Shares of Samsung Electronics (KRX:005930) saw their steepest single-day decline in nearly two decades on Tuesday, closing 13.4% lower at 220,000 won. The sell-off erased the initial enthusiasm surrounding the company's landmark $200 billion collaboration agreement with Broadcom (NASDAQ:AVGO), as investors digested the deal's composition and its implications for Samsung's foundry ambitions.
In New York, Broadcom shares also retreated, trading 1.6% lower at $377.02 by mid-morning. The broader market reaction extended to other memory-related names, with SK Hynix (KRX:000660) falling 14.7% and the KOSPI index shedding 10.8%. Analysts pointed to concerns over AI financing risks and accelerating Chinese memory capacity as contributing factors to the widespread sell-off.
Memory Dominates the Deal Structure
The central issue for investors is the deal's product mix. JPMorgan (NYSE:JPM) analyst Harlan Sur estimates that memory products will account for 90% to 95% of Broadcom's purchases under the memorandum of understanding. Foundry wafers, which many had hoped would be a significant component, are projected to represent only 5% to 10% of the total.
At the $200 billion baseline, this breakdown implies foundry work worth roughly $10 billion to $20 billion over the agreement's five-year horizon, or about $2 billion to $4 billion annually. While still material, this is far below the headline figure and comparable in scale to Samsung's previously disclosed $16.5 billion chip supply contract with Tesla (NASDAQ:TSLA).
Market Implications and Competitive Context
The distinction matters greatly in Samsung's ongoing contest with Taiwan Semiconductor Manufacturing Company (NYSE:TSM) for advanced foundry leadership. While the Broadcom deal can support factory utilization and help Samsung qualify new customers, it does not represent the $200 billion transfer of advanced foundry demand from TSMC that some investors had anticipated. TSMC's U.S.-listed shares fell 3.9% during Tuesday's trading session.
For Broadcom, the sheer scale of the arrangement remains substantial. The company reported $10.8 billion in AI semiconductor revenue for its fiscal second quarter, with total quarterly revenue reaching $22.2 billion. A straight-line calculation of $10 billion in quarterly collaboration value would equate to roughly 93% of AI revenue and 45% of total revenue. While not directly comparable, these figures underscore the enormous supply requirements involved.
Deal Details and Investor Concerns
Samsung's official release described the arrangement as a memorandum of understanding, with the companies estimating collaboration exceeding $200 billion through 2030. However, they did not disclose annual volumes, pricing, or minimum purchase commitments, leaving significant uncertainty. The scope includes high-bandwidth memory (HBM) used alongside AI processors, contract chip manufacturing on Samsung's two-nanometer and smaller processes, and advanced packaging to more closely integrate memory and logic components.
"AI is driving unprecedented demand for tightly integrated semiconductor technologies," said Young Hyun Jun, head of Samsung's chip business, in a statement. He cited memory, logic, and advanced packaging as linked components of future AI systems.
Outlook and Key Risks
Samsung is scheduled to publish detailed second-quarter results on July 30 at 10:00 a.m. KST. Its preliminary guidance showed sales near 171 trillion won and operating profit of approximately 89.4 trillion won. Investors will be closely watching for updates on foundry losses, HBM shipment details, and any timetable for Broadcom purchases.
Key risks include the fact that the $200 billion figure remains an estimated MOU value rather than a disclosed shipment schedule, with pricing and minimum volumes still unknown. Weaker AI spending or faster Chinese capacity growth could also erode memory margins. For now, the cleaner investment thesis remains Samsung's memory franchise, while a broader foundry re-rating will require firm wafer volumes, stable yields, and evidence of profitable two-nanometer production.
