NVIDIA to Meet with Samsung and SK Hynix to Discuss Physical AI Collaboration

Three Giants Convene as Physical AI Collaboration Looms

According to Sina Finance, senior NVIDIA executives are set to meet with counterparts from South Korea's two semiconductor powerhouses, Samsung Electronics and SK Hynix, to discuss collaboration in the field of physical AI. The news has quickly drawn intense attention from the industry, as a deep partnership among the three companies could have far-reaching implications for the global AI hardware supply chain.

Physical AI: NVIDIA's Next Strategic Priority

Physical AI is a core concept that NVIDIA CEO Jensen Huang has repeatedly emphasized in recent years, referring to artificial intelligence systems capable of perceiving, understanding, and interacting with the physical world. Unlike purely digital AI, physical AI encompasses critical domains such as autonomous driving, robotics, industrial automation, and smart manufacturing.

Huang has stated on multiple occasions that physical AI will be "the next wave of artificial intelligence," with a market size potentially reaching trillions of dollars. NVIDIA has already launched several key products around this vision, including the Isaac platform for robotics development, the DRIVE series of chips for autonomous driving, and the Omniverse platform for digital twin simulations.

However, the hardware demands of physical AI are far more complex than those of large language models — requiring not only powerful GPU computing power but also high-bandwidth, low-latency memory solutions to support real-time data processing. This is precisely where Samsung and SK Hynix hold core advantages.

Why Samsung and SK Hynix?

Samsung Electronics and SK Hynix are the world's two most important memory chip manufacturers, commanding a dominant position in the HBM (High Bandwidth Memory) market.

SK Hynix is currently NVIDIA's primary HBM chip supplier, with its HBM3E products already deployed at scale in NVIDIA's latest H200 and B200 series GPUs. Samsung Electronics, meanwhile, is accelerating its pursuit, actively pushing mass production certification of its HBM3E products in a bid to secure a larger share of NVIDIA's supply chain.

The focus of this meeting on "physical AI" suggests that the collaboration may extend beyond traditional chip supply relationships into deeper technological synergies. Physical AI scenarios introduce entirely new requirements for edge computing, low-power high-performance chips, and heterogeneous computing architectures. The three parties are expected to explore cooperation in the following areas:

• Customized HBM Solutions: Specialized memory chips optimized for robotics and autonomous driving scenarios
• Advanced Packaging Technology: Tighter integration and packaging of GPUs with HBM to improve energy efficiency
• Edge AI Chips: Compact, low-power solutions for physical AI endpoint devices
• Supply Chain Coordination: Ensuring production capacity for large-scale commercialization of physical AI

Industry Landscape and Competitive Dynamics

The backdrop of this three-way discussion is the increasingly fierce global AI competition. The United States, China, South Korea, and other nations all regard AI as a national strategic technology priority, and physical AI — as a key area where AI and manufacturing converge — is a focal point of contention.

For SK Hynix, deepening its partnership with NVIDIA helps solidify its leading position in the HBM market. For Samsung, this meeting represents a crucial opportunity to catch up and potentially redefine its role in NVIDIA's supply ecosystem through the emerging physical AI track.

At the same time, other memory manufacturers such as Micron Technology are also aggressively positioning themselves in the HBM market, and the competitive landscape is evolving rapidly. NVIDIA's decision to engage both Korean giants simultaneously also reflects its strategic consideration of "diversified supply" to avoid excessive dependence on any single supplier.

Outlook: The Physical AI Era Accelerates

From a broader perspective, the collaboration discussions between NVIDIA, Samsung, and SK Hynix signal that the AI industry is rapidly transitioning from the phase of "training large models" to a new stage of "AI entering the physical world." When AI is no longer just a chatbot running in the cloud but is driving real-world robots, vehicles, and factories, the entire semiconductor supply chain will face restructuring.

This meeting among industry titans may well be an important prelude to the large-scale commercialization of physical AI. The specific details and progress of subsequent collaborations deserve continued attention from the industry.