The Forgotten AI Investment: Memory

Monday in the US brought another big AI moment.

AMD shares exploded ~25% higher after announcing a multibillion-dollar chip deal with OpenAI.

The partnership will see OpenAI deploy up to six gigawatts of AMD GPUs over multiple years, starting with one gigawatt in 2026. As part of the agreement, OpenAI could acquire up to 10% of AMD if certain milestones are met.

But while everyone fixates on flashy chip deals, a big story hides underneath.

The memory makers everyone forgot about

While investors chase trillion-dollar tech valuations, something extraordinary happened to the ‘boring’ memory suppliers.

Over the past three months, these companies aren’t making headlines. They don’t have celebrity CEOs. They just make the essential storage that keeps AI systems running.

Yet their one-year performance tells a different story entirely:

• Seagate Technology [NASDAQ:STX]: Market cap ~$52 billion, +194%
• Western Digital [NASDAQ:WDC]: Market cap ~$44 billion, +120%
• Micron Technology [NASDAQ:MU]: Market cap ~$210 billion, +87%

These aren’t small companies riding a speculative wave.

We’re talking about established players with massive market capitalisations experiencing explosive growth typically reserved for early-stage companies.

The AI infrastructure buildout isn’t slowing down. It’s accelerating. And every data centre, every AI training facility, every edge computing deployment needs massive amounts of reliable storage.

The second-order effects of AI
(and one tiny ASX stock)

Here’s where things get interesting for forward-thinking investors.

While everyone focuses on the obvious plays — GPUs, data centres, cloud providers, a quiet revolution is happening in memory technology itself.

You see, memory in computers and AI chips can be divided into two main parts:

• Long-term memory that holds the bulk of the data in hard drives (Seagate & Western Digital’s realm)
• Short-term memory that’s designed for speedy calculations (Micron’s niche)

Both are critical for AI; however, recently, short-term memory has been seen as the vital building block for advancing robotics and advanced AI applications.

Enter Resistive Random Access Memory, or ReRAM.

This is a new type of storage technology.

ReRAM combines the best of both worlds: it remembers data even when the power is off (like a hard drive) but works almost as fast as regular computer memory (RAM).

It’s speedy, more energy-efficient, and critically for AI applications, it can perform computations within the memory itself.

That last point matters enormously.

AI inference involves massive matrix multiplication operations that typically require constant data movement between memory and processors.

ReRAM’s ability to perform ‘in-memory computing’ eliminates this bottleneck, delivering superior inference speed and energy efficiency.

The Australian dark horse

Which brings us to a small ASX company called Weebit Nano [ASX:WBT].

(Note: this isn’t a recommendation, this is simply an example of a memory company that could benefit from the AI push)

With a market cap of ~$790M, this ASX-listed company has quietly built an advanced ReRAM technology.

It isn’t the only ReRAM company out there, but it has the advantage of having some well-placed big brothers.

Weebit is progressively licensing its technology to tier-one semiconductor suppliers like OnSemi.

The company’s technology aims to address a critical AI infrastructure challenge.

Weebit’s ReRAM technology is specifically designed for the edge AI applications that will define the next wave of tech.

From autonomous vehicles to IoT devices to real-time AI inference, these applications demand memory that can adapt and learn at the edge without constant cloud connectivity.

Point is…

Sometimes the best investments aren’t the ones making the biggest headlines.

They’re the ones quietly building the technology that makes the headlines possible.