Samsung union votes on strike that could hit chip output

Samsung Electronics’ labor union is set to begin a vote on industrial action on March 9 that could lead to a full-scale strike and, observers say, disrupt semiconductor production schedules.

Industry sources said the union’s “joint struggle headquarters” will run the vote from March 9 to 18 and that membership is estimated at about 89,000.

The union said a majority yes would trigger a rally on April 23 and an 18-day strike from May 21 to June 7.

Union leaders plan office monitoring and a reporting channel, which critics say pressures non-participants.

Wage talks broke down over bonus calculations, with the company offering raises, stock grants, and special bonuses.

A May strike could disrupt HBM4 memory production for Nvidia’s AI accelerator, Vera Rubin, due to production lead times.

🔗 Source: Chosun Daily

🧠 Food for thought

Implications, context, and why it matters.

A potential strike could hit Samsung at a sensitive moment in its HBM4 ramp

A proposed May strike may squeeze the launch window for Samsung’s High Bandwidth Memory (HBM)4 chips, while full-scale supply is expected around June 1.
HBM4 is part of Samsung’s comeback plan in the high-demand HBM market, where it has trailed rival SK hynix 2.
In the third quarter of 2025, SK hynix held 53% market share and Samsung had 35%, which raises the stakes for this rollout 2.
Samsung executives have tied the company’s reputation to the launch, repeating a customer line that “Samsung is back,” which adds pressure to avoid hiccups 2.

Samsung’s labor dispute could worsen the global AI chip bottleneck

Any production slip can affect Nvidia, since its next-generation Vera Rubin AI platform is set to use HBM4 memory, which limits a fallback to older versions 3.
A delay may push Nvidia further toward SK hynix, which can boost the rival’s leverage on pricing and supply 2.
Broader knock-on effects may reach the wider tech sector, which already faces an AI hardware “capacity crunch” expected to last two years 2.
HBM3 and HBM3E are expected to stay the main standards through at least Q1 2026, and many HBM3 and HBM3E nodes are sold out through late 2026, so a meaningful slowdown in HBM4 from a key supplier can drag the pace of AI development across the board 3.