Samsung Electronics has announced it is discussing short-term cooperation with NVIDIA on HBM4 memory and foundry services.
This is a strategically timed move to secure a significant role in NVIDIA's next-generation AI platform, known as 'Rubin'. The discussion comes just after NVIDIA CEO Jensen Huang officially confirmed that Samsung, alongside SK hynix and Micron, is qualified and in production for HBM4 memory. This confirmation is crucial; it transforms Samsung's technical achievements, like its first HBM4 shipments in February 2026, from a simple milestone into a tangible, near-term revenue opportunity.
But the story isn't just about memory. The 'foundry services' part of the announcement is equally important. The world's leading chip manufacturer, TSMC, has stated it will be a 'long time' before it can meet all customer demand. This creates a significant bottleneck in advanced chip production and packaging. Samsung is positioning its foundry business as a viable second option for NVIDIA, offering an integrated package that combines cutting-edge memory and logic chip manufacturing. This 'all-in-one' approach is what Samsung calls its 'Total AI Solution'.
Tracing back the events, this moment was built on several key steps. First, Samsung had to regain trust after falling behind in the previous memory generation. They achieved this by successfully passing NVIDIA's HBM3E qualification in late 2025. Second, they proved their technical roadmap by starting HBM4 mass production early this year. Third, public acknowledgments from NVIDIA about Samsung's capabilities, both in memory and foundry, at major events like GTC 2026, paved the way for these direct, high-stakes negotiations.
The broader market context also plays a vital role. The global demand for AI chips remains incredibly high, partly sustained by U.S. export policies that keep the market in a state of structural scarcity. For a company like NVIDIA, which sits at the center of the AI boom, relying on a single supplier is risky. Therefore, diversifying its supply chain across multiple partners for both memory and foundry is not just a preference but a strategic necessity. Samsung's push for 'short-term cooperation' is a direct attempt to capitalize on this need, turning its comprehensive technical portfolio into confirmed orders for the Rubin platform.
- HBM (High Bandwidth Memory): A type of high-performance memory stacked vertically to provide faster data transfer speeds, essential for training large AI models.
- Foundry: A semiconductor manufacturing plant that produces chips designed by other companies. It's a 'factory for hire' for chip designers.
- Rubin Platform: NVIDIA's next-generation AI computing platform, set to succeed the 'Blackwell' platform, which requires even more advanced memory and processing power.