The letter came addressed to the plant manager and it was two paragraphs. As of a date about fourteen months out, the controller in that machine would reach end of support: no security updates, no parts, and the cloud service the machine uses to authenticate its own firmware would be retired.
The machine is from 2016. It runs. It has made the same part to the same tolerance for nine years and there is nothing wrong with it.
What the letter is really saying is that a capital asset the company bought outright is now going to stop, on a date chosen by somebody else, because a piece of it phones a server.
I have watched this arrive at three plants now and I want to separate two things that get argued as one, because the shop-floor version of this complaint is usually nostalgia and the nostalgia is wrong.
The connected machine genuinely does more. Remote diagnostics on that model have twice saved us a service visit and about two days each time, because the vendor's engineer could see the fault log from four states away and tell us which board to pull. Firmware updates have fixed real bugs. The usage data has settled arguments about maintenance intervals that we used to have by shouting. I would not buy the disconnected version if it were offered, and when people romanticise the machine that only needed a screwdriver, they are describing a machine that also broke in ways nobody could diagnose.
So the connection is not the problem. The dependency is, and they are not the same thing, and the industry has been careful not to distinguish them.
Here is the distinction. A machine that can phone home is better. A machine that must phone home to run is a different kind of object. The first is a capital asset with a service attached. The second is a service with a capital asset attached, and the difference shows up on exactly one day: the day the vendor decides.
Nobody sold us the second thing. We bought the first thing in 2016, with a purchase order, depreciated over ten years, and the authentication requirement arrived in a firmware update in 2019 that we accepted because it fixed something else. There was no negotiation. There was a changelog.
The plant's options now are the ones you would expect and all three are bad. Buy the replacement, which is a machine we do not need at a price that assumes we have no alternative. Run unsupported, which our insurer has views about and which means the next failure has no parts behind it. Or pay for the extended support programme, which is real and is offered and costs, annually, a meaningful fraction of what the machine cost, forever.
What I would want is not a regulation and not a boycott, both of which get proposed and neither of which survives contact with a purchasing department that is measured on unit price. It is a line in the quote. Under what conditions does this machine stop running, and for how long are those conditions guaranteed. Two sentences, at the point of sale, comparable across vendors.
Ask that question today and you get an answer that is friendly and is not a commitment. I have asked it four times.
The 2016 machine has a plate on it with a serial number and a date, which is the convention for something meant to outlast the people operating it, and about eighteen inches away there is a network jack that somebody ran in during a retrofit, and the plate is now the less accurate of the two.