Nvidia CEO Jensen Huang has confirmed that all three major memory manufacturers,Samsung, SK Hynix, and Micron,have been certified to supply fourth-generation high-bandwidth memory (HBM4) chips for the company's upcoming Vera Rubin data center platform. This move addresses growing concerns about whether memory shortages could delay the rollout of Nvidia's next-generation artificial intelligence (AI) infrastructure hardware.

Why Does Memory Supply Matter for Nvidia's AI Chips?

High-bandwidth memory, or HBM, is a critical component in modern AI accelerators. It allows data to move between the processor and memory at extremely high speeds, which is essential for training and running large language models (LLMs) and other AI workloads. The shortage of HBM chips has been one of the most significant bottlenecks in the AI infrastructure buildout, with lead times stretching as far out as 2028 according to some suppliers.

Huang explained his proactive approach to managing these supply chain challenges during a press briefing, stating that he was "constantly just looking across the entire life cycle of the technology and manufacturing across the supply chain" to "prepare our company for growth." He noted that when scaling silicon photonics technology and other advanced components, the world's current manufacturing capacity falls short of what's needed, so Nvidia works directly with suppliers to help them build capacity in advance.

Huang

What Are Nvidia's Supply Chain Strategies?

Huang's approach to securing components goes beyond simple purchasing agreements. The company employs multiple strategies to ensure suppliers can meet demand:

• Direct Investment: Nvidia has invested billions in suppliers, including $2 billion into Marvell Technology, a connectivity chip maker essential for disaggregated AI data center architecture.
• Prepayment and Forecasting: The company provides advance payments and long-term demand forecasts to help suppliers plan manufacturing capacity expansion.
• Equity Stakes: When technology capacity is particularly constrained, Nvidia may take equity positions in suppliers like Ayar Labs, Lumentum, Corning, and Coherent, which specialize in co-packaged optics (CPO) technology.

This strategy reflects what Huang called his "chimaek summit" approach, referencing a famous incident in which he, along with influential Asian business leaders, secured initial memory supply commitments during an informal dinner at a fried chicken restaurant.

When Will Vera Rubin Launch and What Should We Expect?

During Computex 2026, Huang revealed that Vera Rubin is already in full production, with initial shipments expected later in 2026. Prior reports had suggested that memory constraints and power consumption issues could reduce Vera Rubin shipments from 29 percent to 22 percent, but the certification of all three major memory suppliers appears to address at least part of that concern.

Samsung began mass production of HBM4 chips in February 2026 and has already announced samples of an enhanced version called HBM4E. SK Hynix and Micron have also met Nvidia's required threshold for data transfer speeds of 10 gigabits per second, though SK Hynix had to push back its production timeline.

The broader AI chip market remains under intense pressure. Taiwan Semiconductor Manufacturing Company (TSMC) CEO C.C. Wei warned this week that demand for AI chips could exceed supply for several years, underscoring just how critical Huang's supply chain management efforts have become.

How Is Nvidia Supporting Its Broader Supplier Ecosystem?

Beyond memory, Huang has made strategic investments across Nvidia's supply chain to ensure the company can scale its most advanced technologies. During his Computex keynote, he called Marvell Technology "the next trillion-dollar company," highlighting the critical role connectivity infrastructure plays in modern AI data centers. Marvell reported record quarterly revenue of $2.418 billion in May 2026, up 28 percent year-over-year, with management describing AI-related bookings as "exceptional".

The company is also investing heavily in optical component suppliers as it prepares to launch co-packaged optic versions of its networking products. These optical components allow data to move between chips and across data centers with minimal latency and power consumption, making them essential for the next generation of AI infrastructure.

Huang's willingness to invest directly in suppliers reflects a broader recognition that Nvidia's growth depends not just on its own manufacturing capabilities but on the entire ecosystem of companies that provide components and services. By securing supply agreements with Samsung, SK Hynix, and Micron, Nvidia is taking concrete steps to ensure that Vera Rubin can reach customers on schedule, even as global chip demand continues to surge.