Legacy Equipment Replacement 2026 — Four Deadlines Before the Parts Vanish

Legacy equipment replacement usually gets discussed only after something breaks, or after the asset is fully depreciated. But what actually ends a machine’s working life is not wear in the mechanical section. It is the manufacturer’s calendar for control parts. A cast-iron frame will last 30 years, while the CPU, the touch panel and the servo amplifier are discontinued in 10 to 15 years. Using a Japanese-owned metalworking plant in Rayong, Thailand as a model case, this article shows that there are four deadlines rather than three, that the choice is not binary, and that upgrading the control section alone costs 18.3% of a full machine replacement and pays back in 2.6 years. The formulas are all disclosed.

What decides the life of legacy equipment is the manufacturer’s calendar, not wear

Production equipment in Japan is already very old. An editorial in the Nikkan Kogyo Shimbun (March 19, 2025, citing Japan Machine Tool Builders’ Association data) reports that more than 60% of installed equipment is over 10 years old, and more than 35% is over 20 years old. Two machines in every three are ten-year veterans, and one in three is a twenty-year veteran.

At overseas plants the numbers work out worse. It is common for a factory to have been started up with equipment relocated from a plant in Japan. Ten years of use in Japan, shipped to Thailand, another 15 years here, and the machine is 25 years old. The JETRO FY2025 Survey on Business Conditions of Japanese Companies Operating Overseas (Asia and Oceania edition, November 2025) collected valid responses from 5,109 Japanese companies operating in 20 countries and regions. That many Japanese-owned operations are running overseas, and a great number of those sites are now 20 or 30 years old. Equipment age has accumulated accordingly.

The mechanical section lasts 30 years, the control section disappears in 10 to 15

The point to grasp here is that the phrase “the equipment is old” actually mixes two completely different stories.

The first story is about the mechanical section. Cast-iron beds, frames, spindles, ball screws, linear guides. These wear, but wear can be measured. Because it can be measured, it can be managed through replacement, regrinding and adjustment. Machine tools that still hold accuracy after 20 or 30 years do so because this kind of upkeep is possible.

The second story is about the control section. CPU units, I/O units, touch panels, servo amplifiers, power supply units. These do not die from wear. Their life ends the moment the manufacturer stops production. The unit may still be running, but when it fails, the same part is no longer available. These are two entirely different kinds of deadline.

Note that the “why replacement is forced” column above summarizes general tendencies by section. The specific number of years varies widely by model, manufacturer and operating environment. Always confirm the deadlines for your own equipment model by model in the manufacturer’s discontinued products list.

Build the plan from a control-parts deadline register, not the asset register

One practical conclusion follows from this structure. A legacy equipment replacement plan should be built from a control-parts deadline register, not from the acquisition dates in the asset register.

Most plants keep an asset register listing machine number, manufacturer, model, acquisition date and book value. All you can extract from that is “oldest first.” But what determines replacement urgency is not age. It is when the control parts installed on that machine become unobtainable. A machine built in 1998 whose control section was upgraded in 2015 still has time. A machine built in 2012 whose CPU is discontinued in 2026 comes due first.

As long as the priority discussion happens over an asset register sorted by age, that reversal stays invisible.

There are four deadlines, not three — you hold the fourth one

Manufacturers publish three deadlines. Order deadline, discontinuation, and end of repair support. Most plants look at those three and conclude that there are still years left before repair support ends.

In practice, though, the deadline that arrives first is usually the fourth one.

The fourth deadline is a deadline on information and people, not parts

The fourth deadline has nothing to do with parts inventory. It is the deadline on whether the information and the people needed to repair that machine still exist inside your company.

Concretely, this is what gets lost. The as-built wiring diagram of the control panel was never updated through successive modifications and no longer matches the hardware. The PLC program exists only inside the PLC, not as source code. That program is password protected and the person who set the password has left. The touch panel screen data will only open in the version of the screen editor used at the time. The servo parameter backup vanished along with the laptop of a departed engineer. The production engineer who attended the commissioning has been repatriated to Japan.

Any single one of these gaps means you cannot repair the machine even with the part in hand. More precisely, you can repair it but you cannot restore it to how it was. Without the parameters, you start over from trial operation. Without a readable program, you rewrite it while guessing at the intended behavior. A one-day recovery becomes two weeks.

The more equipment came from Japan, the sooner the fourth deadline arrives

At overseas plants the fourth deadline arrives early for structural reasons. There are three of them.

The commissioning was done by a production engineering department in Japan. The drawings and manuals are in Japanese. The program passwords and the design intent live in the head of an engineer on the Japanese side.

At a plant in Japan, the department that did the design sits in the same building, and someone still remembers ten years later. That does not happen overseas. Three to five years after start-up the engineer is repatriated, subsequent modifications are done locally, and the Japanese-language drawings stay on the shelf without ever being updated. The plants most confident that they are safe because manufacturer repair support still runs for years are precisely the plants that reach the “parts available but unrepairable” state first. That follows directly from this structure.

Measuring your own fourth deadline is easy. Pick one machine and ask, “who can open the source of this machine’s PLC program today, password included?” If no name comes back, that machine’s fourth deadline has already passed.

Deadlines running right now — discontinuation dates get extended and withdrawn

Rather than staying abstract, here are two deadlines that are actually running.

Mitsubishi Electric MELSEC-Q Universal model QCPU

Mitsubishi Electric has announced an order deadline of September 30, 2026 and discontinuation in October 2026 for the Universal model QCPU in the MELSEC-Q series (14 model numbers from Q03UDCPU through Q26UDEHCPU). Repair support, according to the company’s notice, runs for seven years from discontinuation.

As of August 9, 2026, the publication date of this article, 52 days remain until the order deadline.

That 52-day figure means more than “new units can no longer be bought.” It is also the last date on which spares can be secured. After discontinuation you are hunting through distributor stock, prices rise, and counterfeit risk appears. The option of extending life through spare parts itself disappears at the order deadline.

On the specific date of the end of repair support, this article goes no further than the published wording of “seven years from discontinuation.” We were unable to fix a date from the relevant section of the primary source. If your equipment is affected, always confirm model by model in the Mitsubishi Electric discontinued products list and technical news.

OMRON extended deadlines and withdrew some entirely

Deadlines do not always move earlier.

In a notice dated November 5, 2024, OMRON extended the discontinuation of the CJ2H-CPU6□-EIP from the end of March 2025 to the end of March 2027, and the CJ2M-CPU3□ from the end of March 2026 to the end of March 2028. For the CS1D-CPU6□H/P and the CJ1W-NC□□3, the discontinuation was withdrawn altogether.

This matters a great deal in practice. A replacement plan built in early 2024 on the assumption that “it ends in March 2025” had its premise changed by this notice. Conversely, a plant that rushed to secure budget without knowing the premise had changed has invested two years earlier than necessary.

That is why the deadline register needs an annual review

Deadlines get extended, and they get withdrawn. Only one operating rule follows from that fact. A deadline register cannot be built once and left alone. It needs an annual stocktake.

There are three things to check. Whether any model number you hold has newly become the subject of a discontinuation announcement. Whether a deadline you already track has been extended or withdrawn. Whether successor model numbers and migration methods have been published.

All three can be confirmed by visiting the manufacturers’ discontinued products lists once a year. The work itself takes half a day. Even so, few plants do it. Most run on the basis of “we will think about it when the trading company calls us.”

The choice is not life extension versus full replacement — four options for legacy equipment replacement

With the deadlines visible, the next question is what to replace and how far.

Discussion on the shop floor usually becomes binary. Buy spares and extend the machine’s life, or swap the whole machine out. Frame it that way and the sequence is always the same. A quotation for a full machine replacement arrives, the figure is too large for approval, life extension continues by default, and the deadline keeps closing in.

In reality there are two options in between.

The four options

Option B is what is usually called a retrofit. On a machine tool, the cast-iron bed, the spindle and the ball screws all stay in place while the control and drive sections move to the current generation. Option C adds the panel enclosure itself, the terminal blocks, the breakers and the internal wiring.

The reason these two get skipped in the discussion usually comes down to where information comes from. Ask the machine builder and you get a quotation for a new machine. Ask the trading company and you get a quotation for spare parts. Options B and C come from neither of them. Upgrading the control section is not a machine builder’s product and not a trading company’s product. It is control engineering work.

Option B is 18.3% of option D, option C is 23.8%

A later chapter builds up the per-machine cost, but here is the conclusion first. In the model-case estimate, option B is 18.3% of option D, and option C is 23.8%.

Submit a 7,840,000 baht capital request, fail to get it approved, and keep extending the machine’s life. Or spend 1,435,000 baht and push the deadline out by more than 10 years. That gap is not a difference in technology. It is the difference between listing two options and listing four.

Note also that upgrading the control section is a different matter from adding IoT to existing equipment. Retrofitting sensors to collect operating data is covered in IoT retrofit for existing equipment, but that is about collecting data without changing the machine. Collecting data does not extend a control part’s deadline by a single day.

Comparing equipment upgrade cost across the four options on one machine

From here the numbers are built up on a model case.

Model case assumptions

A Japanese-owned metalworking plant in Rayong, Thailand. It has pressing, machining and assembly processes and 40 machines. It runs two shifts, 16 hours a day, 250 days a year.

First, here is the result of an inventory in which the control parts of every machine were checked against the manufacturers’ discontinued products lists.

Of the 40 machines, 14 (35% of the total) were already covered by discontinuation announcements. Six of those have repair support ending within five years. And nine hold no spare parts at all. For those nine, the moment a single control part fails, recovery begins with a hunt through distributor stock.

From that group we take one representative machine, a CNC lathe built in 1998 whose control section was upgraded in 2009, and compare the four options. The figures below are model-case estimates. Real quotations vary with the model, the number of axes, the condition of the panel and the allowable downtime window.

Option A, life extension (securing maintenance stock)

The initial cost is 85,000 + 45,000 + 180,000 + 60,000 = 370,000 baht. On top of that, storage and an annual power-on check cost 12,000 baht a year. Over the seven years to the end of repair support, 370,000 + 12,000 x 7 = 454,000 baht, or 64,857 baht a year.

Around sixty thousand baht a year. On the numbers alone it looks cheapest. But what those 454,000 baht buy is seven years of time and nothing else, and after seven years nothing remains. Once a spare is used you need to buy another, and by then a new unit will most likely be unobtainable. And the fourth deadline, the deadline on drawings, programs and people, is not extended by a single day under this option.

Option B, control section only (retrofit, mechanical section reused)

The total is 220,000 + 95,000 + 480,000 + 140,000 + 260,000 + 180,000 + 60,000 = 1,435,000 baht.

What deserves attention in this breakdown is not the hardware cost (795,000 baht for CPU, HMI and drives) but the 500,000 baht made up of 260,000 baht of program migration, 180,000 baht of commissioning and 60,000 baht of drawing updates. These are costs that leave no physical object behind, so they are the first things cut when a quotation needs to look cheaper. What the cut actually produces is a failure to reset the fourth deadline. If the drawings are not updated and the program source is not delivered, exactly the same problem returns in ten years.

Option C, new control panel

1,435,000 + 320,000 + 110,000 = 1,865,000 baht. The gap against option B is 430,000 baht.

Paying that difference is worth it in cases like these. The inside of the existing panel has been modified so many times that it can no longer be traced. The terminal blocks and breakers are aging too. The drawings do not match the hardware. And above all, downtime drops from 10 days to 7 days. Because the panel is built off site in advance and the on-site work is limited to swapping, terminating and tuning, the line stops for fewer days. On a line running at high utilization, those three days can be worth more than 430,000 baht. The approach to designing and building the panel itself is set out in control panel design and manufacturing.

Option D, full machine replacement (a new equivalent-class CNC lathe)

6,800,000 + 420,000 + 380,000 + 150,000 + 90,000 = 7,840,000 baht. Note that 1,040,000 baht falls outside the machine price itself. Estimate an equipment upgrade cost on the machine price alone and that 1,040,000 baht surfaces later.

The four options side by side

1,435,000 / 7,840,000 = 18.3%, and 1,865,000 / 7,840,000 = 23.8%. The 454,000 baht seven-year total for option A is roughly a third of option B, but what remains after seven years is not the same.

And option D does not stand up on replacement alone. As the later chapter shows, its payback is 14.4 years. Option D is justified only when effects unrelated to replacement can be added on top, such as capacity expansion, new orders, or a larger workpiece envelope. Put the other way, propose option D on the grounds that the machine is old and it will always stop at the investment criteria.

What a control panel upgrade actually touches — CPU, HMI, drives and panel internals

Let us look more concretely at what the 1,435,000 baht of option B is spent on. Without that, you cannot judge whether a quotation is reasonable.

Swapping the CPU and I/O

Migrating to a current-generation CPU is not a straight swap. The base unit form factor changes, I/O assignments change, and network standards can change. Where an old-generation expansion unit has no equivalent in the successor series, the function has to be engineered another way.

One thing that must always be in the quotation here is a physical count of I/O points. Machines whose actual I/O count differs from the drawing are common. Added sensors never reflected in the drawing, unused wiring still sitting on the terminal block, and so on are the norm. Order on the drawing’s point count without a physical check and you run short during on-site work, which means extra cost and a longer schedule.

Upgrading the touch panel (HMI)

The touch panel is one of the parts of the control section whose deadline tends to arrive earliest, because the supply of LCD panels and backlights can end on a shorter cycle than the PLC itself.

The problem at upgrade time is the screen data. Manufacturers provide tools to convert screen data from the old model to the new one, but elements get lost in conversion. Macros, recipe functions, communication settings, fonts. Man-hours are needed to check every screen visually after conversion. “Gap correction” is included in the 260,000 baht program migration cost precisely to cover this work.

Replacing servo amplifiers and motors

480,000 baht for three axes of servo amplifiers and motors. This is not only a large number, it is also the interface with the mechanical section, so there is a lot to verify technically.

What needs checking is the fit between motor shaft diameter and coupling, mounting flange dimensions, encoder resolution, and gain tuning. Gain in particular depends on the wear state of the mechanical section. The backlash of a 20-year-old ball screw is not the backlash of a new one. Loading the same parameters as the old system and running the machine is not an option. That is why 180,000 baht is allocated to trial operation and tuning. How to choose the control method itself is set out in servo control versus inverter control.

Panel wiring and the panel enclosure

Option B reuses the panel enclosure and reworks the wiring inside it. The 140,000 baht covers those materials and that labor. Option C takes the full scope of option B and adds 320,000 baht for a new enclosure, terminal blocks and breakers, plus 110,000 baht for installation and removal of the existing panel. Rewiring is included in both options, so the 430,000 baht difference is the cost of the enclosure and everything around it.

Here are some rules of thumb for the decision. Temperature rise inside the panel is high. Terminal blocks have been extended to the point where there is no spare capacity. The existing breakers are to an old standard. The as-built wiring diagram does not match the hardware. If two or more of these apply, option C is worth considering. If the panel is tidy and the drawings match, option B is enough.

Updating drawings, programs and documentation

The last 60,000 baht is in fact the item that directly resets the fourth deadline.

What has to be updated is the as-built wiring diagram, the I/O list, the parameter list, the PLC program source (with comments), the touch panel screen data, and the operating procedures. Only when these are delivered in a state that matches the hardware can an engineer ten years from now repair the machine.

An upgrade project that cuts this item to compress cost and schedule simply rebuilds the same wall ten years later. The manufacturer’s three deadlines reset when you fit new parts, but the fourth deadline resets only if you require the documentation as a deliverable.

Loss avoidance alone is not enough — three effects that change the payback period

The single biggest reason replacement proposals fail approval is how the benefit is written.

Legacy equipment replacement does not increase output, and it does not dramatically cut defects. So the benefit tends to be written as nothing more than “we avoid the loss when it breaks.” But divide by loss avoidance alone and the payback period stretches.

Divide by expected loss alone and you get 5.6 years

Start with the loss avoided.

The scenario is a long stoppage caused by an inability to procure a control part. With a loss of 9,500 baht per hour of downtime, 21 days to recovery, and 16 operating hours a day, downtime is 21 x 16 = 336 hours. 336 x 9,500 = 3,192,000 baht per occurrence. Set the annual probability of occurrence at 8% and the expected loss is 3,192,000 x 8% = 255,360 baht a year.

That 8% is an assumed value used in the model-case estimate. It is set from the conditions of a machine holding no spare parts at all whose major control parts are already announced as discontinued. When you run the calculation for your own plant, replace it by counting how many times in the past five years recovery was drawn out by control-part procurement.

Divide the 1,435,000 baht of option B by that expected loss alone and 1,435,000 / 255,360 = 5.6 years.

Five point six years. At a company whose investment criterion is three or five years, the proposal is rejected right there. This is the cause of replacement proposals failing approval.

The two effects that belong in the same calculation

Beyond loss avoidance, a control section upgrade produces two more effects that can be expressed in money.

First, reduced electricity consumption. Moving from an old-generation servo amplifier to the current generation lowers power consumption through better regeneration and switching efficiency. The model-case machine consumes 96,000 kWh a year. Set the reduction at 11% and 96,000 x 11% = 10,560 kWh saved, which at an electricity price of 5.9 baht/kWh gives 10,560 x 5.9 = 62,304 baht a year.

Second, fewer unplanned stoppages caused by the control system. On this machine, control-system trouble (contact faults, communication dropouts, lost parameters, unresponsive panel touch and so on) caused 60 hours of unplanned stoppage a year. Assume 36 hours a year after the control section upgrade, a reduction of 24 hours (a 40% cut). 24 x 9,500 = 228,000 baht a year.

255,360 + 62,304 + 228,000 = 545,664 baht a year.

Counted together, option B pays back in 2.6 years

Divide each option by that 545,664 baht.

1,435,000 / 545,664 = 2.6 years, 1,865,000 / 545,664 = 3.4 years, and 7,840,000 / 545,664 = 14.4 years.

Option B, which took 5.6 years on expected loss alone, becomes 2.6 years when the effects are counted together. Same investment, same real-world effects, different write-up. And option D comes to 14.4 years. Option D does not stand up on replacement alone unless the benefits of capacity expansion or new orders are added separately.

What should be stressed here is that counting these together is not padding. Both the electricity saving and the reduction in control-system stoppages are numbers you can measure after the upgrade. If anything it is the 255,360 baht expected loss that is harder to verify, being an estimate multiplied by a probability. Leaving out the verifiable effects and dividing by the unverifiable one is exactly what happens at plants where the proposal fails.

Note that these three count separate events, so there is no double counting. The first is an event where a part cannot be procured and recovery drags on. The third is an event where the parts exist but the control system faults out. The causes differ, and so does the order of magnitude of the recovery time. When you run the numbers for your own plant, split your stoppage records into “prolonged by waiting for parts” and “recovered the same day” and count them separately.

Why option A must not be divided by the same figure

Option A has deliberately been left out of the tables above. Divide option A by the same 545,664 baht and 454,000 / 545,664 comes to under a year, apparently crushing option B’s 2.6 years. That does not hold. There are three reasons.

First, option A captures only the first of the three effects. Holding spares does lower the probability of a long stoppage caused by unobtainable parts. But it lowers only part of it. The spares held are for four items, CPU, touch panel, servo amplifier and power supply, and those do not exhaust the causes of a control section failure. Communication units, encoders, cables inside the panel, and the loss of parameters or the program itself cannot be recovered from a spares shelf. Of the 255,360 baht of expected loss, option A eliminates only a fraction.

Second, the remaining two effects produce not one baht under option A. The 62,304 baht electricity saving and the 228,000 baht reduction in control-system stoppages, 290,304 baht a year together, only appear once the parts are new. Putting spares on a shelf changes neither the power consumption nor the downtime of the machine that keeps running.

Third, the time horizons differ. The probability that option A lowered returns to its original level 7 years later, when repair support ends. In fact it gets worse, because once the spares are consumed there is no procurement route left. The 1,435,000 baht of option B buys more than 10 years of continued benefit. The 454,000 baht of option A buys 7 years of grace. Line them up on the same payback yardstick and that difference in horizon disappears from view.

Option A should therefore be judged on a different metric, namely “how much are you paying for 7 years?” rather than on payback. Deciding to buy seven years of grace at 64,857 baht a year can be perfectly rational in some situations. But only if there is a plan to execute option B or option C within those seven years. Life extension without a plan is spending 64,857 baht a year to arrive, seven years from now, at a position where option D is the only choice left.

Choosing life extension means giving up BOI privileges

There is one more hidden cost in option A. The privileges offered by the BOI, the Thailand Board of Investment.

Measures to improve production efficiency have a minimum investment

The BOI operates an incentive category called measures to improve production efficiency. It supports the replacement and upgrading of machinery at businesses that are already in operation.

As a general criterion, the condition is an investment of at least 1,000,000 baht, excluding land cost and working capital. An improvement that uses automation machinery qualifies for a 3-year corporate income tax exemption, capped at 50% of the investment in the automation machinery. Further, where at least 30% of the automation investment is purchased from manufacturers in Thailand, the exempted amount is adjusted upward within a limit not exceeding 100% of the automation investment.

Applying that criterion to the four options gives the following.

The 370,000 baht of option A does not reach the 1,000,000 baht investment threshold, so it does not qualify for this category at all.

Option B, on the other hand, gives an indicative corporate income tax exemption of 50% of the 1,435,000 baht investment, or 717,500 baht, rising to 1,435,000 baht if local Thai sourcing is at least 30%. Note, however, that the source expresses the cap in terms of investment in automation machinery. The figures here are a rough calculation based on the total of option B, and how far program migration and commissioning costs fall within scope is decided case by case.

This is what it means to say that life extension looks cheap while quietly giving up the privileges. Under option A you spend 454,000 baht over seven years, keep nothing, and receive zero privileges. Under option B you spend 1,435,000 baht, push the deadline out by more than 10 years, and stand up an exemption of as much as 1,435,000 baht. Put that comparison into the capital request and the terms of the discussion change.

The two measures, and the Industry 4.0 category

The measures to improve production efficiency comprise a first measure (replacement or upgrading of machinery for energy conservation, alternative energy use, or reduction of environmental impact) and a second measure (improving production efficiency by upgrading machinery).

Option B is likely to fall under the second measure, though a control section upgrade that reduces electricity consumption could also fall under the first. Which way it is classified depends on the content of the case, and this article cannot state it definitively. Confirm with the BOI or your agent before you apply.

A larger category is the incentive for Industry 4.0 transformation. It offers exemption from import duty on machinery plus a 3-year corporate income tax exemption, covering 100% of the investment. It does require a transformation investment plan certified by the National Science and Technology Development Agency (NSTDA), however, so it is not a category you can use for a one-machine control section upgrade. It is an option to consider if you hold a transformation plan for the whole plant.

The BOI conditions cited here are based on a BOI document dated November 18, 2021. Incentive content and conditions are revised over time, so always check the latest announcements when you apply. Read the figures in this article as a guide for comparing replacement options.

How to build the deadline register — taking stock of 40 machines in one day

Everything discussed so far assumes you know your own deadlines. So how do you establish them?

The short answer is that at a scale of 40 machines you can build the register in one day. What you need is the asset register, time spent looking inside the control panels, and the manufacturers’ discontinued products lists. No system is required.

The columns the register needs

The first five columns identify the machine and its model numbers, the next three are the manufacturer’s deadlines, the ninth column on spares shows how much room there is to extend life, and the last three are the internal deadlines. A register without those last three columns will miss the fourth deadline.

The half-day procedure

Collecting the model numbers off the hardware means opening the panel door and photographing the nameplates with a smartphone. For 40 machines, two people need less than half a day. Transcribe the model numbers from the photos and search the manufacturers’ discontinued products lists. Both Mitsubishi Electric and OMRON offer model number search on the web.

Restrict this to three items, the CPU, the HMI and the servo amplifiers. Chase every component in the panel and you will never finish. Those three dominate both on cost and on procurement difficulty. Power supplies and I/O can be dealt with alongside when those three are upgraded.

Spares held are counted physically in the maintenance store. The three internal columns, the location of the source, the password and the drawing status, are filled in by interview. The machines where “unknown” piles up are the ones that rise to the top of the deadline register.

Setting the replacement order from the register

Once the register exists, decide the sorting criteria. Acquisition year is not one of them.

The first axis is the number of years remaining until repair support ends. The second is whether spares are held. The third is the hourly loss when that machine stops. The fourth is how many “unknown” entries appear in the three internal columns.

Rank on those four axes and sort by the lowest total rank. In the model case, that sort put the overlap between the 9 machines holding no spares and the 6 machines with repair support ending within five years at the top. That is where the work starts.

The question of how much day-to-day maintenance to keep in house and where to draw the line for outsourcing also becomes more precise once this register exists. Equipment maintenance outsourcing sets out how to draw that line.

Seven items to settle before you place the order

There are items that belong in the specification before you order an upgrade project. Leave them undecided and you end up with the work complete but the fourth deadline not reset.

The three most often dropped

Of the seven, the ones most often dropped in practice are items 2, 3 and 7.

Item 2, password escrow, is very often set by production engineering at the Japanese head office and never handed over. An upgrade project is the one opportunity to reset this. Write into the specification who holds the passwords, where they are recorded, and how they are handed over when staff change.

Item 3, documentation language, is specific to overseas plants. Japanese-only drawings cannot be used by local maintenance staff, so they never get updated, and drawings that are not updated diverge from the hardware within a few years. Require English, or Thai, or at minimum drawing symbols and signal names that local staff can read.

Item 7, the deadline list, appears in almost no specifications. Without it, the next ten years begin without knowing the next deadline for the machine you just upgraded. Insist on receiving a list of the model numbers of the CPU, HMI and servo amplifiers actually installed, at handover.

Note that how and to whom you contract PLC program development itself needs to be worked out separately from the upgrade project. Source code rights, comment conventions and change control are covered in outsourcing PLC program development.

Five common failure patterns

Failure 1. Taking comfort in the repair support deadline alone

Seeing the notice that “repair support runs seven years from discontinuation” and deferring the upgrade. This is the most common pattern. But the fourth deadline, the deadline on drawings, programs, passwords and people, does not appear in the manufacturer’s notice. You reach the state of having parts but being unable to repair well before repair support ends.

Failure 2. Framing the discussion as life extension or full replacement

Skip options B and C and the discussion becomes a 7,840,000 baht capital request or a 370,000 baht spare parts purchase. The former is not approved, the latter is chosen, and the deadline keeps closing in. Option B is 1,435,000 baht, 18.3% of option D.

Failure 3. Writing the benefit as loss avoidance only

Divide option B by the 255,360 baht a year of expected loss alone and you get 5.6 years. Add the 62,304 baht electricity saving and the 228,000 baht reduction in control-system stoppages for 545,664 baht a year, and it becomes 2.6 years. The cause is leaving out the verifiable effects and dividing by the unverifiable one.

Failure 4. Cutting the program migration and documentation costs

Within the option B breakdown, program migration at 260,000 baht, commissioning at 180,000 baht and drawing updates at 60,000 baht leave no physical object behind. They are the first to be cut when a quotation is compressed, but cutting them means the fourth deadline is never reset. The work is complete, and the same problem arrives ten years later.

Failure 5. Building the deadline register once and never updating it

As OMRON’s notice of November 5, 2024 shows, deadlines can be extended and they can be withdrawn. The CJ2H-CPU6□-EIP moved from the end of March 2025 to the end of March 2027, the CJ2M-CPU3□ from the end of March 2026 to the end of March 2028, and discontinuation was withdrawn for the CS1D-CPU6□H/P and the CJ1W-NC□□3. Skip the annual review and you plan on stale assumptions.

Frequently asked questions

How much does legacy equipment replacement cost?

The range is more than tenfold, depending on what you replace and how far. For the single CNC lathe in the model case, life extension through spare parts came to 454,000 baht over seven years, a control section upgrade to 1,435,000 baht, a new control panel to 1,865,000 baht, and a full machine replacement to 7,840,000 baht. The control section upgrade is 18.3% of the full machine replacement. When you request quotations, align everyone first on which of the four options is being quoted. Options B and C often do not surface whether you ask the machine builder or the trading company, so you have to specify them yourself.

What is the difference between a PLC upgrade and a PLC replacement?

In practice the terms are used almost interchangeably, but using them to define scope makes the discussion more precise. A PLC upgrade, or PLC replacement, generally refers to swapping the CPU and I/O for successor models. A retrofit, by contrast, refers to bringing the whole control section to the current generation, including not just the PLC but the touch panel, servo amplifiers, motors and panel wiring. Option B in the model case is a retrofit, and the 220,000 baht within it is the PLC hardware upgrade. Replace only the PLC and leave the touch panel and servos in place and the deadlines on what you left behind arrive a few years later, so you end up doing the work twice. Decide the scope first.

Can a retrofit be used on machine tools?

Yes. Machine tools are in fact the equipment where a machine tool retrofit delivers the most. Mechanical elements such as the cast-iron bed, the spindle, the ball screws and the linear guides can be maintained for 20 or 30 years, while the control and drive sections are discontinued in 10 to 15 years. The 1998 CNC lathe in the model case had its control section upgraded once in 2009, so this would be its second upgrade. Always check the condition of the mechanical section first, however. If ball screw backlash or spindle runout is outside specification, new control hardware will not restore accuracy. The correct sequence is to measure mechanical accuracy first and decide the option after seeing the result.

Is upgrading only the control panel worthwhile?

It is. In the model case, the control section upgrade alone (option B, 1,435,000 baht) produced 545,664 baht a year in benefit and paid back in 2.6 years. That breaks down into 255,360 baht of expected loss from long stoppages due to unobtainable parts, 62,304 baht of electricity savings, and 228,000 baht from fewer control-system unplanned stoppages. Even option C (1,865,000 baht), which rebuilds the panel enclosure as well, pays back in 3.4 years on the same benefit. A full machine replacement (option D) pays back in 14.4 years, so on investment efficiency an upgrade limited to the control section or the control panel is clearly better. This is a model-case estimate, though. When you run it for your own plant, measure your actual hourly loss and your historical control-system downtime and substitute those figures.

What should I do first when I receive a discontinuation notice for a part?

First, count how many of your machines carry that model number. Discovering that a part you thought was on one machine is actually on five is common. Next, confirm the order deadline date. That is the last day on which extending life through spare parts remains an option. The Mitsubishi Electric MELSEC-Q Universal model QCPU (14 model numbers) has an order deadline of September 30, 2026, leaving 52 days as of this article’s publication date of August 9, 2026. Then confirm the internal deadlines for the affected machines, namely where the program source is, who knows the password, and whether the drawings match the hardware. If those three are not all in place, securing the part may still leave you unable to repair the machine, so consider an upgrade rather than life extension.

Can BOI privileges be used for equipment replacement?

There is a category called measures to improve production efficiency, which covers the replacement and upgrading of machinery at businesses already in operation. As a general criterion the condition is an investment of at least 1,000,000 baht, excluding land cost and working capital. An improvement using automation machinery qualifies for a 3-year corporate income tax exemption, capped at 50% of the investment in the automation machinery. Where at least 30% of the automation investment is purchased from manufacturers in Thailand, the amount is adjusted upward within a limit not exceeding 100% of the automation investment. Life extension through spare parts (370,000 baht) does not meet the threshold and is not eligible, whereas a control section upgrade (1,435,000 baht) gives an indicative exemption of 717,500 baht, rising to 1,435,000 baht if the local sourcing condition is met. The conditions cited here are based on a BOI document dated November 18, 2021, so check the latest announcements when you apply.

How many days will the machine be down during the upgrade?

The model case assumes 10 days for a control section upgrade (option B), 7 days for a new control panel (option C), and 3 days for a full machine replacement (option D) where parallel installation is possible. Option C is shorter than option B because the panel is built off site in advance and the on-site work is limited to swapping, terminating and tuning. On a line where downtime translates directly into lost production, those three days can be worth more than the 430,000 baht difference between options C and B. If you want to minimize downtime, option C with its pre-built panel is the better choice. Actual days vary with the number of axes, the volume of wiring and the accuracy of the existing drawings, so ask for a downtime schedule with the quotation.

Summary

The first thing to fix in legacy equipment replacement is the basis for the decision. Build the plan from a control-parts deadline register, not from acquisition dates in the asset register. A cast-iron frame lasts 30 years, but the CPU, the touch panel and the servo amplifier are discontinued in 10 to 15 years. In Japan more than 60% of equipment is over 10 years old and more than 35% is over 20 years old, and overseas plants running machines relocated from Japan are older still.

There are four deadlines, not three. Alongside the order deadline, discontinuation and end of repair support sits the deadline on whether the drawings, the programs, the passwords and the people remain inside the company. The plants most confident that they are safe because repair support runs for years are the ones that reach the “parts available but unrepairable” state first. The Mitsubishi Electric MELSEC-Q Universal model QCPU (14 model numbers) has an order deadline of September 30, 2026, 52 days from this article’s publication date. OMRON, meanwhile, extended deadlines in its notice of November 5, 2024 and withdrew some entirely. That is why the deadline register needs an annual update.

The choice is not binary. Between life extension (option A, 454,000 baht over seven years) and full machine replacement (option D, 7,840,000 baht) sit a control section upgrade (option B, 1,435,000 baht) and a new control panel (option C, 1,865,000 baht). Option B is 18.3% of option D and option C is 23.8%. These two get skipped because neither the machine builder nor the trading company will offer them.

Loss avoidance alone is not enough as a benefit. Divide option B by the 255,360 baht a year of expected loss alone and you get 5.6 years, but add the 62,304 baht electricity saving and the 228,000 baht reduction in control-system stoppages for 545,664 baht a year and option B pays back in 2.6 years, option C in 3.4 years. Option D takes 14.4 years, so it does not stand up on replacement alone unless capacity expansion or new order benefits are added.

And life extension looks cheap while giving up the privileges. The BOI measures to improve production efficiency require an investment of at least 1,000,000 baht, so a 370,000 baht spare parts purchase is not eligible. Option B gives an indicative exemption of 717,500 baht, rising to 1,435,000 baht where local Thai sourcing is at least 30%. These conditions are based on a BOI document dated November 18, 2021, so check the latest announcements when you apply.

Finally, require the documentation as a deliverable when you place the order. The program source, the password escrow arrangement, an as-built wiring diagram matching the hardware, and a deadline list for the new parts. An upgrade project without these resets three deadlines and leaves the fourth untouched. The same wall arrives ten years later.

It is perfectly fine to start with nothing more than a conversation about how to build the deadline register. Collecting the model numbers inside the control panels takes two people half a day at a scale of 40 machines. Including the cross-check against the manufacturers’ discontinued products lists, one day is enough to have a register in shape. Deciding which of options A through D to quote comes after seeing that result, and that is where the real evaluation begins. Where your equipment currently sits against each deadline, how to sequence the upgrades, what a quotation includes and what it leaves out. These are the conversations we have every day. Get in touch through our contact page whenever you are at the point of looking into it.

References

• Nikkan Kogyo Shimbun, “Editorial, the machine tool industry’s sense of crisis, act early on aging production equipment” (March 19, 2025. Reports, citing Japan Machine Tool Builders’ Association data, that more than 60% of installed equipment is over 10 years old and more than 35% is over 20 years old)
• Mitsubishi Electric, “MELSEC-Q discontinued products list” (Searchable by model number for order deadline, discontinuation date and end of repair support)
• Mitsubishi Electric, “Technical News FA-D-0418-A” (Notice on the discontinuation of the Universal model QCPU and migration to successor models)
• OMRON, “Discontinued products list” (Confirms discontinuation dates and successor models by model number)
• Takashima Denki, “Notice of extension and withdrawal of discontinuation dates for certain OMRON PLC models in the CS1 and CJ series” (Content of the November 5, 2024 notice. Extension of CJ2H-CPU6□-EIP and CJ2M-CPU3□, withdrawal for CS1D-CPU6□H/P and CJ1W-NC□□3)
• Thailand Board of Investment (BOI), “Investment promotion policy for the automation machinery and robotics industries in Thailand” (Document dated November 18, 2021. Conditions and exemption caps for the measures to improve production efficiency. Check the latest announcements when you apply)
• JETRO, “FY2025 survey on business conditions of Japanese-affiliated companies overseas, Asia and Oceania edition” (Published November 2025. Valid responses from 5,109 Japanese companies operating in 20 countries and regions)
• JETRO, “Incentives for foreign investment in Thailand” (Overview of the BOI investment promotion system and the types of privileges available)