Samsung Electronics just fired a shot across the bow of the AI chip market. The company has begun mass production and global shipment of its next-generation high-bandwidth memory chips, known as HBM4, sending its stock price surging 6.4% to close at KRW 178,600.

In technical terms, Samsung says the new chips deliver over 40% higher processing speeds compared to previous generations. That’s a meaningful jump for AI data centers, where memory bandwidth has increasingly become the bottleneck preventing faster model training and inference.

Samsung itself is calling these chips “industry-leading” for AI data center applications.

The AMD partnership changes the equation

On March 18, 2026, Samsung and AMD signed a strategic memorandum of understanding for Samsung to supply HBM4 chips specifically for AMD’s upcoming Instinct MI455X GPU.

AMD has been steadily gaining ground in the AI accelerator market, positioning itself as the primary alternative to Nvidia’s dominant data center GPUs. Securing a dedicated HBM4 supply agreement with Samsung gives AMD a second major memory supplier alongside SK Hynix, reducing supply chain risk at a time when AI chip demand continues to outstrip production capacity.

Samsung’s broader AI memory play

HBM4 isn’t Samsung’s only bet on the AI infrastructure boom. The company’s AI memory strategy spans multiple product lines, including HBM, DDR5, and what the company describes as near-memory compute technologies.

Near-memory compute is an emerging approach that moves processing capabilities closer to where data is stored, rather than shuttling everything back and forth between separate memory and processor chips. It’s the kind of architectural shift that could matter enormously for AI workloads that are increasingly constrained by data movement rather than raw computation.

What this means for investors

The 6.4% stock jump tells you the market was hungry for this news. Samsung’s memory division has been a source of frustration for investors who watched the company fall behind SK Hynix in qualifying HBM chips for Nvidia’s platforms.

If Samsung’s entry into mass HBM4 production increases overall supply, it could help moderate the memory pricing premiums that have been a persistent cost driver for AI infrastructure deployments.