One of the fundamental challenges in bringing physical products to market stems from the shortcomings of the software giants dominating the industry. These incumbents have failed to innovate meaningfully since the 20th century and have weaponized proprietary file formats to lock engineers into rigid workflows. The strict flow of knowledge and process is all downstream from their cash cow: CAD. Knowledge moves from CAD → Simulation → PDM → PLM → CAM → MES → ERP with a handful of other less critical tools mixed in.

The proprietary data formats used by these incumbent software providers intentionally create integration barriers by limiting the ability to extract meaningful data from CAD models and transfer it downstream into other software. This strategy, in theory, would force organizations to adopt entire suites of tools from a single provider, effectively locking them into an ecosystem that is grossly overpriced, but integrated. In reality, different providers are better at different tools, so hardware organizations end up picking different software from multiple providers, all of which are siloed and not integrated.

The lack of integration between these critical tools creates severe implications for hardware enterprises. As knowledge and data moves through the fragmented toolchain by word of mouth and manual entry, errors accumulate and are amplified, potentially leading to costly mistakes in design, manufacturing, and purchasing processes. For large companies, these compounding errors can result in millions of dollars of unnecessary spending on rework, material waste, and production delays. Even more alarmingly, for upstarts and smaller businesses with limited resources, these cascading inaccuracies can be catastrophic, potentially leading to product failures and even the demise of the entire venture. The inability to seamlessly transfer accurate data between tools not only cripples efficiency but also introduces significant business risk, underscoring the urgent need for a more integrated and error-resistant approach to hardware development.

It has been well-reasoned that the commodity software-for-hardware challengers need to be brought in-house to retool and unite the toolchain is the CAD format and models themselves. Any company that successfully owns and integrates the entire toolchain end-to-end stands to unlock hundreds of billions of dollars of value.

Industry loves “first-principles approaches” to solving new problems. In this case, the first-principles answer, that a small pocket of the internet also loves to talk about, is a new CAD.

The CAD market has been relatively stagnant over the last 20 years with the biggest challenger to this ecosystem being OnShape, founded in 2012. OnShape's founder, also the ex- founder of SolidWorks, brought unparalleled tribal expertise and commanded obscene amounts of funding, everything you’d need to go build the next great CAD, but it hasn’t played out as the industry hoped it would. Venture Capital and PTC contributed at least $650,000,000 to OnShape (it’s likely well over), but even after 12 years of development, the product is nowhere near feature complete, let alone parity with the rest of the market. They’ve shown flashes of brilliance with their approach to version control & collaboration, but there is no serious viability of the product eating up market share over the course of the next decade. Even if OnShape achieves feature parity over the coming decade, they would still require step functions up in network capabilities, only to in their best case, eat into margins of Dassault and Siemens. If Dassault and Siemens are smart, they will gradually drop prices enough to retain customers which, in the end, leaves OnShape neglected.

A new closed-format CAD hasn’t successfully been able to eat its way into the market, but there is a lot of discourse surrounding an open source CAD. I talk about OnShape because even with ~15 years, probably ~$1 billion in funding, and some of the most experienced CAD engineers, they were not able to succeed in CAD parity, but somehow there's this expectation that open-source CAD will. I think it is ridiculous to entertain the idea of open source CAD based off of the lack of resources alone, but if we want to have the conversation, here is my take.

The primary benefit people seem to hope from an open source CAD surrounds data-compatibility as there would be a universal interface to accessing data from models - a necessary part of building universal software to build an integrated and agnostic toolchain. In addition to the traditional building blocks (kernel, sim, CAM), an open source CAD would also require the ability to lossless-ly transfer from all closed-format CAD’s to this open version. Every single kernel is different, so creating a universal format as the base of a 1:n mapping could cripple the integrity of CAD files due to issues like rounding errors. Ironically, the format you’d end up creating would not be universally and bi-directionally transferrable (therefore not open) and you’d end up spending however many hundreds of thousands of man-hours and hundreds of millions of dollars building a new open-source, but closed format CAD. This is the same form factor in which the ecosystem exists today.

The last time we tried to do this (minus the billion dollars) was OCC → FreeCAD, and we all know how that turned out.

Despite how unbelievably hard a new CAD is, we need it. It’s the substrate that stands between us and a truly AI-native engineering experience—one where models are interoperable, intent is machine-readable, and the rest of the toolchain becomes composable by default. If we want to unlock real abundance for humanity, we must develop new technology faster and we have to rebuild the thing everything else is downstream from.