'm going to tap into 15 years of experience in building cloud solutions for 3D engineering software. My goal: help software vendors become the NVIDIA, not the Intel.

Personal Experience

Over those 15 years, I built a cloud business for CAD and BIM applications twice: more than a decade ago, a private cloud at a VMware/EMC/Atos joint venture. And since late 2023 at Designair. I also held an executive role at a major CAD vendor and watched them try building cloud. And I've won customers from another major vendor's cloud offering, giving me unique insights as well.

The 10 Factors

Before I tell you what happened, let me explain why combining CAD and Cloud is so hard. And i can't think of a better way to do this than contrasting NVIDIA and Intel. NVIDIA designs chips and outsources manufacturing. They keep a 75% gross margin and a $3 trillion market cap. Intel designs and manufactures chips as well, but with vastly different results. Their gross margin is 33% and falling. In 2024, they split the company to show investors the truth: Intel Products (the design side) is profitable. Intel Foundry (the manufacturing side) lost $13 billion.

Building a cloud offering means building a managed service to underpin the CAD software. Managed services run on 8-12% net profit margins - and 30% of providers don't make money consistently. That's before you add GPUs. A managed service that turns software into SaaS brings the horrible economics of the foundry to a software company - even when it's automated well, and running on a hyperscaler.

Here's why:

#1: Margin Dilution. Software has near-100% margin per unit sold. Managed services runs on 8-12% net profit. Mix the two, and watch your margins collapse. Think about what Intel's margin would have been without that $13 billion loss from manufacturing.

#2: Opportunity Cost. NVIDIA's $3 trillion market cap isn't just from 75% margins - it's because the market is betting everything on AI, and NVIDIA sells the hardware. Every dollar an ISV spends on cloud infrastructure is a dollar not spent on AI capabilities. Building cloud is a distraction from what's obviously more promising.

#3: Operating Model Mismatch. Designing chips - or software - means getting it right once. Manufacturing chips - or running cloud infrastructure - means getting it wrong zero times. Software development is agile: sprints, iterations, 9-to-5. Managed services is ITIL: process, control, zero tolerance for error, 24/7 operations. Different culture. Different mindset. Different capabilities.

#4: Single-Vendor Limitation. A vendor will only offer cloud for their own software. But engineers don't use one tool - they run 3D modeling, clash detection, 2D drawings, and Excel simultaneously. A single-vendor cloud will never support the full workflow.

#5: Version Flexibility. Single-vendor clouds typically offer only the latest version to stay close to "true SaaS". But plant engineering projects run for years and freeze on a specific software version at kickoff. Upgrading mid-project - even if contractually allowed - introduces unacceptable risk. A cloud offering with only the latest release is useless for these customers.

#6: Scale Economics. NVIDIA has massive scale and still doesn't manufacture. If anyone in CAD/BIM had the scale to make this work, it would be Autodesk - the 800-pound gorilla. Yet none of their flagship desktop CAD or BIM authoring tools are available as true SaaS - or "360" in Autodesk speak. Autodesk looked at cloud in the early teens. Even with their scale they couldn't make the economics work. Any smaller single CAD vendor faces a very hard, and much lower ceiling.

#7: Compliance Burden. ISO 27001, SOC 2, NIS2 - these are the security certifications most enterprise customers demand. They're about how infrastructure is run, how data is handled, how incidents are managed. Completely different expertise from software development. Different audits, different ongoing commitments. Building cloud versus getting certified to run it for regulated industries is a whole different ball game.

#8: Channel Conflict. Most CAD/BIM vendors generate over 50% of revenue through VARs and implementation partners. A cloud offering changes the go-to-market motion, the buying journey, the pricing expectations. Done well, the cost of sale drops significantly - but it kills the channel. And once partners are gone, there's no going back.

#9: Fear of Vendor Lock-in. Cloud from the software vendor creates double dependency: customers are locked into both the software and the infrastructure. Switching CAD systems already means retraining teams and migrating projects. Adding infrastructure lock-in makes the decision even harder. Customers know this, and it makes them hesitate.

#10: SLA Liability Exposure. Software licensing is "as-is." Cloud means uptime commitments, incident response SLAs, breach notification obligations. A software bug means a support ticket. Cloud going down means customers lose money and expect compensation. Most software vendors aren't structured for this liability - it's a fundamentally different risk profile.

Here's to those who tried

PTC announced Creo+ - CAD running in the Cloud - in May 2023 at LiveWorx with CEO's Jim Heppelmann's full support and backed by hyperscaler partnerships. Almost three years later, PTC's own chatbot confirms: "Creo+ primarily runs on your own computer."

This isn't failure - it's how hard this is, even when you do everything right, and you have the talented PTC team working hard to achieve this.

AVEVA did launch a cloud solution for E3D. Here's what we learned from AVEVA customers who switched to Designair. To avoid what I suspect is margin dilution, they priced it so high that customers balked - solving one problem by creating another. Their operating model mismatch is visible everywhere: backups that should be swift are a 72-hour struggle, the traditional software development and cloud departments are disconnected, and upgrades arrive with little upfront notice and catch everyone by surprise. And they offer only the latest version - completely against industry best practice for plant engineering projects that freeze on a specific version at kickoff.

What ISVs Should Do Instead

NVIDIA's brilliance isn't just focus - it's knowing what to outsource. They design the chips the world runs on, then hand manufacturing to TSMC. TSMC operates the fabs, manages the yields, runs 24/7 operations. NVIDIA keeps the margins, TSMC takes the operational complexity.

Smart ISVs should do the same. Design great software. Let someone else run the cloud.

That's what we do at Designair. We're the TSMC for engineering software - purpose-built cloud workstations for CAD and BIM applications. The ISV focuses on what they're great at. We handle the GPUs, the uptime, the support tickets at 2am.

Want to talk?

If you're an ISV considering building your own cloud, let's grab 30 minutes and a cup of coffee. I'm happy to spar - and explore what a partnership with Designair could look like.

You can book me via this link.