What must a robotic cell prove?
Uptime versus the production calendar. Changeover time. Scrap and rework before and after — see OTD, scrap, and rework. Who programs it, and whether that person transfers. Whether the cell is in the quoted capacity or a demo in the aisle.
CNC useful life covers pallets and lights-out options on mills. Plastics files care about robots on the press and scientific-molding discipline. Electronics lines obsolete faster than either. Unused is unused.
Deloitte / Manufacturing Institute: as many as 3.8 million workers needed 2024–2033, 1.9 million potentially unfilled (Source: Deloitte / MI). Demonstrated automation is more interesting in that market. It still sits inside the current size-band × sub-sector cell. A sde-under-1m job shop with a pallet pool does not inherit 6.5x–9.0x medical EBITDA math.
BLS May 2025 operator and programmer medians remain the comparison for “we saved two heads.” Burden the package. Then show the heads actually left — or were redeployed to a bottleneck. A robot plus the same overtime is not substitution.
How should unused automation be treated?
As iron. Commission OLV if the dollars are material. Do not put crate-value into trailing EBITDA. Do not split the difference between a going-concern fantasy and an auction flyer. Equipment age and capex is the cash-chip page.
Fed G.17 utilization 76.0% (July 2026; Source: Fed G.17) means a lot of U.S. iron is not screaming. Idle automation in a 60% bay is a balance-sheet story. Capacity should say so.
What automation will not replace
Quoting. Closing. CAPA. Customer calls. Skilled labor still needs a deputy. Owner-only integration of a cell is owner dependence with a nicer photo.
ISM August 2026 PMI 54.6 (Source: ISM) does not re-rate a robot. BizBuySell’s 2.75x SDE median (2021–2025; Source: Iconic recap) already describes many shops that have a single cobot and a lot of manuals. Price the mix.
| Automation claim | Proof | Valuation treatment |
|---|---|---|
| Labor substitution | Headcount / overtime file | May support high end of cell |
| Scrap reduction | Before/after yield | May support high end of cell |
| Lights-out | Hours file vs calendar | Capacity story if real |
| In crate / unused | PO and location | Capex / OLV, not earnings |
| AI / software slide | Pulse: 67% no material effect | Not a new multiple |
How buyers should diligence the cell
Watch a changeover. Export alarms. Ask for the integrator’s remaining support. Confirm safety and guarding. Confirm the program library is backed up. Confirm the cell is on the insurance schedule and the machine list.
Sellers who want credit for automation should put the before/after in the CIM, not the adjective “automated.” Buyers: buy a manufacturing business. Sellers: sell my manufacturing business. The valuation pillar keeps the cells.
How to diligence a cobot and a full cell without the same checklist
A cobot tending a lathe needs safety, program library, changeover time, and whether it actually runs second shift. A six-axis welding cell needs fixturing ownership, seam tracking, and a named programmer. An SMT line’s robots are the line — obsolescence is the story. Do not use one “robotics” adjective for all three.
Ask for the integrator’s remaining hours and the spare-parts list. An orphan cell with no support is a remaining-life problem. Maintenance should include the integrator invoices.
Before-and-after exhibits that can support the high end of a cell
Headcount on that operation, overtime on that operation, scrap on that operation, and OTD on those part numbers — twelve months before and twelve months after go-live. If the seller cannot produce that, the automation is décor. OTD and scrap is the metric page. This page is the causal claim.
IBBA’s 67% “no material AI effect” finding is the tone for software slides. A scheduling tool can help. It is not a new official multiple. Do not staple it to a robot photo and call it Industry 4.0 value.
When automation should push you toward OLV instead of a growth story
If the cell has never run in the quoted capacity, if the book-to-bill is under 1.0, and if utilization on the rest of the bay is already low, you may be looking at unused iron. Capacity plus asset versus earnings is the test. A crate does not get a growth multiple and an OLV bid on the same day without someone noticing.
CNC pallet pools that actually run lights-out can support the high end of the CNC cells — 2.0x–3.6x SDE under $1M, 5.0x–6.5x at $1–3M EBITDA for precision machining — if the hours file agrees. They do not borrow the medical row.
A cell-by-cell memo a searcher can take to credit committee
Write one page per automated cell. Name the operation, the year it went live, the integrator, the uptime last twelve months, the changeover time, the scrap before and after, the people it actually removed or redeployed, and whether it is in the quoted capacity. If you cannot fill those lines, the cell is not yet a valuation input. It is a photo.
Credit committees have learned to fear crates. They have also learned to fear cells that run only when the owner is on site. A cobot that the owner starts every morning is owner dependence with a nicer payload. Put the start-up ritual in the memo. Put the backup operator in the memo. If there is no backup, say so.
Compare the memo to the insurance schedule and the machine list. Orphan cells that are not on the list will be found. Cells that are on the list and never warm will be found. Cells that are warm and not on the capacity claim will be found. The memo is how you look like an adult before those three finds.
Then decide the stack. Demonstrated substitution can support the high end of the current size-band and sub-sector cell. Unused iron goes to OLV. Do not put the same robot in both stacks at full credit. IBBA Pulse Q1 2026 still sits in the background: most advisors did not see material AI effects on value. A scheduling overlay does not change that sentence. Keep software in the operations narrative. Keep iron in the iron narrative.
Jason Taken is a business broker, not an integrator. Educational information only. No value is guaranteed.
A crate-versus-cell decision tree that keeps stacks from mixing
Has the cell run in the quoted capacity for twelve months. Yes or no. If no, it is iron. Commission OLV if the dollars are material. Do not put crate-value into trailing EBITDA.
If yes, show substitution and scrap. Heads actually left or redeployed. Overtime actually down. Yield actually up. If those files do not exist, the cell is décor with hours. Hours without substitution do not earn a high-end seat in the cell.
If substitution exists, keep the published size-band and sub-sector cell. A pallet pool does not create a medical row. A press robot does not create an aerospace row. IBBA Pulse still reported that most advisors saw no material AI effect on value. Keep software in the operations narrative. Keep iron in the iron narrative. Do not staple them into Industry language and call it a turn.
A closer read on cell memo
One page per cell: operation, go-live, integrator, twelve-month uptime, changeover, scrap before and after, heads actually removed or redeployed, in-capacity or not. Empty lines mean the cell is a photo. A cobot the owner starts every morning is owner dependence. Credit committees fear crates and they fear unused hours. Compare the memo to the insurance schedule. Decide the stack. Demonstrated substitution can support the high end of the current cell. Unused iron goes to OLV. Do not put the same robot in both stacks at full credit. A pallet pool does not create a medical row. A scheduling overlay does not create a new official multiple. Keep software in the operations narrative. Keep iron in the iron narrative.
What a cobot that needs the owner every morning is worth
A collaborative robot that the owner homes, teaches, and restarts after every crash is not a labor substitution. It is an owner-dependence story with a photogenic arm. Credit committees have learned to ask who starts the cell on a Monday after a holiday. If the answer is the seller, the cell does not support the high end of the current size-band and sub-sector earnings cell. It may still have orderly liquidation value as iron. Those are different stacks. Do not put the same arm in both at full credit.
A full cell with twelve months of uptime, a scrap-before-and-after exhibit, and heads actually removed or redeployed can support the high end of the cell you already sit in. It cannot move commodity fabrication into an aerospace row. It cannot create a new official multiple. Deloitte and the Manufacturing Institute’s 3.8 million workers needed from 2024 through 2033, with 1.9 million potentially unfilled, is climate. Your cell memo is the plant. Climate does not replace the memo.
Unused automation on the floor is a crate problem even when the crate has been opened. Hours in capacity versus hours as a tour prop decide the stack. Compare the insurance schedule to the cell memo. A robot that is on the schedule and not in the hours file is a photograph. A scheduling overlay or a pallet-pool slogan without a second-shift superintendent is the same photograph in software form.
Integrators who still have a support contract, and integrators who disappeared, change remaining life more than the year on the teach pendant. Write the vendor. Write the last service. Write whether the current programmer can touch the teach pendant without a plane ticket. That paragraph belongs next to the skilled-labor page, not inside a growth multiple.