AMD CEO Lisa Su calls Samsung and SK Hynix important for long-term supply chain

AMD CEO Lisa Su calls Samsung and SK Hynix important for long-term supply chain

AMD is stretching its memory planning horizon to three to five years as AI demand outruns high-bandwidth memory production

AMD CEO Lisa Su has named Samsung and SK Hynix as important partners for the chipmaker’s long-term supply chain. The comment comes as the AI boom turns a specialized type of memory into one of the scarcest commodities in tech.

That memory is high-bandwidth memory, or HBM. Su says AMD now has to plan its supply years further out than it ever did before.

A longer planning horizon for a tighter market

Su visited Taiwan and South Korea in early October 2026. The trip was aimed at locking in long-term supply agreements with memory makers.

Su emphasized that AMD’s capacity planning has to stretch from the previous one to two years out to a new timeline of three to five years.

On October 6, 2026, Su announced that AMD plans to significantly increase supply in 2027. The ramp reflects a strategic pivot as the company responds to how fast the AI market is growing.

Advertisement

AMD is working with memory chipmakers to address HBM supply constraints that are expected to persist through 2028.

Samsung gets the priority seat

The backbone of AMD’s memory strategy is a deal with Samsung Electronics. On March 18, 2026, the two companies signed a memorandum of understanding naming Samsung as AMD’s priority supplier for sixth-generation HBM4.

That HBM4 is destined for AMD’s new Instinct MI455X AI accelerators. The same agreement also covers advanced DDR5 memory for AMD’s EPYC “Venice” server processors.

Samsung began shipping HBM4 in February 2026. It was already the primary supplier of HBM3E, the prior generation, for AMD’s earlier chip models.

The relationship is not new. AMD and Samsung have worked together for nearly two decades, so the MOU formalizes a priority role inside a long-running partnership.

The SK Hynix problem

SK Hynix is the other name Su highlighted, and it is the biggest player in the space. The company holds a market share of approximately 57-58% in the HBM segment.

SK Hynix has primarily partnered with Nvidia, AMD’s chief rival in AI accelerators. That arrangement has limited AMD’s options, helping explain AMD’s deeper ties with Samsung.

What this means for AMD and the AI chip race

AMD’s heavy reliance on Samsung concentrates exposure. Any production hiccup or yield issue on Samsung’s HBM4 lines could ripple directly into AMD’s accelerator roadmap.

With roughly 57-58% of the HBM market, any shift in how SK Hynix allocates capacity between Nvidia and others could reshape the competitive picture.

Also worth tracking: whether AMD’s MOU with Samsung turns into firm, multi-year commitments that match Su’s new planning window of three to five years.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.
AMD CEO Lisa Su calls Samsung and SK Hynix important for long-term supply chain
AMD CEO Lisa Su calls Samsung and SK Hynix important for long-term supply chain

AMD is stretching its memory planning horizon to three to five years as AI demand outruns high-bandwidth memory production

AMD CEO Lisa Su has named Samsung and SK Hynix as important partners for the chipmaker’s long-term supply chain. The comment comes as the AI boom turns a specialized type of memory into one of the scarcest commodities in tech.

That memory is high-bandwidth memory, or HBM. Su says AMD now has to plan its supply years further out than it ever did before.

A longer planning horizon for a tighter market

Su visited Taiwan and South Korea in early October 2026. The trip was aimed at locking in long-term supply agreements with memory makers.

Su emphasized that AMD’s capacity planning has to stretch from the previous one to two years out to a new timeline of three to five years.

On October 6, 2026, Su announced that AMD plans to significantly increase supply in 2027. The ramp reflects a strategic pivot as the company responds to how fast the AI market is growing.

Advertisement

AMD is working with memory chipmakers to address HBM supply constraints that are expected to persist through 2028.

Samsung gets the priority seat

The backbone of AMD’s memory strategy is a deal with Samsung Electronics. On March 18, 2026, the two companies signed a memorandum of understanding naming Samsung as AMD’s priority supplier for sixth-generation HBM4.

That HBM4 is destined for AMD’s new Instinct MI455X AI accelerators. The same agreement also covers advanced DDR5 memory for AMD’s EPYC “Venice” server processors.

Samsung began shipping HBM4 in February 2026. It was already the primary supplier of HBM3E, the prior generation, for AMD’s earlier chip models.

The relationship is not new. AMD and Samsung have worked together for nearly two decades, so the MOU formalizes a priority role inside a long-running partnership.

The SK Hynix problem

SK Hynix is the other name Su highlighted, and it is the biggest player in the space. The company holds a market share of approximately 57-58% in the HBM segment.

SK Hynix has primarily partnered with Nvidia, AMD’s chief rival in AI accelerators. That arrangement has limited AMD’s options, helping explain AMD’s deeper ties with Samsung.

What this means for AMD and the AI chip race

AMD’s heavy reliance on Samsung concentrates exposure. Any production hiccup or yield issue on Samsung’s HBM4 lines could ripple directly into AMD’s accelerator roadmap.

With roughly 57-58% of the HBM market, any shift in how SK Hynix allocates capacity between Nvidia and others could reshape the competitive picture.

Also worth tracking: whether AMD’s MOU with Samsung turns into firm, multi-year commitments that match Su’s new planning window of three to five years.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.