A deepening global memory shortage is reshaping the balance of power in the tech industry, pushing some of the world’s biggest companies into intense, on-the-ground negotiations in South Korea. With high-demand memory products like HBM (high-bandwidth memory) and LPDDR in short supply, major cloud and AI players including Microsoft, Google, and Meta are working directly with leading suppliers such as Samsung, SK hynix, and Micron to secure long-term supply.
South Korea has effectively become a central battleground for these talks. Procurement leaders from large tech firms are spending extended time near the headquarters of top memory makers as discussions drag on and supply commitments become harder to lock down. The reason is simple: production capacity for HBM at key suppliers is already running at maximum levels, leaving little room for additional volume—especially on the timelines hyperscalers want.
Tensions are reportedly boiling over. In one recent example, Microsoft executives visited SK hynix to negotiate a new long-term memory supply agreement. When told it would be “difficult” to meet Microsoft’s requested conditions, one executive allegedly walked out of the meeting in frustration. It’s a snapshot of the pressure facing hyperscalers as memory becomes one of the biggest constraints on AI expansion.
With supply tight and demand surging, some of these buyers are now placing open-ended orders, signaling they’ll take whatever HBM they can get—at virtually any price. This is particularly true for companies racing to scale AI infrastructure, where HBM is a critical component for modern accelerators and training workloads.
Google, meanwhile, appears to be taking an even tougher internal stance. The company relies heavily on HBM for its custom AI chips, known as TPUs. As availability worsened, Google reportedly sought additional HBM volume from SK hynix and Micron, only to receive responses described as effectively “impossible.” In the wake of that setback, Google is said to have fired a procurement executive, blaming them for not securing long-term agreements earlier. Samsung reportedly supplies around 60 percent of the HBM used on Google’s TPUs, underscoring how concentrated and fragile these supply chains can be.
This kind of disruption doesn’t just trigger conflict—it also creates opportunity. Big tech companies are now expanding supply-chain teams across Asia, attempting to strengthen sourcing strategy and reduce the chance of being caught short again. Google has reportedly posted a role for a Global Memory Commodity Manager focused on data-center memory products such as DRAM and NAND flash. Meta is also said to be planning hires for dedicated memory sourcing manager positions, reflecting how procurement expertise is becoming a competitive advantage in the AI era.
The memory crunch is also rippling beyond cloud and AI. Apple is reportedly being pushed into paying a steep premium for LPDDR5X memory—estimated at roughly 230 percent above previous levels. Adding to the pressure, some of Apple’s existing long-term supply agreements with major memory providers are expected to expire in January, which could open the door to additional pricing increases.
Taken together, these developments point to a new reality: memory supply—especially HBM and advanced mobile DRAM—is no longer just a component cost. It’s becoming a strategic choke point that can determine how fast AI platforms scale, how quickly devices ship, and how aggressively companies can compete on performance and price.
