Forge Automation Made the Machine Shop Behave Like Software

Ask a software engineer what it feels like to wait two weeks for a build to compile and they will look at you like you asked whether they enjoy stubbing their toe. That absurd wait is exactly what mechanical engineers have accepted as normal for decades. Need a custom aluminum bracket to test a robot arm? Email a shop, wait for a quote, wait some more, and hope the part shows up in two or three weeks. Forge Automation looked at that gap and decided it was a bug, not a feature.

The Toronto company, part of Y Combinator's Winter 2025 batch, runs what it calls software-enabled factories. The pitch is short enough to fit on a business card: custom CNC parts in four days or less. An engineer uploads a CAD model to forgejobshop.com, gets an instant quote, pays, and the finished metal shows up days later. No sales call. No spreadsheet of estimates. The machine shop, reorganized to feel like an online checkout.

Imagine having to wait 2 weeks to compile your code - you would hate your life. Forge Automation, YC launch

The problem, stated plainly

01Hardware moves at the speed of its slowest part

Every hardware team - robotics, consumer electronics, aerospace, medical devices - lives and dies by iteration speed. The faster you can hold a real part in your hand, the faster you learn whether your design works. But the custom parts that feed that loop have historically been slow to get. Traditional shops quote by hand, schedule by phone, and program their machines one job at a time. Two to three weeks of lead time is standard, and each delay stretches the gap between a good idea and a working prototype.

Forge's founders had felt this personally. Both are University of Waterloo engineering graduates who did their time inside serious hardware companies before starting up.

Who is behind it

02Two engineers who got tired of waiting

Timothy Seto, co-founder and CEO, has more than a decade of machining experience and interned at Apple, Formlabs, and iRobot before graduating from Waterloo engineering in 2022. His co-founder, Walter Raftus, studied AI and Mechatronics at Waterloo and did stints at Nvidia, Trexo Robotics, and Intel PSG. One knows the shop floor. The other knows how to make software think. That split shows up directly in the product: a real factory with real machinists, wrapped in code.

They started Forge in 2022 and spent two years quietly shipping parts before their public YC launch in early 2025. By then they had customers ranging from early-stage startups to Fortune 10 companies, and thousands of parts out the door - a useful thing to have before you tell the world what you built.

How the thing actually works

03From CAD file to cut metal

The ordering flow is deliberately boring, which is the point. You log in, drop in a STEP file, and the site gives you a price on the spot. What makes the instant quote possible is the software underneath.

The engine in the middle is Foundry, Forge's in-house CAM and design-for-manufacturability software. Traditionally, turning a CAD model into a program a milling machine can run is skilled, slow, human work. Foundry automates a large chunk of it - reading the model, making manufacturing decisions, and generating the toolpaths the machine follows. That automation is what lets Forge quote instantly and hold a four-day promise instead of drowning in manual programming.

There is a small but meaningful design choice buried in the workflow: customers can add manufacturing intent directly onto their CAD model instead of writing a separate spec sheet or emailing notes back and forth. Tolerances, finishes, and the things a machinist would normally have to ask about get baked into the file. It sounds minor. It is the kind of minor that removes an entire round of human clarification from every order, which is exactly how you make instant quoting believable rather than a marketing line.

Software-enabled factories that manufacture CNC parts in 4 days or less. Forge Automation

The speed gap, drawn out

04Four days versus three weeks

The clearest way to understand the wedge is to put the lead times side by side. A traditional custom-CNC order commonly lands somewhere around two to three weeks by the time quoting, scheduling, and machining are done. Forge compresses that window.

That difference is not cosmetic. For a team running weekly design reviews, cutting the part wait from three weeks to four days can mean testing several revisions in the time it used to take to test one.

How it makes money

05Two ways to buy machine time

Forge sells the same underlying capacity in two shapes. The first is straightforward on-demand ordering - you pay per part for flexibility and speed. The second is more interesting: Forge Reserve Machines, where a customer reserves a dedicated CNC machine on Forge's floor, staffed by Forge machinists, with a guaranteed production SLA. It is the difference between hailing a cab and leasing a car with a driver. Teams that need a steady flow of parts get predictable turnaround; Forge gets recurring, committed demand.

Reframing a job shop as reservable capacity is a quietly clever move. It borrows the logic of cloud computing - reserve versus on-demand - and applies it to physical machine time.

Option	Best for	Trade-off
On-demand	One-off or bursty prototype needs	Higher per-part price, maximum flexibility
Reserve Machines	Teams needing steady 4-day turnaround	Committed capacity, guaranteed SLA

Where it sits in the market

06Not the only shop online, but a fast one

Forge is not the first company to put custom manufacturing on the internet. Xometry, Protolabs, Fictiv, PartsBadger, and Protolabs' Hubs network all offer instant-ish quoting for machined parts. What Forge is betting on is the combination of speed, price, and software depth - owning the CAM automation and the factory floor rather than mainly brokering orders to a network of third-party shops. The four-day promise and the reserved-machine model are its sharpest points of difference.

The company positions itself squarely at hard-tech builders: consumer hardware, robotics, and the kind of teams for whom a slow part is a real tax on progress. That focus keeps the pitch honest. Forge is not trying to make everything for everyone; it is trying to make the prototype-to-part loop fast for people who iterate constantly.

The best startups often attack the one step in a workflow nobody bothered to optimize. For hardware, that step is getting the part in your hands.

The bet

07Atoms and bits under one roof

The hard part of Forge's model is also the interesting part: it is running a real factory and a real software company at the same time, with a team of around twenty people. Most startups pick one - pure software, or pure hardware. Forge is arguing that the two belong together, that the way to make manufacturing faster is to let code make the decisions humans used to make by hand, while keeping the machines and machinists in-house so the promise is actually enforceable.

The expertise that makes this credible is the unusual pairing at the top. Seto has spent years on the shop floor and knows what actually breaks when you try to machine a tricky part fast. Raftus builds the software that has to encode those hard-won judgments. Hardware companies often stumble because their software people do not understand the physical constraints, or their machinists do not trust the automation. Forge's answer is to keep both kinds of people in the same building, on the same team of about twenty, arguing with each other until the software gets the physics right.

The seed round it raised alongside Y Combinator's Winter 2025 batch is early-stage money for an ambitious idea. Whether Forge becomes the default way hardware teams get their parts depends on execution across both halves of the business - keeping the software smart and the floor fast as volume grows. Scaling a factory is not the same as scaling a web app; every new machine is capital, floor space, and a machinist to hire. But if Foundry keeps taking more of the manual work out of each order, the math gets easier as volume climbs rather than harder.

For now, the proposition is refreshingly concrete: send us a CAD file, get metal back in four days. In an industry that taught engineers to expect three weeks, that alone is worth paying attention to. Forge Automation is betting that the future of custom manufacturing looks less like a phone call to a shop and more like a page you already know how to use - upload, quote, checkout, and wait a few days, not a few weeks.