Scaling from Pilot Line to Gigafactory

Announcing a gigafactory is the easy part. The industry has seen a great many announcements, and a far smaller number of plants actually running at rate. The distance between the two is not mostly about money or intent. It is about the unglamorous work of taking a process that works at small scale and making it work, reliably and repeatedly, at large scale.

India shows this clearly, with commissioned cell capacity sitting well below what has been announced, a gap looked at in the piece on EV battery production trends in India. The pattern is not unique to India, and the reasons behind it are worth understanding for anyone planning a scale-up.

This article walks through the stages of scaling a battery operation, what each stage is actually for, and the handful of decisions that make the difference between a smooth ramp and a painful one.

Each Stage Has a Different Job

The most useful thing to understand about scale-up is that the stages are not smaller and larger versions of the same thing. Each one exists to answer a different question.

Stage	What it is really for	What success looks like	Common mistake
Lab and prototype	Proving the design works at all	Working samples, however they were made	Hand methods that no production line could ever copy
Pilot line	Learning the process and setting its parameters	Stable parameters and a clear list of what goes wrong	Treating it as a small factory and judging it on output
Industrialization line	Proving the process on production-representative equipment	Customer qualification and a yield trend heading the right way	Changing equipment type between here and volume
First volume line	Reaching rate and yield together	Output at target with defects under control	Chasing rate before the process is stable
Multi-line plant	Repeating what works, reliably	Each new line ramps faster than the last	Improving every new line differently instead of copying the proven one

The pilot line row is where most confusion starts. A pilot line is a learning tool. Its job is to find out how the process behaves, where it breaks and what the parameters need to be. Judging it by how many units it produces misses the point entirely, and pushing it to produce volume usually means it stops teaching you anything.

What a Pilot Line Should Actually Teach You

A pilot line earns its cost if, by the end, you can answer a specific set of questions with confidence.

• Which parameters matter most, and how tightly do they need to be held?
• What goes wrong, how often, and what causes it?
• How sensitive is the process to variation in incoming materials?
• What does a good unit look like in the data, and what does a bad one look like?
• Which checks actually catch problems, and which just add cycle time?

If the pilot phase ends and those answers are vague, the problems it should have found will simply turn up later, on a volume line, where they cost far more to fix.

The Equipment Trap

Here is the single most common and most expensive mistake in battery scale-up. The pilot line is built with whatever equipment is convenient, and the volume line is built with different equipment chosen for speed. Then everyone discovers that the carefully developed parameters do not carry over.

The reason is simple. A welding parameter set developed on one laser type does not transfer cleanly to a different laser type. A test limit set on one measurement method does not mean the same thing on another. The process knowledge gained on the pilot line is tied to the technology it was gained on.

The fix is to keep the process-relevant technology the same from pilot to volume, even if throughput differs enormously. The pilot welder can be slower and simpler than the volume welder, but it should use the same joining method. The pilot tester can handle one cell at a time instead of many, but it should measure the same way. That way what you learn on the small line still applies on the big one.

Customer Qualification Locks the Process In

Automotive customers in particular qualify a specific process, not just a product. They want samples from production-representative equipment, and once they have approved a process, changing it can mean going through qualification again.

That has a direct consequence for scale-up planning. The industrialization stage has to use equipment and methods that genuinely represent volume production, because that is what gets approved. Qualifying on a pilot setup and then switching to something different for volume can put a program months behind at exactly the moment it was meant to be launching.

Scaling Up and Scaling Out Are Different Problems

Battery operations grow capacity in two quite different ways, and mixing them up leads to the wrong expectations.

	Scaling up	Scaling out
How capacity grows	Bigger equipment doing the same job faster or wider	More copies of equipment that already works
Typical in	Cell manufacturing: wider coaters, longer ovens, larger formation	Pack assembly: extra lines, parallel stations at the bottleneck
Process risk	High; each size step is partly a new process	Lower; the process is already proven at the size being copied
What transfers	Some learning, but parameters often need re-establishing	Almost everything, if the copy really is a copy
Where it goes wrong	Assuming a parameter that worked small will work large	Letting each new line drift from the original design

Cell manufacturing mostly scales up. Coating lines get wider, drying ovens get longer and formation capacity multiplies, and each step up in size behaves partly like a new process. Parameters that worked at one width do not always work at the next. That is a big part of why cell plants take quarters to reach target yield, a subject covered in the guide to battery cell manufacturing.

Pack assembly mostly scales out. Once a pack line works, capacity grows by adding another line, or by adding a parallel station at whatever step is holding the line back. The process being copied is already proven at the size being copied, so the risk is much lower.

That difference matters for planning. A pack assembly operation can grow in steps with relatively predictable results. A cell operation should expect each size increase to bring its own learning curve.

Copy What Works, Exactly

When a second or third line is added, there is a strong temptation to improve it. The first line taught everyone lessons, and it feels wasteful not to build those in.

Resisting that temptation is often the better choice. A new line built as an exact copy of a proven one ramps quickly, because nearly every question has already been answered. A new line that has been improved in a dozen small ways is partly a new line, and it brings a partial learning curve with it.

The disciplined approach is to collect improvements, prove them properly, and then roll them out across all lines together, rather than letting each new line become its own slightly different experiment. Plants that do this find each additional line ramps faster than the last. Plants that do not find they are running several different processes that happen to make the same product.

Yield and Rate Have to Rise Together

During a ramp there is pressure to show output, and output is easier to push than yield. A line can be run faster than its process is ready for, producing more units and more scrap at the same time.

That trade rarely pays. Scrap in battery manufacturing is expensive, and in pack assembly a scrapped unit usually contains a lot of value by the time the problem is found. Worse, running fast before the process is stable makes problems harder to diagnose, because too much is changing at once.

The more reliable pattern is to stabilize the process first, then raise rate in steps, checking that yield holds at each step before moving on. It feels slower and it usually gets to full output sooner.

Design the Line for the Ramp, Not Just the End Point

A line specified only for its final output spends its first year expensive and underused. Designing it to grow in stages avoids that, and the principles behind it, takt, balancing and where to leave room for parallel stations, are set out in the guide to battery production lines.

In practice that means leaving floor space where a parallel station is likely to be needed, allowing control capacity for equipment that is not installed yet, and planning which steps start semi-automatic and move to full automation as volume grows. The one thing that cannot be phased in later is data capture. A line that starts recording in its second year has no record of its first, and the first year is when most of the learning happens.

Infrastructure Scales Too

It is easy to plan the production equipment carefully and forget that the building has to keep up. Compressed air, power, cooling, cell storage and fire protection all have to grow with the lines, and some of them cannot be extended easily once the plant is running. Those facility-level decisions are covered in the article on battery plant automation.

The common pattern is a first line that runs well and a second line that underperforms for reasons nobody initially connects to the building: an air system at its limit, storage that has run out of space, or a utility that was sized exactly for one line with no margin for two.

People Scale Slowest

Equipment can be bought and installed on a schedule. Skilled people cannot. Process engineers who understand why a weld behaves the way it does, maintenance teams who can keep an automated line running, and operators who recognize when something looks wrong all take time to develop.

Many scale-ups find that the real constraint on the second and third lines is not equipment or money but experienced people to run them. Building that capability early, often by having future teams work on the pilot and industrialization lines, is one of the better investments a scaling operation can make.

Where Cybernetik Fits

Cybernetik builds battery pack assembly automation rather than cell manufacturing equipment, so its work sits on the scale-out side of the picture described above.

Cybernetik capability relevant to scale-up	Specification
Line structure	Three zones covering cell to module, pack assembly and end of line testing
Line rate	Up to 6,000 cells per hour
Cell formats	Cylindrical 18650, 21700 and 32140, plus prismatic, pouch and blade
Cell diameters	32, 33, 35, 40, 42 and 46 mm
Module flexibility	Assembly lines adaptable to multiple module configurations
Joining	Laser and resistance welding with inline weld integrity testing on every joint
Verification	OCV, IR and ACIR on every cell, insulation detection and polarity checking
Traceability	Barcode and RFID capture with MES connectivity and process logging
Validation	Factory acceptance testing before dispatch
Delivery model	Design, build, installation, commissioning and support from one engineering team

Several parts of that specification matter specifically during scale-up. Handling several cell formats and multiple module configurations helps while designs are still settling, which is exactly the situation early in a program. Using the same joining and testing technology across the line means parameters learned early still apply later. And factory acceptance testing before dispatch means each new line is proven at works rather than debugged on the plant floor during a ramp that is already under pressure.

Cybernetik has been building automation for more than three decades. It is headquartered in Pune with facilities in Gujarat and Raigad and offices in the United States and UAE, and has installed over 6,000 systems in more than 30 countries, including over 400 custom robotic solutions. Battery work has been delivered for manufacturers including Hero MotoCorp, TVS Motor, Livguard and Matter. More background is on the Cybernetik about page.

Why manufacturers choose Cybernetik

Frequently asked questions