AMD Rocketed this Week — But It’s Not the Only AI Trade at Hand

AMD shares took off like a rocket on Monday after the company confirmed a multi-year supply deal with OpenAI. 

The agreement covers up to 6 gigawatts of AMD GPUs and includes warrants allowing OpenAI to acquire roughly 10% of AMD if certain deployment targets are met. Investors saw the move as confirmation that OpenAI is broadening its chip sourcing beyond Nvidia, sending AMD shares sharply higher.

Our June MegaTrends entry is now sitting +61% higher, one of the fastest movers in the portfolio this year. The gains reinforce what we’ve been tracking for months: enterprise demand for AI compute continues to expand.

But there’s more to this trend than GPUs. As new data centers come online and model sizes multiply, a parallel wave of suppliers is seeing rising orders.

The same expansion creating GPU demand is also tightening supply in the layers beneath it, particularly in memory and storage.

These AI pin-action names provide the high-bandwidth memory, storage capacity, and controller technology that keep AI servers running…

The Broadening Trade: Memory and Storage

Every GPU depends on lightning-fast access to data. That dependency is turning memory and storage into the next major investment theme inside the AI buildout. 

Contract DRAM and NAND pricing climbed up to 20% this year as suppliers exercised capacity discipline and AI demand increased.

Memory stocks are surging:

• SK Hynix (000660.KS) – The world’s largest supplier of high-bandwidth memory (HBM), which feeds data to Nvidia’s H100 and AMD’s MI300 accelerators. Its latest HBM3E chips can move over 1.2 terabytes of data per second.
  

• Micron (MU) – The only U.S.-based HBM producer, ramping next-generation HBM4 and expanding domestic DRAM capacity. Contract pricing for memory is up roughly 15–20% year over year, reflecting tight supply.
  

• Western Digital (WDC) – Major player in enterprise solid-state and hard-disk storage. As AI models grow, data-center customers are buying more drives to store training datasets, inference logs, and retraining archives.
  

• Seagate (STX) – Leading supplier of nearline hard drives used by hyperscalers for large-scale AI data storage and backups.
  

• Montage Technology (688008.SS) – Developer of memory controllers supporting Compute Express Link, a standard that allows multiple processors to share pools of memory across servers.

You can see the stock boom in this sector below – nearly all have doubled (or more) in value over the last year:

What Comes After the Memory Boom?

The setup for memory and storage remains strong. Demand is rising, pricing power has returned, and producers are exercising more discipline than in past cycles. 

But we’re also considering how the balance could evolve:

• Capacity will expand. SK Hynix, Samsung, and Micron have new fabs scheduled to open in 2026 and 2027. Once production ramps, pricing could begin to stabilize.
  

• Efficiency is improving. Model optimization and compression are reducing the amount of memory needed to train or run AI systems, which could ease demand over time.
  

• Compute Express Link adoption will take time. Pooling memory across servers is promising, but operating systems and data-center software need to catch up.

The sharpest investors are already preparing for the next layer of opportunity: the companies that build the infrastructure supporting both compute and memory.