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 market is likely underestimating how much HBM4 shifts bargaining power from GPU vendors to memory suppliers. In the near term, NVDA benefits because a preferred supply relationship reduces the risk of shipment delays and lets it preserve product cadence into Vera Rubin; that is more a schedule-risk story than an immediate margin story. The larger economic winner is the supplier that gets the first-call allocation, because HBM remains one of the few semis where incremental capacity can still translate into durable pricing power rather than just volume growth.

Second-order, this is a capex reallocation event inside the memory stack. Every dollar diverted into HBM is a dollar not going into commodity DRAM/NAND, so the winners can be more concentrated than consensus expects while weaker memory players face a slower recovery in non-AI segments. For hyperscalers, rising memory content per accelerator raises total cluster cost and can force a slower rollout of frontier training capacity, which paradoxically may support pricing discipline longer than the equity market is modeling.

The key risk is execution timing: if HBM4 qualification slips, the narrative stays bullish but the earnings impact gets pushed out 1-2 quarters. Over 6-18 months, the trade only works if HBM supply remains constrained and Nvidia’s platform transition keeps memory content rising; if compute, networking, or power become the binding constraints instead, HBM multiple expansion could stall. The consensus is probably right on direction but too aggressive on duration; a lot of the value is already reflected in the obvious AI beneficiaries, while the real upside may be in the supplier with the cleanest allocation and yield advantage.

Contrarian view: the current read may overstate how durable a single-customer HBM4 relationship is. If SK Hynix wins too much share, Samsung and Micron have stronger incentives to price aggressively, subsidize qualification, or accelerate capacity, which could compress the supercycle economics faster than investors expect. Watch for any NVDA commentary on memory availability improving, because that would be the fastest signal that the scarcity premium is peaking.