Microsoft And Google Escalate The Memory Supply Fight In South Korea As HBM And LPDDR Allocation Tightens
An intensifying shortage of high value memory products is pushing hyperscalers into a very visible, very human procurement battle, with executives effectively relocating to South Korea to secure long term supply agreements for critical components like HBM and LPDDR. According to this Seoul Economic Daily report, Microsoft and Google procurement leaders have been spending extended time on the ground in Korea to negotiate supply terms with Samsung and SK hynix, as capacity remains constrained and allocation decisions increasingly determine who can scale AI infrastructure on schedule.
The report describes a recent Microsoft visit to SK hynix headquarters to pursue a memory focused LTA and price negotiations. When the supplier reportedly responded that it would be difficult to supply memory products under Microsoft’s requested conditions, a Microsoft executive is said to have left the meeting in anger. In practical terms, that anecdote signals how tight the negotiating window has become when HBM and related memory products are already heavily committed, and when suppliers hold the leverage because production lines are operating at full utilization.
The same report frames the current environment as a scramble to lock in HBM, DRAM, and enterprise SSD supply for AI chips and data centers, driven by the reality that modern accelerators and platforms need high bandwidth memory for performance and that AI data centers lean on low power memory for efficiency at scale. It also underlines a hard constraint for buyers: only a small set of companies can produce leading edge HBM and LPDDR at the volumes demanded by hyperscalers, and forward capacity is already spoken for into next year.
Google’s situation appears even more escalated. The report says Google has taken punitive action after being told that additional HBM volume was effectively not available from suppliers it approached, and it highlights internal accountability tied directly to whether procurement secured LTAs early enough. It also notes that Google’s TPUs currently receive a significant share of their HBM supply from Samsung, while demand for TPUs has reportedly exceeded expectations, increasing pressure to source incremental supply elsewhere.
Executives who fail to secure memory supply get fucking fired LMAO https://t.co/zh25EI1Nll pic.twitter.com/YWkYlCPAPR
— Jukan (@jukan05) December 25, 2025
From an industry operations lens, this is what a supply chain war room looks like at hyperscaler scale: open ended demand signals, aggressive allocation negotiations, and organizational changes to put decision makers closer to the manufacturing base. The report points to a broader hiring push across Asia for memory sourcing and supply chain roles, with Google recruiting for data center memory sourcing strategy positions and Meta pursuing memory silicon sourcing managers to strengthen technical collaboration and improve supply security.
For gamers and PC enthusiasts, the key takeaway is not that your next DDR5 kit vanishes overnight, but that AI driven memory competition can reshape pricing, priorities, and product cycles across the whole ecosystem over time. When hyperscalers are willing to accept “any volume at any price” for scarce memory categories, suppliers naturally optimize allocation toward the highest strategic value contracts, and that pressure can cascade into broader market dynamics that enthusiasts eventually feel in platform costs and availability.
Do you think hyperscalers will normalize longer LTAs for memory the way they did for GPUs, or will we keep seeing quarter to quarter allocation drama driving the market?
