AMD’s Lisa Su Meets Samsung Chief as AI Chip Demand Strains Memory Supply

AMD CEO Lisa Su is scheduled to meet Samsung Electronics’ semiconductor chief in South Korea as shortages of AI-related memory intensify. The meeting places supply security at the center of a relationship spanning processors, advanced packaging and high-bandwidth memory, or HBM. Bloomberg reported that the encounter is the latest in-person discussion between the companies’ leaders as AI infrastructure demand accelerates.

The visit also reflects AMD’s push to secure components beyond its own accelerator designs. Samsung is one of the world’s largest memory manufacturers, while AMD depends on a broader manufacturing network for the chips and packaging needed to turn demand into shipments. A Samsung plant visit and discussions about expanding cooperation were described by WMBD as part of Su’s itinerary, citing an unnamed source.

AI chip demand is exceeding production capacity, creating pressure across the supply chain rather than at a single manufacturing step. High-bandwidth memory is particularly important because it feeds large volumes of data to graphics processors and AI accelerators. Samsung’s HBM output therefore matters directly to AMD’s ability to increase accelerator shipments.

Chosun reported that Su secured Samsung’s HBM4 supply, but the broader meeting coverage did not provide terms, volumes or a formal announcement. The reported development still highlights the commercial value of memory allocation as chipmakers compete for capacity. AMD’s effort comes as cloud companies and AI developers place large orders for computing hardware, tightening access to both advanced logic chips and the memory attached to them.

Su has described chip demand as “very high” and said it will remain elevated for the next few years. CNBC, Bloomberg AI and Seekingalpha all tied that outlook to the continuing buildout of AI infrastructure, while Yahoo Finance characterized the forecast as “very, very high.” The comments frame the supply squeeze as a structural planning challenge rather than a short-lived ordering spike.

AMD is responding by working to add supply and by supporting capacity expansion at manufacturing partners. TSMC’s expansion received praise from Su, according to Seekingalpha, but additional wafer capacity alone does not solve constraints in memory production, advanced packaging or testing. The result is a coordinated race among chip designers, foundries and memory suppliers to synchronize several scarce inputs.

Samsung’s role in the AI hardware chain is expanding as customers seek alternatives and additional sources for HBM. A supply commitment involving HBM4 would give AMD another resource for future accelerator generations and strengthen Samsung’s position in a market led by intense demand from data-center customers.

The companies’ talks also carry competitive significance. AMD is challenging Nvidia in AI accelerators, while Samsung competes with SK Hynix and Micron for advanced memory orders. A closer AMD-Samsung relationship could reshape sourcing decisions as customers weigh performance, delivery schedules and supply diversification. The dinner meeting between Lee and Su, reported by Thelec, adds a relationship-building dimension to negotiations that otherwise center on production targets and technical qualification.

AMD’s supply concerns show why AI hardware expansion remains tied to industrial capacity. Demand for accelerators is rising alongside investment by cloud providers and enterprises, but each system requires more than a processor. Memory bandwidth, packaging capacity, networking equipment and power infrastructure all determine how quickly new data-center capacity can come online.

Su has also urged AI companies to cooperate while warning that demand will remain high, Bloomberg AI reported. That message reflects a supply chain where competing firms share dependence on a limited group of manufacturers. It also comes amid wider debate about AI safety, showing that the industry’s expansion is being shaped by both physical constraints and questions about how quickly new systems should be deployed.

The immediate test is whether AMD and Samsung turn executive discussions into confirmed supply agreements, production schedules and qualified HBM4 shipments. Public evidence of a meeting indicates active engagement, while the reported HBM4 arrangement points to a possible step forward. Neither development, as presented in the coverage, establishes shipment volumes or the effect on AMD’s near-term revenue.

Investors and customers will watch AMD’s accelerator availability, Samsung’s HBM4 ramp and TSMC’s capacity additions together. If memory output remains constrained, extra processor capacity will leave AMD unable to satisfy all demand. If suppliers expand in sync, AMD can use the prolonged AI buildout to narrow the gap with Nvidia and strengthen its position in data-center computing.