Jensen Huang Said Nvidia Will Be the First Customer for HBM4. Here's the AI Memory Stock That Has Reportedly Locked Up 70% of Those Orders.

The explosive growth of artificial intelligence (AI) has ignited a powerful supercycle in the memory semiconductor landscape. As large language models (LLMs) and generative AI systems scale to trillions of tokens, the bottleneck is shifting from raw compute to the speed and capacity of data movement.

Nvidia (NVDA 1.97%) is a dominant force in this ecosystem, not merely as the leading designer of graphics processing units (GPUs) but as the primary driver of demand for specialized high-bandwidth memory (HBM). The company’s GPUs support the majority of hyperscale training clusters and inference workloads, forcing memory producers to align their roadmaps with Nvidia’s performance targets.

Three companies possess the technology and manufacturing expertise to produce HBM4 at scale: SK Hynix (SKHY +1.13%), Samsung, and Micron Technology (MU +0.04%). Each company is aggressively expanding capacity and refining its manufacturing processes to meet Nvidia’s specifications.

Nvidia CEO Jensen Huang is taking the memory supercycle incredibly seriously. Over the last couple of years, Huang has quietly dropped some breadcrumbs that can be traced to Nvidia’s favorable memory suppliers. Let’s take a look at what Huang has to say about the memory bottleneck, and explore which AI memory stock you may want to put on your radar right now.

Why is memory becoming so important for AI?

Memory is emerging as one of the most important components in the chip stack because AI applications are fundamentally memory-bound. Training and running models require transferring enormous volumes of data between hundreds of thousands of GPUs at extremely low latency.

Conventional DRAM struggles to keep pace with these bandwidth demands, leading to underutilized compute resources and longer training times. HBM solves this by stacking DRAM dies and stitching them together through vertical interconnects. This delivers bandwidth that is meaningfully higher than that of traditional memory solutions while consuming less power.

HBM4 represents the next step forward in the memory evolution. Nvidia requires HBM4 to power its next-generation platforms because these chips enable larger model sizes, faster token generation, and more efficient scaling of GPU clusters. In particular, Nvidia’s Vera Rubin architecture is expected to rely heavily on HBM4 to deliver the performance leap customers anticipate.

SK Hynix is expected to be the leader of HBM4

During CES back in January, Huang said that Nvidia will be the “only customer” of HBM4 for quite some time. While that’s great news for memory suppliers, it’s especially beneficial for one company in particular.

In early June, Nvidia and SK Hynix announced a multiyear partnership focused on co-developing advanced memory solutions for AI factories. This isn’t entirely surprising, as Huang has been relentlessly pursuing SK Hynix’s memory wafers for years.

According to reports dating back to 2024, Huang had asked SK Hynix to expedite its HBM4 manufacturing by six months. While SK Hynix plans to double its wafer capacity by 2030, Huang warns this may not be enough supply given the explosion of AI workloads. Huang explains:

AI factories are the engines of the next industrial revolution, and advanced memory is essential to their performance. SK Hynix has been an extraordinary partner to Nvidia, playing a central role in delivering advanced memory technologies for Nvidia AI computing platforms. Together, we will codevelop the next generation of memory for AI factories and support the accelerating global expansion of AI infrastructure — from frontier model training to agentic and physical AI.

Against this backdrop, industry analysts estimate that SK Hynix could lock in between 50% and 70% of Nvidia’s anticipated HBM4 orders. This outsize allocation would provide SK Hynix with durable revenue tailwinds as the company stands to capture a disproportionate share of HBM revenue while its competitors ramp up their own capacities. With this in mind, SK Hynix’s strengthening relationship with Nvidia could be seen as a subtle competitive advantage.

For investors seeking exposure to the memory pocket of the AI infrastructure theme, SK Hynix appears to be one of the clearest beneficiaries of sustained demand over the coming years. This makes the stock worth a look as growth investors continue to take profits and rotate out of the obvious winners seen so far during the AI revolution.