Samsung-Broadcom pact valued at up to $200 billion
Samsung Electronics said on July 25 that it signed a memorandum of understanding with Broadcom covering up to $200 billion through 2030 across high-bandwidth memory, sub-2-nanometer foundry services and advanced packaging. The deal pairs Broadcom's application-specific integrated circuit design work with Samsung's integrated device manufacturer capabilities, and is expected to include HBM supply for Broadcom's next-generation AI accelerators. Broadcom's communications chips for high-speed data transfer are slated for production on Samsung's sub-2nm process, with mass production anticipated at Samsung's Pyeongtaek campus. Vice Chairman Jun Young-hyun said Samsung's combination of memory, logic and packaging technology had become "core competitiveness" in the AI era. The pact adds to Samsung's recent foundry wins, including a 2025 Tesla AI6 contract and a recent MOU with Anthropic.
SK Group commits $750 billion, anchored by Nvidia partnership
SK Group unveiled cooperation agreements totalling $750 billion, including a five-year long-term memory supply deal with Nvidia valued at more than $500 billion covering HBM4 and joint development of next-generation AI memory. The remaining roughly $250 billion bundles additional big-tech contracts, including mid- to long-term cooperation with Microsoft on memory for AI servers. SK Hynix and Nvidia also said they would co-develop and optimise HBM solutions for large language model training and physical AI applications. SK Group Chair Chey Tae-won said Nvidia's projected memory requirements over the next five years could prove higher than current estimates and that Broadcom's CEO had indicated the company's memory demand was likely to surpass available supply.
SK Telecom plans 2-gigawatt data centre powered by HBM4
As part of the Nvidia-linked initiative, SK Telecom plans to build a 2-gigawatt data centre using Nvidia's Vera Rubin chips paired with SK Hynix's HBM4 memory, with the facility due online in 2027. SK Hynix and Nvidia framed the project as part of an effort to accelerate large-scale AI infrastructure development alongside next-generation memory supply.
Apple and Micron spar over Chinese memory imports
Apple has lobbied the White House, Commerce Secretary Howard Lutnick and Treasury Secretary Scott Bessent to allow it to source memory from ChangXin Memory Technologies and Yangtze Memory Technologies for Apple devices sold outside the United States, according to The Wall Street Journal. Micron, the only sizeable American memory maker remaining, has pressed the same officials in the opposite direction, with CEO Sanjay Mehrotra warning that Chinese memory firms in any supply chain could undercut the domestic industry. Apple argues Chinese chips would ease a memory crunch that has pushed prices up roughly fourfold over the past year, per TechInsights, and accuses Micron of gouging, pointing to gross margins above 80%. A White House spokesman said the administration would pursue economic relief while "safeguarding our national security." Both firms remain on US entity lists or military-company designations.
Industry backdrop and diplomatic framing
The Korean announcements were made during South Korean President Lee Jae-myung's visit to San Francisco for an AI summit that included meetings with OpenAI's Sam Altman, Nvidia's Jensen Huang, Anthropic's Dario Amodei and Broadcom's Hock Tan. Presidential chief of staff for policy Kim Yong-beom told reporters in San Francisco that the two Korean chipmakers and the US tech firms had agreed to "pursue cooperation" in the sector. Samsung Executive Chairman Jay Y. Lee separately visited OpenAI's headquarters on July 25 to discuss expanded AI semiconductor cooperation. The combined $950 billion figure frames Korean memory leaders and their US counterparts as long-term infrastructure partners rather than transactional suppliers, with memory contracts shifting from one-year cycles to agreements of five years or more.
Share this article
