FAQ¶
Common questions about Anvil, organized by topic. These are written for the team — if you're explaining Anvil to someone, start here.
About the Business¶
Q: What does Anvil actually do?
We turn CAD files into finished, inspected machined parts in 3-5 days. We own and operate standardized CNC facilities powered by an AI-CAM engine that automates the programming bottleneck.
Q: Why vertical integration instead of licensing software?
Selling AI-CAM to existing shops means dealing with hundreds of different machines, tool libraries, and fixtures — an intractable generalization problem. By owning our machines, we control every variable. We also close the learning loop: every part we cut generates data that improves our models.
Q: Who are our target customers?
Hardware teams building robotics, EV, climate tech, medical devices, industrial equipment, and drones — anyone iterating on physical products who needs parts in days, not weeks. We're not targeting aerospace/defense initially (certification overhead) or hobbyists (Protolabs serves that). See Targets & Scaling for Alpha Facility specifics.
Q: How do we make money?
Per-part revenue at ~$600 average selling price. Margin comes from automation: AI-CAM removes the 1-2 hours of manual programming per part that traditional shops need, and standardized cells run at higher utilization. Target is 50-60% gross margins at scale.
Q: What stage is the company?
Pre-revenue, seed-funded. We're building the AI-CAM engine and preparing to open the Alpha Facility. First revenue expected mid-2026.
About the Technology¶
Q: How does AI-CAM actually work?
We combine ModuleWorks (proven commercial CAM kernel) with our proprietary AI layer. The system analyzes part geometry, recognizes machinable features, selects tools from our calibrated library, determines fixturing and orientation, and generates collision-free toolpaths. Reinforcement learning improves decisions based on actual machining results. See the CAM Pipeline for technical details.
Q: What's the path to 90% automation?
Start narrow: prismatic aluminum parts with standard features (holes, pockets, faces, fillets) — this covers 60-70% of prototype demand. Train on real customer parts with expert machinists validating and correcting CAM programs. RL optimizes based on actual cycle times and quality. Expand coverage incrementally as the model proves capable.
Q: Why one machine type, one material?
Constraining the problem makes it tractable. One machine (DVF 5000), one tool library (120 HSK-A63 tools), one material (6061-T651 aluminum) means the AI-CAM only needs to learn one set of physics. Once solved, we expand scope deliberately.
About the Market¶
Q: How are we different from Xometry?
Xometry is a marketplace — they route orders to third-party shops. They don't own machines, don't control quality, and don't have manufacturing data. Their lead times are stuck at 7-10 days because the underlying shops still use manual CAM. We own the machines and automate the CAM. See Competitive Landscape for the full breakdown.
Q: How are we different from Hadrian?
Hadrian is building software-defined factories for aerospace/defense. Their center of gravity is factory orchestration, with CAM still requiring meaningful manual labor (CAM programmers, CMM programmers). Our core is AI-CAM automation — we automate the programming bottleneck first, then scale. Different wedge, different sequencing, different market.
Q: How are we different from Protolabs?
Protolabs automates CAM by restricting geometry to simple 2.5D parts. That caps their market. We're building AI that handles complex 5-axis geometry, targeting the much larger high-mix, tight-tolerance segment they can't serve. See Competitive Landscape for more.
About Working Here¶
Q: Where are we located?
The Alpha Facility will be in the Boston area (scouting locations in Sudbury, Danvers, and Waltham). Engineering work is also based in the Boston area.
Q: How big is the team?
We're a team of 4 founders growing to ~13 in Year 1.
Q: What's the Alpha Facility?
An 8,000 sqft facility in the Boston area with 1-3 DVF 5000 machines. It's our proving ground — where we validate that AI-CAM works on real customer parts and that unit economics hold before scaling. See the Alpha Facility overview for details.