Around three hundred and forty thousand dollars. That is the average annual bonus that roughly forty eight thousand workers in Samsung’s semiconductor division become eligible for under the agreement their union approved this week, about ninety minutes before a strike was due to begin at the largest memory chip maker in the world.
It is an extraordinary number and it is not a pay rise. The distinction is the entire deal, and it is worth setting against the way this problem is normally solved.
The problem both approaches are trying to solve
A workforce in a violently cyclical industry wants a share of the good years. That is not greed; it is a reasonable response to having absorbed the bad ones.
Semiconductors are as cyclical as any large industry on earth. Demand swings, capacity arrives in enormous indivisible increments because a fabrication plant costs tens of billions and cannot be built by halves, and the result is a market that alternates between shortage and glut on a period of a few years. Operating profit in a good year can be many multiples of a bad one.
So the question is how to give people a claim on that without breaking the company in the downturn, and there are two established answers.
The conventional answer
Raise wages.
It is simple, it is what unions have historically asked for, and it has one decisive property: it is permanent. A wage negotiated this year is paid next year and the year after, in a downturn, when the profit that justified it has gone.
That permanence is the point from the worker’s side and the problem from the company’s. A fixed cost base set during a peak is exactly what destroys companies in a trough, and every management in a cyclical industry knows it, which is why wage negotiations in these industries are fought as hard as they are.
What Samsung did instead
It agreed to distribute 10.5 percent of the chip division’s operating profit to employees in the form of stock, plus a further 1.5 percent in cash, over ten years, contingent on the division hitting profit thresholds: 200 trillion won of annual operating profit from 2026 to 2028, and 100 trillion from 2029 to 2035.
Three features of that structure are doing the work.
It is variable. In a year when the division misses the threshold, the obligation is nothing. The company has bought ten years of labor peace at a cost that automatically disappears in precisely the years it could not afford to pay it. That is a considerably better trade than any wage settlement available to it.
It is mostly stock. Of the twelve percent, ten and a half is equity and one and a half is cash, which is a ratio of about seven to one. The workforce is therefore exposed twice: once to whether the division makes the profit, and again to what the market pays for the company while they hold the shares. Those are correlated but not identical, and the second exposure was not obviously what anybody was asking for.
The threshold falls. Two hundred trillion won for the first three years and one hundred thereafter. Read one way that is an acknowledgment that the current boom is the boom and the back half of the decade will be harder. Read another it is a floor set low enough to keep the agreement alive through a downturn, which is what makes a ten year deal signable.
What the comparison teaches
That the two approaches allocate the same money and completely different risk.
Under a wage settlement the company carries the cycle. It pays in the trough, it borrows if it has to, and the workforce’s income is stable while the shareholders absorb the variance.
Under this arrangement the workforce carries a substantial share of the cycle, in exchange for a share of a peak far larger than any wage negotiation would ever have produced. Nobody was going to negotiate a three hundred and forty thousand dollar annual pay rise for forty eight thousand people.
Which is why roughly seventy four percent voted for it, and it is also why the twenty six percent who did not are not being unreasonable. A worker with a mortgage and a stable job has just accepted that a meaningful part of their compensation now depends on a semiconductor cycle they cannot influence and a share price they certainly cannot.
There is one more wrinkle in the thresholds that is worth noticing. Operating profit is a divisional figure, and a division’s profit depends partly on how a large company allocates shared costs and internal transfer prices between its parts. Nothing suggests anything improper here, and the number is audited. But a bonus pool defined as a percentage of one division’s profit gives everybody involved a reason to care intensely about accounting decisions that were previously of interest to nobody, and that is a new dynamic inside the company whether or not it ever changes a figure.
What it does to everybody else
Immediately, pressure. A deal of this size at the largest employer in the sector becomes the reference point in every negotiation that follows, domestically and beyond, and other Korean industrial unions have already noticed.
That is the ordinary mechanism by which a settlement at one firm propagates. What is unusual here is the form rather than the amount: if profit linked equity becomes the template, a great many workforces in cyclical industries will find themselves being offered upside instead of certainty, and the choice will be presented as a raise.
There is a second order effect worth naming, which is what this does to the labor market inside the company. A bonus of this size, paid in stock that vests over time, is an extremely effective retention device: leaving means walking away from a claim on a future distribution. A competitor trying to hire an experienced process engineer out of that division is now bidding against a number nobody can match with salary.
The part that does not resolve
Ten years is a very long time in this industry, and the agreement’s fate depends on a thing nobody can forecast.
The current profits are being produced by demand for memory to feed artificial intelligence training and inference, which has grown faster than anybody’s plan and which is the reason 200 trillion won is a plausible threshold at all. If that demand holds, this is the best labor agreement any industrial workforce has signed in a generation.
If it softens, the mechanism reverses quietly. The threshold is missed, the obligation is zero, and forty eight thousand people who were told their compensation had been transformed discover that what changed was the variance rather than the level. There is no provision that protects against that, because a provision that did would reintroduce the fixed cost the whole structure exists to avoid.
The workforce has not been given a share of the profits. It has been given a share of the risk, priced generously, and whether that was a good trade will be knowable in about 2031.
