Get quotes from five suppliers for “an ultrasonic welding machine” and the numbers can land four times apart. The phrase covers too much ground to price. Below is the model we use in-house when we cost a job, so you can run your own numbers through it.
Why one number can’t work
Three things move the number more than anything else. None of them appear on a spec sheet.
Start with the acoustic stack. A 20 kHz stack for thick copper busbar and a 40 kHz stack for a 0.5 mm² splice are different products, not the same one with a different label. On some jobs the stack alone is a third of the machine price.
Then everything around the weld. A bench unit with a hand clamp looks much like a station with a servo press, part transfer and in-line resistance monitoring. They don’t cost anything alike.
Volume matters too. One machine for a trial line and twenty with a documented process qualification aren’t the same conversation.
So skip the number and do the arithmetic. Feed it your own figures, and the quotes coming back become comparable. You’ll also see which supplier is pricing a machine and which is pricing a project.
Where the money goes
A welding budget splits into four buckets. Suppliers quote you the first one. The other three show up later, which is why year two feels expensive.
Equipment is the generator, transducer, booster, horn, press frame, fixture and controls. It’s capital, amortised over the machine’s life.
Consumables are horns, anvils, and eventually transducers. This bucket grows with the number of welds you make.
Operating cost is electricity, compressed air, the operator, floor space. It’s driven by cycle time.
Quality is scrap, rework, line stoppage, warranty claims. On a three-shift harness line it usually outweighs the other three combined over five years.
The invoice shows bucket one. The other three decide whether the project was a good idea.
How to price one joint
Compare cost per acceptable joint: one weld that passes inspection, everything included.
Equipment: purchase price divided by joints over the machine’s life. Consumables: horn cost divided by joints per horn, plus the same for the anvil. Operating: hourly rate plus power and air, times cycle time. Quality: scrap rate times the cost of the scrapped part, plus rework.
Run those four lines with real numbers and it looks like this:
| Item | Example input | Cost per joint |
|---|---|---|
| Equipment | USD 30,000 over 5,000,000 joints | USD 0.0060 |
| Horn | USD 400 per horn, 200,000 joints per horn | USD 0.0020 |
| Operator + power + air | USD 9/hour, 3.0 s cycle | USD 0.0075 |
| Scrap and rework | 0.5% at USD 0.50 per part | USD 0.0025 |
| Total | USD 0.018 per joint |
Equipment is the smallest line in that table. Buying the cheapest machine is usually the most expensive one you’ll make.
That leaves quality. Pushing scrap from 0.5% down to 0.1% is worth more per joint than arguing 15% off the machine price. It’s also the hardest line to move, because it means fixing the process, not the purchase order.
Run the same table with a 20-second cycle instead of 3 and it flips. At long cycle times the operator dominates and automation pays for itself fast. That’s why “is ultrasonic welding worth it” has no answer in general. It depends which line is biggest in your process, and you’re the only one with those numbers.
Where it leaks
On lines that are already running, we see the same four losses. None of them is the purchase price.
Most plants change horns on a calendar rather than on a measurement. Somebody picked “every six months” years ago and nobody’s revisited it since. Some horns come out while still good. Others stay in until the amplitude has dropped and welds start drifting. Both directions cost money. There’s a separate guide on how to tell when a horn is done.
Stoppage next. On a battery pack or harness line, one failed splice stops everything downstream. A few minutes of that costs two or three orders of magnitude more than the weld. If you’re deciding whether to pay for in-line monitoring, put this number next to it rather than the weld cost.
Welding outside the window without knowing is harder to spot. If nobody’s characterised the amplitude, pressure and time combination, the line runs on a setting that “works” until a batch of material shifts things. We published measured data for 70 mm² Cu-Al splices that shows how narrow this gets. At 55 µm the pull-off force is 5% better than at 45 µm, while contact resistance is 46% worse. Five percent more strength costs 46% more resistance, and that trade rarely pays.
The fourth loss is buying capacity you can’t fill. A 12 kW machine running at 20% utilisation costs more per joint than a 6 kW machine at 70%. Match the machine to the line rather than to the brochure.
Against crimping and resistance welding
The comparison that matters is against whatever you’re running now: crimping, resistance welding, or brazing. Purchase prices aren’t comparable. Cost per acceptable joint is.
| Ultrasonic | Crimping | Resistance / brazing | |
|---|---|---|---|
| Equipment | High | Low | Medium |
| Consumables | Horn and anvil only | A terminal on every joint | Electrodes, filler, flux |
| Joint resistance | Low (metallurgical bond) | Higher, depends on crimp quality | Low to medium |
| Heat affected zone | None, solid state | None | Significant |
| Cu-Al dissimilar | Yes | Only with plated terminals | Difficult |
Crimping wins on capital and loses on consumables. Ultrasonic loses on capital and wins once volume is high enough for the per-joint saving to pay the machine back. Where that crossover sits is arithmetic, not opinion.
Three days, your parts
The configuration questions collapse the moment you weld your own parts. Skip the price request and send the parts. What you want back is a weld report.
Send the materials, cross-sections and plating you run in production, as parts rather than datasheets. Tell us the cycle time, how many joints per shift, and what acceptance criterion your customer applies: pull-off force, resistance, cross-section, or visual.
What comes back is a trial weld with measured pull-off force and contact resistance, the parameter set that produced it, and photos of the weld zone. Usually inside three working days of the samples arriving.
With that report, the four buckets turn into real numbers and the quotes you receive become comparable. If the trial doesn’t work, you’ve lost three days rather than a purchase order.
Send us your parts. Or read the quality control and selection guide first, if you’re still working out what your acceptance criteria should be.
FAQ
How much does an ultrasonic metal welding machine cost?
There isn’t one figure. The acoustic stack, the automation level and the validation scope each move it independently. The question worth asking is what your cost per acceptable joint will be. Put your figures through the arithmetic above, then ask any supplier to justify their number against yours.
Is ultrasonic welding more expensive than crimping?
On purchase price, yes. A crimp tool costs a fraction of a welder. On cost per joint it depends on volume: crimping eats a terminal on every joint, ultrasonic eats only horn wear. Past a certain annual joint count the welder pays back its capital, and you can work out where that point sits from your own cycle time and material cost.
What’s the biggest hidden cost?
Line stoppage from one bad weld. On an automotive harness or battery pack line, a few minutes down costs orders of magnitude more than the weld. That’s why a characterised process window and in-line monitoring matter more to total cost than the machine price does.
Can you quote without a sample?
We can suggest a configuration range. We can’t give you a meaningful price. Weldability depends on alloy, cross-section, plating and how oxidised the surface is, so two parts that look identical on a drawing can want different amplitudes. A trial weld on real parts runs about three working days and removes most of the guesswork from the quote.