Joe Schmidt, partner at Andreessen Horowitz, came on Bricks & Bytes to defend the firm’s recent thesis that AEC design software has been frozen since 1997. We pressed him on Revit’s grip, the three attack vectors the thesis identifies, what founders should build, and what construction operators should do right now.
The a16z thesis dropped in March and the AEC group chats lit up. The headline (yes, Joe knows buildings existed before 1997) put a crosshair on Autodesk’s Revit and the $150 billion MEP services market sitting on top of it. Some readers loved it. Others called it Silicon Valley tourism. We got Joe on the show to defend the work.
Joe has family in the trades going back 100 years, including a grandfather who invented the concrete pump. He is not a tourist. But he is also not pretending construction will roll over because a16z published a Substack.
Why AEC software has been stuck since 1997
The Revit moat is technical, not just political
It is easy to mock Autodesk for charging more every year while shipping less. The reason no one has displaced Revit is harder than that. The problem Revit solves is genuinely difficult. You are reasoning about 3D geometry, physical constraints, building code, room occupancy, and equipment specs all at once, while a dozen other firms make changes in parallel. Every challenger for two decades has stalled trying to reach feature parity.
The lock-in is baked into the workforce. Almost every architect and engineer learns Revit in school, and firms have spent years building custom families inside Autodesk’s proprietary file formats. The a16z team estimates Revit is doing around $3 billion in ARR on about 95% market share. Autodesk acquired it in 2002 for $133 million, which Joe called one of the greatest vertical software acquisitions ever made. And yet the platform has barely evolved. MEP was bolted on in 2007 and has done very little since. Joe compared collaboration today to finance professionals emailing Excel files back and forth.
It is to AEC what Excel is to finance: the connective tissue between every firm, every workflow, and every deliverable.From the a16z thesis
Joe Schmidt’s a16z thesis, in plain English
Three attack vectors and one preferred bet
Strip out the venture jargon and Joe’s framework comes down to three ways to attack the AEC stack. One: go head to head with Revit and try to replace it. Two: build new software around Revit and own the workflows it handles badly. Three: go after the services budget itself, the human work that has never been touched by software.
If forced to pick one today, Joe would pick the services angle. The biggest MEP firms have already broken their work into discrete, measurable deliverables and offshored a lot of it. AI does not need to change how firms think about it, just who is doing it. The why-now is two things: LLMs and vision models can finally parse the unstructured metadata inside a BIM model, and the data centre buildout has created an MEP capacity gap urgent enough to force adoption.
Get the deals, the deep takes, and the operator playbooks
One sharp construction tech briefing a week. No fluff.
Join 2500+ ReadersThe opportunity for new AEC software founders
What Joe actually wants founders to build
Joe’s cookie-cutter advice was two things. Deeply understand customer incentives: how does your customer make money, what workflows do they use? Outsiders skip this and it kills them. Then pick your attack vector and stay true to it.
The pricing model is where it gets interesting. Companies in his portfolio in legal and insurance are pricing on a share of incremental revenue or savings, not per seat. In construction that could mean charging a slice of the change orders prevented or risk eliminated from a developer’s portfolio. Those numbers are orders of magnitude bigger than the line Revit currently occupies.
The operator playbook: what construction execs should do
Mindset first, then test, then scale
The most useful part of the conversation came at the end. Joe’s three-step plan: shift your mindset by using frontier AI tools as a consumer (pay for the advanced subscription so you see what it can really do). Then find the work inside your firm that is rote and capacity-constrained. Then test, iterate, and scale what works.
He is a VC, so the optimism is on brand. But the point holds. If your competitor is using AI to clear their MEP backlog while you are still hand-routing ductwork, the developer paying for both projects will eventually notice.
| Attack vector | What it means | Example startup | Joe’s read |
|---|---|---|---|
| Replace Revit | Build a cloud-native, AI-enabled BIM platform from scratch | Motif | Hardest play. Joe said it’s more achievable over the next decade, but not within five years |
| Build around Revit | Own the workflows Revit handles poorly, like cross-discipline document review | LightTable | Strong wedge. The legacy human review process currently costs $50-100K per project, per the a16z thesis |
| Services layer | Automate the human work on top of Revit, particularly MEP design | Endra | Joe’s preferred attack vector today. $150B global MEP services market with modular, rules-based work |
Frequently asked questions
Quick answers to the things people keep asking
Joe’s view: no. Over the next decade it becomes more achievable potentially, but he argued you may not even need to displace Revit to build a huge company, because the workflows around it are now in play. (Source)
A $150 billion market where most work is rule-based and code-constrained. Big firms have already modularised and offshored the repetitive tasks, so AI does not have to change the workflow. It just changes who completes it. (Source)
Joe argues the opposite. Design firms have huge backlogs they cannot serve because they do not have the people. AI lets them say yes to projects they would have turned away. The market grows, it does not shrink. (Source)
Joe acknowledged it, but pointed out rework is a liability event for the MEP designer who has to fix it. As developers get more sophisticated and projects get bigger (think data centres), he expects incentives to shift toward GCs that deliver on time, on budget, and in scope. (Source)
Pay for the advanced subscription on a frontier AI tool and use it as a consumer. You cannot make good decisions about deploying this technology if you have not seen what it can do. Then find one rote workflow inside your business and start testing. (Source)
Related Articles
The VC Firm Behind Facebook Just Told Construction: Your Software Is Next
Nemetschek Acquires HCSS: Why the Infrastructure Bet Is the Biggest Signal in ConTech Right Now
Latest Construction Technology Funding Rounds – 13th April 2026
