Samsung engineers defect to SK Hynix over HBM bonuses
A reported 81.5% of Samsung foundry staff want out within two years as rival SK Hynix pays bonuses more than three times larger on the back of AI memory profits.
A bonus gap that has become a talent drain
Samsung is losing chip engineers to domestic rival SK Hynix at a pace that its own internal surveys describe as severe, according to a report published Tuesday. The proximate cause is compensation structure rather than base salary. SK Hynix distributes roughly ten percent of operating profit to employees company-wide; Samsung ties bonuses to individual division performance, which penalises staff in units running at a loss.
The arithmetic that follows is stark. Bonuses at SK Hynix have reached around $476,000, mostly in cash, against roughly $135,000 for workers in Samsung's foundry division. More than 200 unionised Samsung employees moved to SK Hynix over a four-month stretch through April. One engineer quoted in the report said his entire 30-person team applied to the rival, with the exception of two team leads. An internal figure of 81.5 percent of foundry staff wanting to leave within two years indicates the problem extends well past those who have already moved.
Samsung has begun litigating. It secured a court injunction barring two workers from joining SK Hynix for 18 months, an unusually aggressive step between two Korean firms that have historically competed for the same graduates without going to court.
The divergence traces back to a 2019 decision to downsize Samsung's high-bandwidth memory team. HBM is now the memory tier that feeds Nvidia's AI accelerators, demand is running far ahead of supply, and SK Hynix holds the market lead. Both companies have recently crossed $1 trillion in market value on the strength of the AI memory cycle, even as their internal fortunes diverge by division.
Why it matters
HBM supply is the hardest physical constraint on AI accelerator output, and Samsung is the only supplier with the scale to meaningfully loosen it. Its ability to close the gap with SK Hynix depends on process engineers who are currently leaving, and semiconductor expertise does not transfer through hiring campaigns on a timescale that matters to this cycle. A structural talent imbalance between the two Korean suppliers therefore propagates outward — to Nvidia's roadmap, to hyperscaler capacity plans, and to the price of every AI training cluster built over the next several years.