The average collaborative robot contains somewhere between 25 and 50 unique machined parts. Not the electronics. Not the cables. The structural components — the joint housings, the sensor brackets, the encoder mounts, the end effector flanges. And every one of them, from the 15-gram connector plate to the 2-kilogram base joint housing, was machined on a CNC mill or lathe.
I know because I've quoted these parts. Dozens of them. And the pattern is always the same: a robotics company — usually a startup or a scale-up, rarely a multinational — needs small precision parts in batches of 100 to 1,000. They need ±0.01mm on bearing bores. They need the aluminum to be lightweight and strong. They need delivery in four weeks, not twelve. And they need a supplier who understands that robotics isn't just general machining with a different label on the purchase order.
This guide is about what makes robotics CNC machining different — the parts, the materials, the tolerances, the batch economics, and how to find a supplier who gets it.
What separates robotics CNC machining from general CNC work
General CNC machining is capability-driven: can you make this geometry, this tolerance, this surface finish? Robotics CNC machining adds a layer on top of that: can you make it consistently across 500 units, with process data to prove it, on a timeline that matches a product launch?
Three things make robotics parts distinct from general CNC work.
Part count per assembly. A robot isn't one machined part. It's dozens. A typical 6-axis cobot arm has machined components in every joint — the housing, the bearing seats, the motor mount flange, the encoder ring, the cable pass-through — plus structural brackets in the base and the end effector mounting interface. A startup developing a cobot might need 30 unique machined parts sourced simultaneously, each at 50-200 units for the first production batch. That's a sourcing complexity problem, not a machining complexity problem. The shop that handles it well has project management capability, not just machine tools.
Weight is always a constraint. Every gram in a robot arm increases the torque required at every joint upstream of it. A 100-gram weight savings at the wrist reduces the required motor torque at the elbow, shoulder, and base — compounding through all upstream joints. This is why robotics parts are almost always aluminum (6061-T6 or 7075-T6), why they're pocketed and webbed to remove every gram of non-structural material, and why wall thicknesses are pushed to the minimum the material and machining process can hold.
Surface finish isn't just cosmetic. In a robot joint, the surfaces where bearings seat need specific surface roughness — typically Ra 0.8μm or better — to ensure proper press-fit retention and heat transfer. Too rough, and the bearing doesn't seat fully. Too smooth, and the press-fit retention force drops. The surface finish callout on a robotics drawing isn't a preference; it's a functional requirement that affects assembly yield and product lifetime.
The robotics parts landscape: what actually gets machined
Robotics covers a wide range of machines, but the machined parts break down into a few common categories that repeat across almost every robot design:
Joint and structural components
These are the core of any robot arm or automation mechanism. They carry the load, house the bearings, and transmit the motion. Getting them wrong means a robot that chatters, overheats, or loses positioning accuracy over time.
| Part Type | Typical Material | Key Tolerances | Batch Size | Notes |
|---|---|---|---|---|
| Joint housing | 6061-T6 or 7075-T6 aluminum | Bearing bore ±0.01mm, face flatness 0.02mm, bore concentricity 0.02mm | 100-500 | Dominant cost is boring and facing time; pocketing for weight reduction adds cycle time |
| Motor mount flange | 6061-T6 aluminum | Pilot diameter ±0.01mm, bolt circle true position ±0.05mm | 100-500 | Motor pilot fit is critical — a loose fit allows misalignment that accelerates gearbox wear |
| Encoder mount ring | 6061-T6 aluminum or 304 stainless | Bore ±0.005mm, perpendicularity to mounting face 0.01mm | 100-300 | The most precise part in most robot assemblies; encoder accuracy depends on perfectly concentric mounting |
| Cable pass-through | 6061-T6 aluminum | ±0.1mm on through-holes, edge break 0.2-0.5mm | 200-1,000 | Simpler geometry, higher quantity; often overlooked but present on every joint |
| Base plate | 6061-T6 or 7075-T6 aluminum | Mounting surface flatness 0.05mm, dowel pin true position ±0.02mm | 50-200 | Largest machined part in most robot designs; flatness is critical for mounting repeatability |
For a detailed breakdown of aluminum alloy selection, see our aluminum CNC machining complete guide.
End effector and gripper components
End effectors are where customization explodes. Every application needs a different gripper, a different tool mount, a different sensor bracket. And these parts are almost always machined from billet because the quantities are too low for casting and the geometries too complex for stamping.
Typical parts: pneumatic gripper finger blanks (6061 aluminum, ±0.05mm on gripping surfaces, batch 200-500), vacuum cup mounting plates (aluminum or acetal for non-marring surfaces, batch 100-300), tool changer adapter plates (7075 aluminum, ±0.02mm on alignment features, batch 50-200), vision system mounting brackets (aluminum, ±0.1mm typical, batch 50-150).
End effector parts are a good business for a CNC shop because every integrator and every application needs different ones. The same robot arm might have ten different end effectors across ten different customer installations, and each end effector contains 3-8 machined parts. The repeat business comes from the variety, not the volume — which is exactly where an independent precision shop competes best.
AGV and AMR structural components
Autonomous Guided Vehicles and Autonomous Mobile Robots are the fastest-growing segment of the automation market. Unlike robot arms, where castings are starting to displace billet machining at the high end, AGV/AMR structures are almost entirely fabricated from machined and welded aluminum components.
Typical parts: drive wheel mounting brackets (6061-T6 or 7075 aluminum, pilot bore ±0.01mm, batch 100-500), suspension arms (7075 aluminum, bearing bores ±0.01mm, batch 100-300), sensor tower mounts (6061 aluminum, flatness 0.05mm, batch 50-200), battery tray brackets (6061 aluminum, lower precision ±0.2mm typical, batch 200-1,000), lift mechanism components (steel or aluminum, depending on load capacity, batch 50-200).
The drive wheel brackets are particularly interesting. Every AGV has at least two drive wheels, each with a machined bracket that houses the motor, gearbox, and wheel bearings. These brackets take all the vehicle weight through the suspension, plus dynamic loads from acceleration and braking. A typical bracket is roughly 150mm × 100mm × 40mm, machined from 7075-T6 for strength, with bearing bores at ±0.01mm true position. At 100-500 units per AGV model, it's the kind of recurring production work that keeps a CNC shop's spindle turning.
Sensor and perception hardware
Every robot, cobot, and AGV carries sensors — and every sensor needs a mount. These are small parts, often under 50mm in any dimension, but the precision requirements rival encoder components.
Typical parts: LIDAR mounting brackets (aluminum, angular tolerance ±0.1°, batch 100-300), camera housing bodies (aluminum, threaded holes for lens mounts ±0.05mm true position, batch 100-500), ultrasonic sensor holders (aluminum or 304 stainless for outdoor AGVs, batch 200-500), IMU mounting blocks (aluminum, flatness 0.02mm, batch 100-200).
The LIDAR bracket is worth understanding. A LIDAR unit mounted on a robot relies on the bracket to hold it at a precise angle relative to the robot's coordinate frame. The angular tolerance on the mounting face might be specified as ±0.1° — which, over a 100mm bracket length, translates to roughly ±0.17mm of shimming at the mounting points. That's not a tight machining tolerance by CNC standards (±0.1mm is wide open), but it requires flatness, parallelism, and careful deburring so the sensor seats cleanly against the mounting surface without rocking.
Materials for robotics: aluminum is king, but not the whole story
The dominant material in robotics machining is aluminum — specifically 6061-T6 and 7075-T6. Here's the material breakdown across a typical cobot assembly:
| Material | % of Machined Parts | Typical Application |
|---|---|---|
| 6061-T6 Aluminum | 60-70% | Joint housings, brackets, covers, base plates, sensor mounts |
| 7075-T6 Aluminum | 15-20% | High-stress components: motor flanges, bearing housings, structural joints |
| 304 Stainless Steel | 5-8% | Shafts, pins, fasteners exposed to moisture or cleaning chemicals |
| 4140 or 4340 Steel | 3-5% | Drive shafts, gear blanks, high-wear sliding surfaces |
| Brass C360 | 2-3% | Pneumatic fittings, sensor bushings, electrical grounding components |
| Engineering plastics (PEEK, Acetal) | 2-5% | Non-conductive spacers, sliding bearings, gripper pads |
The dominance of aluminum isn't about cost — though 6061 is cheap. It's about weight. A 6061 joint housing weighs roughly one-third of what a steel housing with the same stiffness would weigh. And as I mentioned, weight savings at the wrist compound through every upstream joint. For a 6-axis arm, removing 200 grams from the wrist assembly can reduce the required motor torque at the base by 15-20%. That's the difference between a robot that runs on a standard servo and one that needs a larger, more expensive motor with a bigger power supply and more thermal management.
For a full discussion of material-driven cost and lead time tradeoffs, see our guide on how material choice affects CNC machining cost and lead time.
Tolerances that actually matter in robotics
Not every dimension on a robotics part needs ±0.01mm. The art is knowing which ones do. Here's a practical breakdown of what tolerances are driven by what functional requirements:
| Functional Requirement | Typical Tolerance | Why | What Happens If You Open It |
|---|---|---|---|
| Bearing bore diameter | H7 (+0.015/-0 mm for 20mm bore) | Press-fit bearing retention requires controlled interference | Bearing walks out of housing under thermal cycling |
| Bearing bore concentricity | 0.02mm between opposing bores | Misaligned bearings bind and overheat | Bearing life drops from years to months |
| Motor pilot diameter | ±0.01mm | Motor shaft alignment to gearbox input | Gearbox noise and premature wear |
| Dowel pin true position | ±0.02mm | Repeatable assembly alignment | Robot loses home position accuracy after maintenance disassembly |
| Mounting face flatness | 0.05mm over 100mm | Prevents distortion when bolted to mating surface | Built-in stress causes joint housing to warp under assembly torque |
| Cable pass-through holes | ±0.1mm | Cable clearance, no functional precision required | Nothing — this is where you save cost |
| Non-structural pocket floors | ±0.2mm | Weight reduction pockets, cosmetic | Nothing — another cost-saving opportunity |
The mistake engineers make most often on robotics drawings is calling out tight tolerances everywhere. I see drawings where a cable clearance hole gets ±0.01mm and a bearing bore on the same drawing gets ±0.01mm. The bearing bore needs it. The cable hole doesn't. Each unnecessarily tight tolerance adds inspection time and, in production, adds cost without adding value.
For a comprehensive tolerance reference, see our CNC machining tolerances guide.
Batch production economics for robotics parts
Robotics companies typically follow a three-phase product lifecycle, and the CNC requirements change at each phase:
Phase 1: R&D and prototype (5-30 units). Parts are for internal testing, investor demos, and pilot customer trials. Speed matters more than per-part cost. The shop needs to turn quotes fast, machine parts accurately from sometimes-incomplete drawings, and be comfortable with design changes between batches. Most general CNC shops can handle this phase.
Phase 2: Pilot production (50-300 units). This is the transition point where process control starts to matter. The parts are the same as R&D, but now you need 200 of them identical, not 5. The shop needs to build dedicated fixturing, establish inspection protocols, and start tracking process data — even if the customer hasn't asked for it yet. This is also the phase where shops that are really prototyping shops start to struggle, because their workflow is built around changeovers, not repeatability.
Phase 3: Production scale-up (300-2,000+ units). At this point, the shop needs production capability: multi-part fixturing, SPC on critical dimensions, material traceability, documented non-conformance handling, and the ability to turn repeat orders with consistent lead times. The per-part cost drops significantly from Phase 2 — typically 25-35% — because setup is amortized, tooling is optimized, and the cycle time has been dialed in over several batches.
Here's an actual cost comparison for a cobot joint housing (6061-T6 aluminum, roughly 100mm × 80mm × 60mm, six machined faces, bearing bores, threaded mounting holes) across the three phases:
| Phase | Quantity | Setup Amortization | Cycle Time | Material | Tooling | Inspection | Per-Part Total |
|---|---|---|---|---|---|---|---|
| R&D | 10 pcs | $35.00 (dominant) | $22.00 (28 min) | $5.00 | $3.00 | $15.00 (100%) | $80.00 |
| Pilot | 200 pcs | $1.75 | $18.00 (23 min) | $4.50 | $2.50 | $4.00 (sampling) | $30.75 |
| Production | 1,000 pcs | $0.35 | $15.00 (19 min) | $4.00 | $2.00 | $2.50 (SPC) | $23.85 |
Setup cost drops from $35/part to $0.35/part. Cycle time drops 32% from R&D to production as feeds and speeds are optimized and multi-part fixturing eliminates tool-change dead time. Inspection cost drops from 19% of part cost to 10% as the process becomes capable and sampling replaces 100% inspection.
The shop that can carry a customer through all three phases — without the customer having to re-qualify a new supplier at each transition — is worth more than the lowest-quoted price at any single phase. That continuity eliminates requalification lead time, preserves institutional knowledge about the parts, and avoids the risk of a new shop making the same mistakes the old shop already learned from.
How robotics companies buy CNC services
Robotics companies buy CNC machining differently from automotive companies or general industrial buyers. Understanding the pattern helps you recognize a robotics RFQ when it appears.
They send assemblies, not individual parts. A robotics startup doesn't send an RFQ for "aluminum bracket, qty 200." They send an RFQ for "joint 3 assembly — 6 unique machined parts, qty 200 each." The RFQ is structured around the subassembly, not the individual component. The shop's ability to manage multiple part numbers as a package — coordinating delivery so all six arrive together, not staggered — directly affects assembly line scheduling at the customer.
They care about weight documentation. A robotics company's mechanical engineering team has weight budgets for each joint and each subassembly. They need the actual machined part weight, not the CAD-calculated weight, to validate their models. Shops that can provide post-machining weight data (a simple scale measurement per part, reported in the inspection documentation) get noticed.
They iterate mid-production. It's common for a robotics startup to discover during pilot production that a wall can be thinned by another 0.5mm, or that a cable routing channel needs to move 2mm. The first production batch of 200 might have a design revision at part 50. A shop that handles mid-batch engineering changes without drama — updating the program, documenting the change point, and segregating the old and new revision parts — is worth its weight in gold to a hardware startup.
They need suppliers who can scale with them. The classic problem: a robotics startup finds a prototyping shop that does great work for 20 units. When the startup raises Series A and needs 500 units, the prototyping shop can't handle the volume. The startup has to find and qualify a new supplier — losing three months and paying for new tooling and new first articles. The better model: find a shop that handles 20, 200, and 2,000 units under one roof, and plan to stay there as you scale.
Surface finishes for robotics parts
| Finish | Material | Purpose | Appearance | Cost Impact |
|---|---|---|---|---|
| Type II anodize, clear | 6061/7075 Al | Corrosion protection, moderate wear resistance | Satin silver, can have slight shade variation batch to batch | $2-4/part at small batch |
| Type II anodize, black | 6061/7075 Al | Corrosion + aesthetics for visible external parts | Matte to satin black | $3-5/part |
| Type III hard anodize | 6061/7075 Al | Wear surfaces, sliding components, gripper fingers | Dark gray-green, 25-50μm layer | $5-10/part |
| Electroless nickel plating | Aluminum, steel | Uniform coating, no buildup on threads, moderate corrosion | Bright silver | $5-8/part |
| Black oxide | Carbon/alloy steel | Mild corrosion protection, no dimensional change | Matte black | $1-3/part |
| Passivation | 304/316 stainless | Restores corrosion resistance after machining | No visible change | $1-2/part |
| Bead blast only | Aluminum | Uniform matte finish, hides minor tool marks | Matte silver-gray, uniform | $1-2/part |
For robotics, most external aluminum parts get Type II anodize (clear or black). Internal components — bearing housings, motor mounts, parts hidden inside joint shells — often ship as-machined or with light bead blast. Hard anodize is reserved for wear surfaces: gripper finger contact faces, sliding alignment pins, cam followers.
See our CNC machining surface finishes guide for process details on each option.
Five questions to ask a CNC supplier for robotics parts
"What's the largest batch of identical parts you've run in the last six months?"
If the answer is 20 or 30, the shop is a prototyping shop. That's fine for Phase 1, but you'll need a new supplier for Phase 2 and beyond. If the answer is 500 or 2,000, the shop has production capability and the workflow to prove it.
"How do you handle multiple part numbers in a single project?"
A good answer: "We assign a project lead who tracks all part numbers against a shared delivery date. You get a single point of contact and a consolidated status update." A weak answer: "Each part number gets its own order and its own delivery." The weak answer means you're doing the project management yourself.
"Walk me through how you'd handle a mid-batch design change."
A good answer describes: (1) stop production at a defined change point, (2) update CAM program, (3) machine a first article of the revised design and get customer approval, (4) resume production with clear segregation of pre-change and post-change parts, (5) document the change point in the inspection report. A bad answer is confusion about why this would be necessary.
"Can you provide per-part weight data with the inspection report?"
For robotics, this matters. A shop that says yes without hesitation has worked with weight-sensitive industries before. A shop that says "we can try" or "why would you need that" has not.
"What's your process capability on ±0.01mm bearing bores in aluminum?"
A good answer includes actual Cpk numbers from real production runs — "we typically hold Cpk 1.33-1.67 on H7 bores in 6061." A vague answer about "we hold tolerance" misses the point. Process capability isn't about whether the parts are in spec; it's about whether the process is centered and consistent enough that inspection sampling is statistically valid.
The bigger picture: why small batch precision CNC matters for robotics
The robotics industry is structurally reliant on CNC machining in a way that few other industries are. Automotive and consumer electronics have volumes that justify casting, stamping, and injection molding. Aerospace has volumes too low for those processes but certification requirements that lock out all but a handful of approved suppliers.
Robotics sits in a unique position: volumes too low for dedicated tooling (casting dies, progressive stamping dies), geometries too complex for manual fabrication, tolerances too tight for commodity machining, and certification requirements that are manageable — ISO 9001 is almost always sufficient for non-safety-critical components.
That's the opportunity. It's a growing market where precision CNC machining isn't a cost to be engineered out — it's the enabling manufacturing process for the entire product category.
If you're developing a robot, cobot, AGV, or automation system and need precision machined parts — whether 10 prototypes or 1,000 production units — send your STEP files. I'll review each part for DFM, flag any features that will drive cost at production scale, and return a per-part quote at your target quantity with the level of documentation your project requires. One point of contact, all part numbers managed as a package, delivery coordinated so your assembly line stays on schedule.