Divide before deciding who is right: if 200 percent of target is worth about five months of pay, target is 2.5 months, so an average payout landing near 2.6 months has been running at roughly 104 percent of target and about 52 percent of the advertised ceiling [6][12][13]. That is a plan doing what its label says, which is the awkward part for both sides, because the grievance as reported by Cryptopolitan is about reconstructability rather than about the average [16]. Samsung's 10.5 percent of chip-division profit and SK Hynix's 10 percent of annual operating profit can be checked against a reported result [4]; Micron's plan asks a Taoyuan or Taichung member to accept a figure they cannot rebuild [16].
Read as cash rather than as fairness, the ask is a claim on the cycle. A cap denominated in months of salary pays the same five months whether contract prices are flat or up 80 to 90 percent in a quarter, which the source says is what happened as supply moved toward high-bandwidth memory [8]; a percentage of operating profit pays whatever the upcycle pays [4]. SK Hynix's Kwak Noh-Jung said in late August that he expects the shortage to run to the end of 2030 [9], and Korean semiconductor shipments were last reported up 209 percent year on year at $46.65 billion [15], so these two unions are asking to be moved off a salary-linked instrument and onto an earnings-linked one at the moment the two diverge most.
The leverage is unusually clean. Line work is specialized and the operators are hard to replace, which is exactly why Taiwanese fabs almost never stop [7], and the Taiwan sites carry both DRAM manufacture and advanced packaging that cannot be shifted abroad quickly against very little idle capacity anywhere else [10][11]. With three vendors holding about 94 percent of DRAM [3], an even split would be 31.3 percent each [17], and Micron's real share hardly needs to be known for a fortnight of disruption to price itself.
This is probably wrong, but the likeliest outcome is that Micron pays and concedes nothing structural, because a one-time supplement expires with the cycle while a published percentage reprices every upcycle after it, including the one running to 2030 that Kwak describes [9]. The second path is that a September ballot happens and falls short, since 80 percent of a poll is not 80 percent of a vote [2]. The third, and the more interesting version, is that Micron simply publishes the methodology without changing the money, which answers the stated complaint [16] at close to zero cost.
What would prove the thesis wrong is a settlement that hands over a percentage and keeps it after memory prices normalise. What would confirm it is a September announcement of a large one-off with the Incentive Pay Plan structurally intact. Tim Cook, on his last earnings call as chief executive, warned that memory prices would keep rising and squeeze profits [14], which is a reasonable guide to who funds either version.
