Manufacturing Startups

An analysis

The world of “hard tech” - brought into the popular narrative by Andreessen’s American Dynamism practice and the success of companies like Palantir and Anduril - has spawned a large number of startups focused on manufacturing, what I’d call a subset of the hard tech space. In this post, I try to define categories, examples, and share some thoughts about the business models of some of these ventures.

Since joining the ecosystem in 2016 when starting at SpaceX, there’s been a flourishing of manufacturing-oriented companies. I’ve been fortunate to meet and become friends with a number of the founders of the companies mentioned below, and in full disclosure, have either advised or put in a tiny angel check into a few of these. A common theme amongst them? A higher-than-usual level of grit and work ethic given the challenge of building a company in this industry.

The focus below are on companies that specifically operate in the manufacturing space, as opposed to those who develop products and do final assembly and test themselves, such as Anduril, Base Power, Harbinger, Saronic, and many others. Often, you’ll find that the companies mentioned below build their wares for the defense primes, commercial aerospace companies, energy industry, or automotive.

It’s more important now than ever that this ecosystem matures and thrives in the US.

Manufacturing Software as a Service (MSaaS) companies

This business model is relatively straightforward: develop software that makes the process of making things better, faster, and/or cheaper. Similarly, the VC investment thesis tends to be relatively traditional, with a heavy focus on user and monetization-based metrics such as net dollar retention, user growth, etc.

In general, I find that this business model has some of the same issues that traditional SaaS has; in particular, defensibility. It's somewhat rare to see Manufacturing SaaS companies have some of the advantageous dynamics of traditional SaaS, such as network effects and the ability to land and expand within an organization. However, what manufacturing SaaS has going for it compared to traditional SaaS is its stickiness. Especially for platforms that become systems of record or core to the operations of their manufacturer-customer, stickiness can be a powerful growth lever, especially as new features are added during the build-out of the product.

One dynamic that’s under-appreciated: large (or soon to be large) modern manufacturers with well-developed software capabilities will often decide to build their own platforms as opposed to buying. I’ve found that it's relatively difficult to generalize software across an entire type of product (like a plane), let alone all manufactured goods, and there’s a lot of value in manufacturers developing their own tech. Furthermore, with recent advances in AI, the barrier to building your own internal software for a manufacturer has decreased; decreasing the potential outcome of companies in this category.

Examples

• First Resonance and Epsilon3 build Manufacturing Execution System (MES) software, which runs the daily operations of a factory. MES houses work instructions quality data and process sequencing, while interacting with more financially-minded software stacks like a company’s Enterprise Resource Planning (ERP) system.

• Dirac and Quarter20 build work instruction & visualization software for manufacturers, somewhat of a subset of the feature set described in MES’s above. Complex, precision-assembled products such as those often found in aerospace and defense, need detailed and highly visually-intensive assembly instructions for technicians.

• Sift and Nominal build test infrastructure and data management software for complex testing and simulation runs such as flight test of a drone.

• Adam, Hestus, Camfer (and many others popping up in this space, mostly from YCombinator) are building AI-powered CAD software; essentially using natural language to automate the simpler portions of creating 3D models of hardware or using AI in other parts of the CAD workflow.

• Stell builds software to manage quality data between vendors and manufacturers and internal to a manufacturer.

• Galvanick builds cybersecurity software for manufacturers that targets the protection and management of physical equipment.

Dirac’s BuildOS software for automated work instruction generation

“Make-parts-for-other-people” companies

This one’s relatively self-explanatory: these companies make parts for other people, as opposed to for their own use in a higher level assembly. There’s some interesting variations in the examples below, but I see some common themes throughout this list:

1. All are focusing on one “method” of making parts as a start, oftentimes with ambitions to expand to others.

2. They are generally not focused on changing the physical way the part is made (more on that in the next section), and therefore are normally buyers of existing capital equipment and machinery

3. All are or have already developed software to manage the process by which they make parts, oftentimes creating similar feature sets to those developed by the companies above. As I mentioned earlier, these feature sets typically tend to be hyper-specific to the manufacturing method or product type that these companies are working on and thus not built for sale.

4. Given the capital-intensive nature of these businesses, they typically raise a lot of capital, and/or find ways to creatively finance equipment and facilities outside of traditional VC equity funding

Examples

• Hadrian, an early entrant into this space, builds highly-automated, software-defined machine shops for making metallic high precision components. They’ve built a successful business, abstracting away a lot of the complexities of machining and inspecting tight-tolerance parts into an internal software stack that manages most aspects of the business.

• Senra Systems builds wire harnesses - often the last-designed and most often changed part of any complex piece of hardware, as wire harnesses connect systems to each other. Senra has built design and production automation software that speeds up the time to production and improves quality of finished parts

• Layup Parts builds, well, parts that require a layup process: composites. Similar to others, they have developed internal software tools to improve documentation, creation of manufacturing processes from design files, and development of tools that are used to make the parts.

• AMCA, a recent entrant into the space, buys manufacturers with existing products and improves their operations and engineering, expanding their capabilities to their customers in the aerospace and defense industry.

• Re:Build is an interesting example that (in my words) is part PE firm part manufacturing tech startup. They are buying up companies with various manufacturing methodologies - including some non mfg companies like engineering firms - in order to produce finished assemblies or whole finished products. This one’s a bit unique for the category.

Hadrian’s automated machine shop, ran by their in-house developed full suite of software

“Change-the-way-the-part-is-made” companies

These are a rare breed of manufacturing companies which innovate on the physical part fabrication methodology. They’re oftentimes iterating on fundamental research with respect to material science and related fields, to improve quality, speed, or throughput of their given process.

While also doing that, these companies also make parts for other people, like the category above, but instead use their own methods to do so. In certain cases, these companies license the technology that they have developed to others. As such, the thesis with these companies is that they are able to both achieve high levels of sophistication on their technology, in addition to the ability to build factories and operate them: perhaps the toughest aspect herein.

Examples

• Rangeview has developed a novel method of making investment castings: often used for parts with highly complex geometries made from exotic alloys. By directly printing the mold, or “shell”, multiple steps of the traditional process are removed, saving cost, lead time, and speeding up iteration cycles on this notoriously high upfront engineering process.

• Machina Labs has developed a way to make stampings - complex geometry formed sheet metal parts - without dies. With a somewhat similar value proposition to Rangeview, this method increases the speed of iteration on parts like this, which often require months of time and huge capital investment to “tool up” before production. Their system uses two opposing robots to push and pull the metal into shape. Its really cool.

• Atomic Industries has developed a novel method of creating tool and die designs for injection molding. Tools are oftentimes more complex than the parts that they make - essentially ‘negatives’ of the final plastic part. Similar to the companies above, the AI-powered design of the tools enables faster iteration cycles, removes the tribal knowledge associated with designing dies, etc.

Machine Labs’ “robotic craftsmen” producing complex geometry sheet metal parts without a die