Foundries and factories both reshape raw materials into products that power industries, but their processes, equipment, and cost structures diverge sharply. Knowing when to choose a foundry versus a factory can save millions in tooling, lead time, and post-processing expenses.
Foundries melt metal. Factories assemble, machine, or mold parts from already-solid materials. That single difference ripples through every design decision you will ever make.
Core Process Differences
A foundry’s heartbeat is the melt deck: cupolas, induction furnaces, or electric arc vessels that reach 1 600 °C. The liquid metal is poured into expendable sand, investment, or die molds that are destroyed or re-coated after each shot.
Factories keep materials solid. They subtract metal with CNC mills, inject plastic into hardened steel tools, bend sheet on press brakes, or join parts with welders and riveters. No melt deck means no ladles, slag, or solidification shrinkage to manage.
Because metal changes state, foundries must model phase diagrams, nucleation rates, and gating ratios to prevent porosity. Factories worry instead about chip evacuation, tool wear, and weld penetration depth.
Heat as a Cost Driver
Maintaining a 30-tonne iron pour at 1 400 °C consumes 550–650 kWh per batch. Even with recuperative burners, energy can top 15 % of total part cost. Factories running cold processes like stamping or injection molding rarely see energy exceed 4 % unless they run cryogenic medical-grade clean rooms.
Foundries therefore cluster near cheap power or co-generation plants. Factories locate near labor or logistics hubs. Your site-selection matrix should weight kilowatt-hour tariffs more heavily when quoting cast solutions.
Tooling Economics
Foundry patterns cost 3–20 k USD for sand-cast aluminum manifolds, but the same geometry milled from 6061 billet can demand 80 k USD in fixture and cycle time. The crossover volume is roughly 300–500 pieces depending on complexity.
Die-cast tools jump to 150 k USD yet deliver 100-micron wall thickness and sub-millimeter repeatability for 500 k shots. CNC shops would need pallet pools and five-axis cells to match that pace, pushing capital past 1 M USD.
Investment-cast turbine blades require ceramic shell trees that last once; the tooling is the wax injection die, not the mold. Re-usable steel tooling inside factories lets you amortize cost over millions of cycles, but only if the design is frozen.
Break-even Calculators
Build a simple sheet: list pattern or die cost, expected volume, hourly rate, and scrap value. Set equal total cost equations and solve for quantity. You will find that sand casting wins below 1 000 parts, machining dominates below 100, and die casting owns the middle ground for non-ferrous alloys.
Remember to add the hidden line items: foundries charge for metallurgical testing, factories add inspection probes. Omitting these can shift the break-even by 30 %.
Material Spectrum
Foundries swim in metals: gray iron for vibration damping, ductile iron for crankshafts, Ni-resist for corrosion, Inconel for 1 100 °C exhausts. Each alloy family needs its own charge makeup, inoculant, and cooling curve.
Factories machine bar, plate, or forgings of the same alloys, but they are limited by mill availability. Want 42CrMo4 seamless tube in 180 mm OD with 3 mm wall? The mill minimum is 5 tonnes, and delivery is 14 weeks. A foundry can pour that chemistry tomorrow in a sand mold shaped to near-net.
Plastics, composites, and wood sit outside foundry reach. If your product mixes metal chassis with over-molded TPU grips, you will need both a foundry and a factory, plus an over-molding cell.
Alloy Switching Flexibility
Today’s pour can be 316L stainless, tomorrow’s can switch to Hastelloy by changing the charge recipe. A CNC shop stuck with 316L bar stock cannot pivot without ordering new billet and risking 20-week mill lead times. Foundries hold the upper hand on rapid alloy pivots for prototype iterations.
Yet flexibility has limits: magnesium cannot be poured in the same line as iron without cross-contamination. Segregate furnaces or accept costly crucible replacement.
Geometric Freedom
Cast walls as thin as 2 mm with internal oil galleries that snake in three dimensions. Machining those galleries would require five-sided access and long L:D ratio drills, driving scrap past 40 %.
Ribs, bosses, and undercuts that are free in a sand mold become nightmares on a five-axis mill. Foundries reward complexity; factories penalize it with extra setups.
The flip side: tight tolerances. A cast bore can arrive at ±0.5 mm, needing a 3 mm machining allowance. If your mating surface must sit within 10 µm, you will machine anyway, erasing some cost advantage.
Net-Shape Roadmap
Start with 3 mm extra stock on critical datums, 5 mm on non-critical faces. After first article inspection, reduce stock by 0.5 mm per production batch once dimensional capability is proven. You can often reach 1 mm final stock, cutting machining time in half without risking scrap.
Use simulation software like MAGMA or ProCAST to predict shrink porosity before cutting the pattern. A single virtual iteration costs 2 k USD but can save 50 k USD in weld repair.
Surface Finish Realities
Investment casting delivers 1.6 µm Ra as-cast, good enough for human-visible consumer housings. Sand casting lingers at 12.5 µm Ra, demanding 250 µm coat of paint or 0.5 mm machining skim.
Die casting lands near 3.2 µm Ra, but solder build-up on the die steel can create cold-flow laps that leak under 3 bar pressure. Factories polishing CNC parts hit 0.4 µm Ra in the same cycle time.
If your product touches food or pharma, you need <0.8 µm Ra for cleanability. Budget a final vibratory polish or electropolish pass, whether the part is cast or machined.
Finishing Cost Allocation
Allocate 8 % of selling price for sand-cast parts that will be painted, 4 % for die-cast parts that only need chromate. CNC parts anodized hard-coat can consume 12 % because racks, racking labor, and lost jig fingers add up.
Production Speed and Lot Sizes
Foundries need 4–6 weeks for pattern machining and first article, but once approved they can pour 1 000 parts per day through horizontally parted flask lines. Machining a 5 kg bracket from solid on a three-axis mill averages 45 minutes cycle time; 1 000 pieces need 750 machine-hours, roughly five weeks on a single spindle.
Die casting drops cycle time to 45 seconds, so 1 000 parts exit in one shift. The bottleneck becomes trim and shot-blast, not fill time.
Rapid prototyping favors factories: upload a STEP file to a 3-axis router and hold a plastic or aluminum part tomorrow. A sand foundry still needs a pattern, even if 3D-printed, extending lead to 5–7 days.
Bridge Quantities
Machine 50 prototypes to validate design, then switch to sand casting for 500 beta units, and finally move to high-pressure die casting at 10 k annual volume. This staged approach balances cash flow with learning curves.
Negotiate a tooling transfer clause: the same 3D-printed sand pattern can be used to machine the hardened steel die inserts, saving 30 % pattern-to-die rework cost.
Quality Control Protocols
Foundries test chemistry with spark-OES on every ladle, check nodularity in 3-minute metallographs, and X-ray 10 % of lot for porosity. A single 2 mm shrink cavity can trigger a 100 % recall in automotive brake calipers.
CNC factories probe every tenth bore with air-gauging to ±2 µm, log tool offsets automatically, and apply SPC charts in real time. Scrap is caught within the cycle, not after a 500-piece pour.
Weld repair is acceptable in castings up to 20 % of wall thickness if the specification allows. Machined parts rarely permit welding; any rework usually means re-machining or scrap.
NDT Selection Matrix
Use X-ray for internal shrink, dye-penetrant for surface cracks, and ultrasonic for thick sections over 50 mm. For machined shafts, rely instead on eddy-current for grinding cracks and roundness testers for geometry.
Document the NDT gate in your control plan before PPAP submission; customers audit the gate, not the part.
Environmental Footprint
Molten metal emits COâ‚‚, SOâ‚‚, and particulate matter. A 1-tonne iron pour releases 1.9 tonnes of COâ‚‚-equivalent when coke is used as charge carbon. Electric induction cuts this to 0.6 tonnes if the grid is 50 % renewable.
Chip-producing factories generate swarf that must be cleaned of coolant and hauled back to a melt shop, adding 0.1–0.2 tonnes CO₂ per tonne of chips for transport and re-melt. Near-net casting bypasses this loop.
Foundry sand can be thermally reclaimed and re-used 15–20 times, reducing landfill 70 %. Water-soluble coolants in machining plants must be treated for tramp oil and biocides before discharge.
Carbon Accounting
Ask your foundry for kWh per poured tonne and your machine shop for kWh per machined tonne. Multiply by local grid factor. You will discover that machining 316L bar stock can emit more COâ‚‚ than pouring the same part if the billet came from a coal-fired mini-mill.
Labor Skill Requirements
Foundry melt operators hold metallurgy degrees and read carbon equivalent charts like chefs read recipes. A single mis-tap of inoculant can drop tensile strength 15 %.
CNC programmers master CAM, tool-path optimization, and macro variables. One typo in G-code can crash a 400 k USD five-axis head.
Both trades suffer shortages, but foundry roles are harder to outsource because tacit knowledge of slag viscosity cannot be googled. Retention bonuses reach 20 % of salary.
Training ROI
Send your foundry team to 40-hour cupola safety courses; accident rates drop 35 % and insurance premiums fall 8 %. For factories, subsidize CAM certification; cycle time improves 12 % within six months.
Supply Chain Resilience
Foundries stock 30-day piles of scrap steel and pig iron in their yards, insulating them from mill outages. A CNC shop dependent on 7075-T651 bar receives weekly drops; one missed truck halts production.
Yet foundries rely on graphite electrodes and ferroalloy imports from China. A 20 % electrode price spike can add 25 USD per tonne of iron poured.
Dual-source both raw bar and foundry charge. Sign 90-day contracts with index-based pricing to share volatility risk.
Regional Density
Guangdong hosts 300 die-casters within 50 km, enabling same-day mold repair. Michigan’s I-75 corridor clusters gray-iron foundries next to engine plants, cutting freight to 40 km. Map your Tier-1 suppliers before you lock the BOM.
Cost Modeling Checklist
Build a three-column sheet: Foundry, Factory, Hybrid. Line items include raw material, tooling, cycle cost, NDT, freight, and working capital. Run Monte Carlo on volume scenarios 100, 1 k, 10 k, 100 k.
You will see that hybrid routes—cast then machine—win 60 % of the time for aluminum parts over 0.5 kg. Pure machining wins under 0.1 kg or when tolerance is tighter than CT5.
Update the model quarterly; aluminum ingot prices swing 8 % monthly, and wage inflation in industrial hubs tracks 5 % per year.
Hidden Cost Traps
Foundries quote FOB their dock; if your machining vendor is 800 km away, you pay 0.08 USD per kg per 100 km. A 10 kg casting can accrue 64 USD freight before it reaches the mill. Factor this in or negotiate delivered-price terms.
Decision Framework
Start with tolerance: if the drawing demands ±50 µm on mating faces, plan on machining regardless of origin, so favor a near-net casting to minimize machine time. If the spec is ±0.5 mm and surface is hidden, sand casting is king.
Next, weigh annual volume against tooling payback. Below 300 parts, machining from plate or bar avoids cap-ex. Between 300 and 5 k, sand casting dominates. Above 5 k, die casting or investment casting pays off unless wall thickness is below 1 mm.
Finally, audit your team’s core competence. If you lack metallurgists, partner with a foundry that offers design-for-casting reviews. If you lack five-axis capacity, choose a factory that can finish castings in-house.
One-page Scorecard
List eight criteria: tolerance, volume, alloy, geometry, finish, energy, lead time, and risk. Score 1–5 for each process. The highest total wins, but drop any criterion scored 1 to avoid disasters. This simple filter has saved OEMs weeks of indecision.