The filament price shown by a slicer is not the cost of a finished part. At best it is one easy-to-measure subtotal: it excludes tied-up capital, operator work, support and purge losses, finishing, inspection, failed builds and workshop infrastructure. This guide builds a total-cost model usable for FDM, photopolymer, powder-bed and metal processes. Its aim is not a universal cost per gram. It is a traceable calculation in which every assumption is visible, can be updated from operations and can be compared with a supplier quotation under the same acceptance requirements.

Define an accepted part first The denominator is not the number of objects printed but the number accepted. Specify revision, material, quantity, dimensional tolerance, surface requirement, load, documentation and deadline. One prototype may be suitable for a fit check yet unsuitable for end use. Without a quality threshold, internal production can look cheap because inspection is lenient, while a bureau quote looks expensive because it includes a measurement report and traceable material. Cost comparisons become meaningful only when both routes are priced against one technical specification.

The complete equation Cost per accepted part can be written as allocated ownership machine time plus direct material and consumables, active labour, post-processing, quality assurance, facilities and software, expected scrap, and logistics. A batch also carries order or changeover cost divided among good units. Keep cash expenditure separate from economic cost: a fully paid machine still consumes capacity, maintenance and useful life. Add detail where the decision is sensitive, but do not demand more data than the workshop can record consistently. A model that is routinely maintained is more valuable than a theoretically perfect spreadsheet abandoned after one quotation.

Capital means more than list price The acquisition base includes the printer, freight, duties, installation, training, required workstation and every accessory necessary for production. FDM may require drying, extraction or wear-resistant nozzles; resin needs washing, curing and safety equipment; SLS needs powder handling; metal may require inert gas, part removal and heat treatment. Allocate shared optional equipment only by its actual use. Do not mix recoverable tax into a net investment, while non-recoverable tax remains a real cost. Record currency and exchange-rate date, because a seemingly precise international comparison can otherwise be moved more by currency than by process improvement.

Useful life, residual value and depreciation Subtract a cautious residual value from the capital base and divide by planned productive machine hours, not merely calendar years. A five-year machine running five hundred productive hours annually has a very different rate from the same system running two thousand. Accounting depreciation can differ from economic depreciation used for decisions. Technology obsolescence, spare-part support, dependency on a cloud service and resale demand may shorten useful life. Build optimistic, base and conservative cases. A transparent range is more informative than false precision, particularly for a new platform whose long-term reliability and secondary-market value have not yet been observed.

Capital is not free A financed purchase carries interest or lease cost. Even internal cash has opportunity cost because it could fund inventory, inspection equipment or another production constraint. A simple model can apply an annual cost of capital to average net book value; a larger investment may justify discounted cash flow and net present value. Avoid double counting: if a lease payment already contains financing and capital recovery, do not add full depreciation and interest again. Name the method and keep it consistent across alternatives. The purpose is not to make ownership look expensive, but to compare scarce capital honestly.

Productive capacity and utilisation The 8,760 hours in a year are not saleable machine time. Remove planned maintenance, calibration, material changes, cleaning, breakdowns, idle weekends and periods without suitable work or staff. Separate technical availability from demand utilisation because they require different remedies. The denominator of the hourly rate should be realistic productive time. For an internal development machine, value may come from faster iteration or avoided supplier delay rather than revenue. That does not make capacity free; it produces a different benefit. Track scheduled, available, occupied and accepted-output hours separately so a high busy rate cannot hide poor yield.

Building a machine-hour rate Add annual depreciation or lease, cost of capital, service contracts, planned maintenance, insurance, dedicated software and an appropriate fixed facility share, then divide by productive hours. Variable energy and material can remain separate. Do not automatically assign one rate to every printer: acquisition value, reliability, calibration burden and wear differ. Time reserved by a job can include warm-up, automatic calibration, cool-down and mandatory cleaning when the machine cannot accept another task. If preparation occurs off-machine, avoid charging both full engineering labour and occupied equipment time unless both resources are genuinely unavailable.

Net and gross material Finished-part mass is only net material. Gross consumption includes supports, skirts, prime towers, purge, test strips, nozzle flushing, resin left in handling, unrecoverable powder and witness coupons. Where practical, measure stock issued minus properly recovered material instead of copying a generic loss factor. Filament discarded through moisture, expired resin and contaminated or oxidised powder are inventory losses. Landed material price can include freight, duty and certification documents. Track colour and lot separately when purchase price or quality risk differs. A cheap input is not cheap if it adds drying, supervision or failures.

Multimaterial FDM and purge With colour or material changes, purge can exceed model mass, especially for small objects with many transitions. Compare slicer estimates with actual spool consumption and identify support, tower and purge separately. Placing several identical parts on one plate can reduce change waste per unit, but one failure can then affect more output. Low-cost filament may increase total cost through diameter variation, moisture or poor winding. Evaluate accepted yield rather than price per kilogram. If a marketing claim promises zero waste, verify what boundary it uses: waste may have moved from a tower into an internal purge path without disappearing from material consumption.

Resin, SLS and metal use different material logic For photopolymers, include resin retained in the vat and filters, wash-fluid contamination, film and tank wear, and supports. For SLS, refresh ratio, packing density and cool-down govern economics; powder filling the build is not necessarily consumed, but it must be handled and may degrade. Metal requires more than powder or wire: build plates, inert gas, filtration, separation, heat treatment and possibly hot isostatic pressing can be necessary. A single price per kilogram cannot compare these technologies fairly. Use the entire validated route from incoming material to accepted component.

Put consumables on their own line Nozzles, build plates, adhesive, wipes, gloves, filters, vat film, wash fluid, abrasive media, protective sheets and packaging are individually small but collectively material. Allocate life-based items across relevant hours or accepted parts; assign event-based items directly. Inventory found in a cupboard is not economically free merely because it was purchased earlier. Consumption logs also reveal abnormal wear, such as rapid nozzle enlargement with abrasive filament or repeated damage to vat film. A realistic standard rate should be reconciled periodically with purchases and physical inventory rather than remaining unchanged for years.

Energy: measure instead of reading the nameplate Power-supply rating is not average consumption. Measure warm-up, stable printing, cooling, washing, curing, extraction, compressed air and drying separately. Multiply kilowatt-hours by the organisation's real tariff and network charges. Prusa's support data likewise shows that measured average consumption depends on environment and settings, so another machine's example is not your operational evidence. Energy is often secondary to labour for a small FDM part, but it can matter greatly for an industrial heated chamber, SLS thermal cycle or prolonged heat treatment. Report the measurement boundary and whether shared HVAC is included.

Labour: touch time versus elapsed time A twenty-hour automated build is not twenty labour hours, but it is not zero. Measure quoting, file repair, slicing, machine preparation, material loading, launch, checks, removal, cleaning, support removal, finishing, inspection and packing. Burden direct wages with employment costs, leave, training and a reasonable overhead allocation. If one operator supervises multiple machines, distribute observed time using a consistent rule rather than charging every machine fully. Automation creates value primarily when it reduces active minutes and error rate. A faster headline print time may save little if setup and finishing dominate the job.

Preparation and engineering CAD repair, orientation, support design, parameter selection, nesting, simulation and production documentation are skilled work, not machine time. For a repeat validated job, initial setup can be amortised over expected volume; for a one-off, the first part carries it. Decide when a revision requires revalidation. Free slicing software does not make its use free: profile maintenance, update checks and file management consume time. Misallocating preparation most strongly distorts prototype-versus-series comparisons because the first part normally costs far more than the next. Preserve the validated build package so repeat orders genuinely avoid work.

The whole post-processing route A part is not complete when the printer stops. FDM may require support removal, deburring, sanding, vapour finishing, bonding or inserts. Resin needs draining, washing, drying, support removal and specified post-curing. Powder-bed and metal routes can require depowdering, separation, heat treatment, blasting and machining. Assign active labour, equipment time, consumables, yield and queue time to every step. If a subcontractor performs an operation, include freight, coordination and rejection risk alongside the invoice. A process that prints rapidly can still have high unit cost when manual finishing becomes the bottleneck.

Inspection and traceability Visual checks, callipers, gauges, CMM, CT, test coupons and records should match application risk. Do not burden every concept model with an industrial inspection package, but do not price a critical component by visual approval alone. Define sampling, instrument calibration, record retention and nonconformance handling. Capturing lot, machine, profile and operator allows causes to be traced. Inspection is not only a cost: it can prevent late claims and reveal a slowly drifting process. Allocate first-article inspection separately when later units use a smaller sampling plan.

Scrap as an expected value If a job family succeeds first time in 90 percent of attempts, the expected direct cost of one good output is not one run; approximately it is one run divided by 0.9 when failures occur at similar cost. A better model distinguishes an early adhesion failure, a mid-build blockage and a dimensional defect discovered at final inspection. For batches, ask whether an event threatens one object or the whole build. Track yield by machine, material, profile and geometry class; a company-wide average is too broad. Use a conservative estimate for a new process until sufficient operational evidence exists.

The business effect of a reprint A failed part consumes more than material and machine time. It can occupy capacity reserved for the next order, require expedited shipping or overtime, trigger penalties and delay a development decision. A defect found after finishing or assembly is especially expensive. Risky jobs may justify schedule buffer or spare units, but expose that cost. A slower reliable profile can therefore outperform a record-speed setting economically. Manufacturer speed claims are not TCO evidence until stable quality and yield are validated on the relevant part family.

Maintenance and spare parts Planned maintenance covers lubrication, filter changes, optical cleaning, calibration and wear replacement; unplanned work includes failure, diagnosis and urgent repair. Include components, service labour, internal labour and lost capacity. Warranty may remove the part invoice but not waiting time or missed output. For a critical machine, locally stocked spares may cost less than several days of downtime. Never leave maintenance hours in the productive denominator while also treating them as saleable. Reconcile model parameters with actual service tickets each quarter and separate recurring wear from exceptional damage.

Downtime and reserve capacity Availability has context-specific value. A single development printer can delay a project; a farm may move the job to another machine but disrupt the schedule. Redundancy is expensive and acts as insurance. Expected downtime cost can use failure frequency, mean time to repair and contribution lost per hour. Do not confuse lost revenue with lost contribution. A service bureau used as emergency capacity becomes much faster to activate when specification and test parts were agreed in advance. The correct resilience choice may be a second internal machine, spare modules, or a qualified external route.

Software, cloud and data handling Licensed CAD, build processors, simulation, MES, remote monitoring and storage may be machine- or user-based. Free systems still need integration, backup, permissions and support. If a printer depends on a cloud service, assess outage, pricing and export risk. Confidential designs also create compliance and cybersecurity work. Allocate shared software through an appropriate driver such as active users or machine hours. Use normal planned-period pricing rather than a temporary trial discount. Record whether service cancellation affects machine operation, archived jobs or access to production data.

Space, extraction and safety Real space exceeds printer footprint because material storage, finishing, safety clearance, movement and waste handling are required. Allocate rent, electrical infrastructure, HVAC, extraction, fire protection, PPE, training and environmental handling proportionately. Safety data sheets and local rules determine requirements for resin, powder and solvents; lasers and metal AM add further hazards. Existing space is not costless when it has alternative use, but allocate only the portion actually caused or occupied. Avoid placing the entire building overhead on one printer simply because it is easy to divide that way.

Inventory and logistics Multiple colours, polymers, resins or powders tie up capital, occupy space and age. Inventory cost includes financing, storage, counting, moisture control, expiry and emergency purchases after stockouts. Add two-way freight, packaging, customs, minimum order value and uncertain lead time to external quotations. Internal production also requires receiving and finished-goods handling. Holding rarely used certified material may cost more than outsourcing, while readily available internal stock can create large time value for frequent iteration. Track slow-moving and obsolete material rather than hiding it in general overhead.

Batch and changeover cost Quoting, file checking, cleaning, material changes and first-article approval often occur once per order. Divide them across accepted batch units rather than multiplying automatically. Machine time is not always linear: more parts improve tray utilisation but extend a build, concentrate risk or increase cool-down. Create cost points for 1, 5, 10, 50 and 100 units. Economic batch size combines demand, inventory carrying, lead time and process risk; it is not simply the maximum number that fits on a build plate. Recalculate after design or material changes because the original validation may no longer apply.

An FDM example using your own inputs For a functional bracket, record net model mass, slicer gross material, measured occupied time and active operator minutes. Multiply gross grams by landed kilogram cost, then add machine hours, measured energy, nozzle and plate wear, preparation, removal, deburring and inspection. Divide the subtotal by observed yield and add each unit's share of the one-time order cost. Do not copy electricity or labour rates from an online example. The template supplies structure; your measured inputs make the decision credible. Retain both estimated and actual columns so quoting accuracy improves after completion.

A photopolymer example Alongside slicer volume, measure support and vat-handling losses. Add draining, staged washing, drying, support removal and material-specific post-curing. Allocate solvent use and disposal, gloves, filters, film, washing and curing equipment. Multiple parts on a tray may add little exposure time per layer, so batch use is crucial. Yet one detached object can damage other models or the film, creating shared risk and inspection work. Include the controlled time before a cleaned part may be handled or measured, not only light-on time.

SLS and metal require capacity models For powder-bed systems, begin with the economics of the complete build. Warm-up, layer cycles and cool-down make a small isolated object disproportionately expensive, while dense nesting spreads fixed cycle cost. Include nesting, refresh, depowdering and queue time. In metal AM, engineering, supports, plates, inert gas, stress relief, separation, machining and inspection can dominate machine exposure. NIST cost studies therefore examine the process chain and application context rather than deriving universal cost from one printer parameter. Validate allocation rules with the people performing downstream operations.

Make or buy Request the same material or documented equivalent, orientation and surface requirement, quantity, records, inspection and delivery date. Add shipping, procurement administration, waiting and change-loop cost to the external price. Put every ownership line and the learning curve into the internal case. A bureau benefits from utilisation, process variety and specialised staff; internal capacity offers fast iteration, data control and direct scheduling. A hybrid strategy often wins: keep frequent stable jobs inside and outsource rare, highly regulated or capital-intensive processes. Revisit the decision when volume or specification changes.

Break-even volume Break-even is not machine price divided by an outsourced unit quote. Each unit saves the difference between outsourced landed cost and internal variable plus risk cost; that contribution must recover internal fixed cost. If the difference is small or volume uncertain, break-even is remote. Avoided downtime from faster access may be added only as a specific, defensible benefit. Do not count work as savings if it was never purchased before. Forecast volume from order history and product plans, not theoretical machine capacity, and show what happens when expected demand arrives late.

Sensitivity analysis Change the most uncertain inputs one at a time: productive annual hours, yield, operator minutes, material price, service cost, residual value and external lead time. Show which ten-percent change moves unit cost most. Sensitivity directs measurement effort: if energy barely matters, do not over-engineer it; if yield dominates, prioritise failure codes and profile validation. Present optimistic, base and conservative scenarios and the decision threshold. A range is more honest than a value with many decimal places built on uncertain assumptions.

A 30-day measurement plan For every job, capture machine, material and lot, profile, net and gross material, occupied time, active labour, finishing, result and failure code. Measure electricity on representative cycles rather than every job. Keep maintenance and downtime logs. After thirty days, group results by geometry and technology class and calculate median and spread; neither a record run nor a disaster should automatically become the standard. Mark missing values as estimates. The system works when entry is quick, ownership is clear and quarterly review actually changes rates and quoting assumptions.

Facts, vendor claims and assumptions Verified facts are invoices, power-meter readings, work logs, acceptance records and documented service events. A nominal speed, promised automation saving or recommended consumable life remains a vendor claim until validated locally. Planned utilisation, residual value and future volume are business assumptions. Store these categories separately. A TCO model is trustworthy not because every number is certain but because uncertainty is visible and updateable, and marketing data is not silently treated as operational measurement. Cite source and date for every externally supplied parameter.

Decision checklist Before approving a quote or investment, confirm that both alternatives use the same specification; gross rather than net material is counted; all active labour is present; productive hours are realistic; finishing, inspection, scrap and downtime are included; depreciation and financing are not double counted; batch and one-time items are allocated correctly; external logistics are included; estimates are labelled; and the model has a review date. The best cost model is not the longest spreadsheet. It is the one that produces the same result from the same evidence and can be reconciled after the job.

Editorial conclusion The economic advantage of additive manufacturing rarely comes from the cheapest kilogram. Faster iteration, tool-less geometry, reduced inventory and avoided downtime can matter more, but only when a specific existing process is displaced. A grams-only quote underprices work, while indiscriminate corporate overhead can make internal production look uncompetitive. Use a complete but auditable model, collect local evidence, compare identical acceptance criteria and show uncertainty as a range. Then TCO becomes an operating decision tool instead of a promotional ROI number.

Containerised local production is a distinct business case

The planned TitanCell deployment in Timor-Leste shows why total cost does not stop at machine hours. A remote containerised LFAM system adds collection, sorting, washing, shredding, drying, blending, power, ventilation, fire protection, finishing, local training, remote support and spare-parts logistics to the printer. Missing any one of these can prevent nominal capacity from becoming an accepted product.

Do not assign a zero price to waste feedstock. The acquisition cost of recovered plastic includes cleaning, rejection and quality testing. Mixed material can create more scrap, slower builds or shorter service life. Those effects belong in cost-per-part scenarios.

Avoided cost and local value

Local production may create value through avoided imports, delay, large tooling and stock-outs rather than a cheaper kilogram. Compare the conventional product's total lead time, freight, duty, minimum order and consequence of downtime. Local employment, repairability and waste reduction can also have financial value, but require transparent methods rather than general sustainability labels.

A business pilot should cover at least three families: a simple non-critical item, a large low-load structure and an application with a major local supply advantage. Record feedstock batch, preparation loss, full cycle time, energy, scrap, finishing and performance in use.

Decision gates for a deployable factory

The first gate is feedstock: can local waste produce a stable formulation? The second is product: does it meet real load and life requirements? The third is operation: can the local team safely maintain and document the process? The fourth is market: are there paying customers at a volume that sustains the cell? Series capacity is credible only after all four gates pass.

A container is therefore a packaging and deployment format, not a cost model. A sound make-or-buy decision compares the cell's full life-cycle cost with the risk-adjusted cost of imports and local conventional manufacture. The apparently cheapest option loses when it cannot deliver consistently accepted parts.