In-house or outsourced busbar fabrication: working the numbers

Most make-or-buy comparisons for busbar are written by someone with a machine to sell. This comparison is written to be run against a specific set of numbers, and it includes the conditions under which the answer is no.

A single arithmetic fact reframes the whole question. At mid-2026 copper prices, the metal in an outsourced busbar is roughly 95% of the price paid for it. A 1,200 mm length of 100 × 10 mm copper weighs 10.7 kg; at a delivered bar price of US$16,500 per tonne, the metal value is US$177. The conversion charge on top of that metal value is a few dollars. The decision is far less about machine efficiency than most vendor material suggests, and far more about whether copper can be purchased as well as a subcontractor can.

The two cost stacks

What outsourcing actually costs

The quoted piece price is the visible part. The rest:

• Freight, both ways: copper is dense. A thousand 1.2 m bars in 100 × 10 is 10.7 tonnes, a full truck. Inbound freight is a cost per part, and outbound freight on a rejected batch is worse.
• Minimum order quantity or value: this is the one that hurts on prototypes, spares and single replacement bars. A US$400 minimum order on a part where only one is needed is a 400-dollar part.
• Lead time, and the inventory it forces: a ten working-day lead time means safety stock equal to demand over ten days plus a buffer, in copper, on the balance sheet.
• Rework turnaround: a wrong hole means the part goes back and comes again: two lead times, plus the argument about whose drawing was wrong.
• Engineering change cost: revised drawing, requote, new batch, and write-off of the batch already produced. In switchgear the busbar is often the last thing frozen and the first thing changed.
• Metal price risk: most subcontractors quote with a copper surcharge, so metal risk is carried either way. The escalation clause should be read before assuming otherwise.
• Goods-in inspection and buyer time: small per part, not small per year.

What in-house actually costs

Worked with explicit assumptions, all of which should be replaced:

The capital charge is an annuity, not straight-line depreciation: at 8% over 7 years the factor is 0.192, so US$150,000 of installed cost is US$28,810 a year. Use the actual cost of capital; at 12% the factor becomes 0.219 and the charge US$32,900.

The floor-space figure should be checked against a machine footprint rather than a guess. An EMAC-BP-60 occupies 9,585 × 3,454 mm, which is 33 m² of machine. Adding bar infeed, part outfeed and aisle access makes 70 m² of served floor realistic. If the building is full, the true cost is not US$120/m² but the cost of the thing displaced.

Two items people expect to matter and do not:

Energy. A 15.7 kW connected load at a 30% duty cycle over 1,800 hours is about 8,500 kWh, or US$1,300 a year at US$0.15/kWh. Under 2% of the fixed cost.

Tooling. Punch and die consumption on copper is modest compared with steel; budget it per hole rather than per year, and see the tooling cost per hole working for how to build that figure. At around US$0.18 per part for a six-hole bar it is a rounding error against the metal.

Scrap is not in the fixed table because it belongs to the metal, and it applies to both options. It is not small: at 5% scrap on issued bar, a 10.7 kg part carries about US$2.10 of unrecoverable copper loss. Whether that counts for or against in-sourcing depends entirely on whose nesting is better, which is covered below and worked in detail in the nesting and scrap cost model.

Break-even, with every assumption exposed

The comparison is defined on a representative part: 1,200 mm of 100 × 10 mm Cu-ETP, six punched holes, sheared to length, no bend. Finished mass 10.7 kg.

Assumptions:

1. Delivered bare bar price: US$16,500/t (LME cash around US$14,500/t in mid-August 2026 plus an assumed US$2,000/t conversion premium). Both figures move; substitution is required.
2. Subcontractor's conversion charge plus margin: US$10.73 per part. This is an assumption. A quote broken into metal and conversion should be requested, and if it cannot be broken out, the same part can be quoted in two sections and solved for.
3. Inbound freight allocated: US$0.50 per part.
4. In-house tooling: US$0.18 per part.
5. In-house fixed cost: US$78,610 per year, from the table above.
6. The subcontractor buys copper better by Δ dollars per tonne. This is the variable that decides the answer.

Break-even volume N is fixed cost divided by the saving per part:

N = 78,610 / (conversion charge + freight + 0.0107 × Δ − tooling)

In tonnes of finished copper, the band runs from about 46 t/year at the favourable corner to about 410 t/year at the unfavourable one, with the base case landing near 125 t/year. One number to carry: on these assumptions the break-even sits somewhere between 50 and 400 tonnes of finished busbar a year, and where the fall inside that band is decided more by copper buying power than by part count.

The "never" cell is not a typo. If the subcontractor buys metal US$800/t better and charges only six dollars to convert it, no volume of simple punched bar justifies a machine. That situation is common for small panel builders buying from a fabricator who also serves the wire and cable trade.

Capacity, for context

At 90 seconds per part including bar handling, one shift of 1,800 hours is about 72,000 parts a year. The break-evens above sit between 6% and 55% of single-shift capacity. Break-even is reached long before the machine is busy, which means the machine is rarely justified by throughput. It is justified by everything in the next section.

What the sensitivity actually tells you

Complexity in-sources before volume does. The right-hand column shows that a simple part needs 38,500 pieces a year to justify the machine; a punched-and-bent part needs 5,600. Subcontract pricing rises steeply with operation count because setup and handling dominate, while marginal cost barely moves once the machine is running. The result holds: complex, high-mix, low-volume work in-sources first, and long runs of plain punched bar in-source last.

Yield is a two-way argument. A subcontractor nesting across many customers has a broader length pool and may run one to two points better on scrap. At 10.7 kg per part and a net loss of about US$3,740 per scrapped tonne, two points of yield is US$0.80 per part, a meaningful fraction of the conversion charge. On the other hand, they nest to their own yield, not yours, and payment is by finished piece.

Mix fragmentation breaks the cycle-time assumption. If parts run across eight sections in small quantities, in-house setup time per batch climbs and the 90-second average fails. The fixed table should be rebuilt with the operator hours actually obtained, not the ones the brochure implies.

Bending changes the machine, not just the number. If a meaningful share of parts needs bends, the single-machine assumption breaks. Either add a dedicated bender such as the EMAC-BB-H12, which raises fixed cost and capacity together, or use an all-in-one such as the SMART-603CNC-S, which punches, shears and bends in one footprint at lower capital and lower throughput. The sections actually run should be checked against the machine's capability with the bending force calculator before assuming one machine covers the range.

The variables that usually decide it

The decision is made on three things the spreadsheet handles badly.

Engineering change lead time

From the moment a drawing is revised, how many days pass before the correct bar is on the assembly bench? In-house, hours to a day. Outsourced, typically five to twenty working days including requote, plus the write-off of the superseded batch.

A typical project carries three busbar changes and each costs a fortnight of float, the machine is being justified on schedule risk, not on piece price. A number should be put on a week of programme slip on the largest contract and compared to US$78,610 a year. For most switchgear builders that comparison is not close.

Rework turnaround

The same argument applies to internal mistakes and external ones. In-house, a mis-punched bar is recut before lunch. Outsourced, it is two transit legs and a commercial conversation. Frequency matters more than cost: the number of rework returns raised last year should be counted and multiplied by the days lost.

Intellectual property exposure

Busbar layout encodes thermal design, short-circuit bracing and enclosure strategy. Handing the full 3D copper set to a subcontractor exposes all of it, and the test is concrete: does the subcontractor, or anyone in their group, build panels? If yes, the design is being sent to a competitor's supply chain. That may still be acceptable, but it should be a decision rather than an oversight.

Verification responsibility does not transfer

Under IEC 61439, verification of the assembly is the original manufacturer's and assembly manufacturer's responsibility, by testing, by calculation, or by comparison with a verified reference design. Temperature-rise verification in particular depends on bar section, spacing and joint quality: the standard's limits allow 105 K for bare copper busbars and 70 K at terminals for connection of external conductors, and those numbers are only meaningful against the geometry verified.

If a subcontractor substitutes a section, changes a hole pattern or alters a joint to suit their tooling, the verification basis has moved and the consequence is owned at type test and in the field. Outsourcing punching and cutting does not outsource that. The drawing and the material specification must be controlled in-house, and incoming inspection must be able to detect a substitution.

The honest case for staying outsourced

• No capital, no floor, no maintenance, no recruitment. The fixed table above is US$78,610 a year that a subcontracting shop does not spend.
• Their metal buying power is usually better, and at current copper prices that single factor can outweigh everything else, as the break-even table shows.
• They absorb peaks. If demand is three large projects a year, a machine sits idle for eight months and the capital charge, the floor and most of the labour are still paid.
• Their tooling library already exists. Every unusual hole size or slot is a tool that would otherwise be bought.
• Process knowledge is already paid for. The first six months in-house will be worse than the subcontractor's tenth year.

If the annual conversion-equivalent volume is under roughly 15,000 simple parts, nobody in the building will own the process, the subcontractor is close and turns work in three days, and the floor is already full, then the answer is not yet. Coming back in two years with twelve months of clean data is a better outcome than buying a machine that runs at 8% utilisation.

What to measure before deciding

Four datasets, all of which are already held:

1. Twelve months of subcontract invoices, split into metal content and conversion charge. This is the single most valuable number and almost never extracted.
2. Engineering changes and their cost. Count them, count the days, count the written-off batches.
3. Rework returns. Count, cause, days lost.
4. The actual part mix, by section, by operation count, by quantity per order. The operation count drives the break-even harder than the quantity does.

With those four the table above can be rebuilt in an afternoon with specific numbers rather than assumed ones, and the answer will be defensible in front of a board.

The hybrid answer is often the right one: prototypes, spares, engineering changes and short runs are brought in-house where responsiveness is worth money, and the long uniform runs are kept outside where someone else's metal buying and yield are better. That splits the volume in a way that makes the machine's fixed cost easier to carry and puts it on exactly the work where its lead-time advantage is worth the most.