The AI Memory Shortage Could Make These 4 Stocks Winners
Listen to the audio version of this article (generated by AI).
Editor’s Note: The headlines coming out of SpaceX’s earnings report focused on billions of dollars in AI spending. My colleague Eric Fry, a top global macro analyst for decades, paid much closer attention to the one thing Elon Musk said he couldn’t simply spend his way around. As Eric explains below, that single constraint could become one of the most profitable investment themes of AI’s next chapter.
If you’d like to see the complete picture, Eric also expands on these ideas in his free Market Shock presentation, where he shares several additional stocks he believes deserve investors’ attention.
The universe wasn’t supposed to do this.
In 1998, astronomers made a discovery so surprising that it eventually earned three of them the Nobel Prize in Physics.
The expansion of the universe was accelerating.
That flew in the face of decades of scientific thinking.
Ever since Edwin Hubble discovered in 1929 that the universe was expanding, astronomers had assumed gravity would gradually slow that expansion over time. The only real question was how much it had slowed.
To find the answer, scientists turned to extraordinarily distant exploding stars known as supernovae. Because these stellar explosions have a predictable brightness, astronomers can use them as mile markers in space, comparing how bright they should appear with how bright they actually look from Earth.
What they discovered turned conventional wisdom on its head.
Those supernovae were fainter – and, therefore, farther away – than expected. Instead of slowing under gravity’s pull, the universe was expanding at an accelerating rate.
Something scientists couldn’t see or explain was pushing space outward.
Observations from the Hubble Space Telescope – named after Edwin Hubble himself – helped confirm the finding and deepen the mystery. Even today, the nearby universe appears to be expanding roughly 5% to 9% faster than our best models predict.
Now another space-related expansion appears to be accelerating, one with much more immediate consequences for investors.
The global space economy recently reached a record $613 billion. Johns Hopkins researchers expect it to approach $1.8 trillion within the next decade, fueled by reusable rockets, private investment, falling launch costs, and entirely new businesses that would have sounded like science fiction only a few years ago.
And now artificial intelligence is accelerating that expansion even further.
AI is already helping companies design spacecraft, process vast quantities of satellite data, automate missions, and explore the possibility of operating data centers in orbit. Space is no longer merely somewhere technology travels. It may become part of the infrastructure where tomorrow’s most advanced computing takes place.
That brings us to Space Exploration Technologies Corp. (SPCX).
SpaceX Earnings Reveal a New AI Bottleneck
This week, investors received their first detailed look inside the newly public company. The results showed a business evolving far beyond rocket launches and satellite communications and spending staggering sums to become a major force in AI infrastructure.
SpaceX generated $7.8 billion in second-quarter revenue, up roughly 92% from a year earlier. But it also spent nearly $16 billion expanding its AI infrastructure, helping produce a quarterly net loss of $541 million.
Yet when Elon Musk discussed what could limit that expansion, he did not point to SpaceX’s losses, its access to capital, or even the availability of advanced AI chips.
“The limiting factor currently is memory,” Musk told investors.
For investors, that admission may prove far more valuable than anything else in SpaceX’s earnings report.
Because the memory shortage constraining Musk’s AI ambitions is also creating severe supply-and-demand imbalances throughout the technology industry. And the relatively small group of companies capable of supplying that memory could possess exactly what investors should look for during a shortage: surging demand, limited competition, rising prices, and extraordinary pricing power.
Today, I’ll show you why Elon Musk believes memory (not money) is becoming AI’s biggest constraint and how that shortage could reshape the industry.
Plus, I’ll introduce you to one memory company I believe is positioned to benefit… and show you where you can find three more memory-related stocks I’m watching before Wall Street fully catches on.
Why AI Memory Is SpaceX’s Biggest Infrastructure Bottleneck
When Elon Musk called memory “the limiting factor” for SpaceX’s AI ambitions, he wasn’t talking about some obscure engineering problem.
He was describing a challenge that now confronts virtually every company trying to build the next generation of artificial intelligence.
SpaceX wants to become much more than a launch company. Musk envisions it operating enormous AI data centers, processing data gathered by Starlink’s thousands of satellites, developing autonomous spacecraft, and ultimately creating an AI infrastructure business that extends well beyond Earth’s atmosphere.
But none of that can happen without memory.
Modern AI systems rely on three essential building blocks:
- GPUs, like Nvidia’s AI accelerators, which perform the calculations.
- HBM (high-bandwidth memory), the ultrafast memory attached directly to those GPUs.
- DRAM (dynamic random access memory), the working memory that allows AI models to “think” in real time.
The first bottleneck of the AI boom was compute. Nvidia Corp. (NVDA) became one of the world’s most valuable companies by solving that problem.
That bottleneck has shifted. Now it’s memory.
Large language models don’t simply perform calculations. They must constantly store, retrieve, and manipulate staggering amounts of information while generating each response.
Training a ChatGPT-sized model can require tens or even hundreds of terabytes of DRAM spread across thousands of GPUs. Without enough memory, those expensive AI systems simply wait.
No memory means no intelligence.
That’s why Nvidia CEO Jensen Huang recently warned that the industry’s “memory bottleneck is severe.”
It’s also why tech companies have reportedly stationed employees in South Korea for months at a time, hoping to secure scarce DRAM allocations from Samsung and SK Hynix Inc. (SKHY). The industry has even given these buyers a nickname: “DRAM beggars.”
SpaceX may have been the first company to say it publicly, but it certainly won’t be the last.
PDF Solutions: A Picks-and-Shovels Play on the AI Memory Shortage
The numbers explain why.
Nearly 100 gigawatts of new AI data centers are expected to come online over the next four years. Yet industry estimates suggest there is enough DRAM supply to support only about 15 gigawatts of new capacity over the next two years.
That imbalance is already driving prices sharply higher. Market researcher TrendForce expects conventional DRAM contract prices to surge 90% to 95% in early 2026, one of the fastest increases the industry has ever experienced.
During SpaceX’s earnings call, Musk added another eye-opening statistic. He expects AI memory demand to grow at nearly 200% annually.
That combination – exploding demand and constrained supply – is exactly the sort of bottleneck I like to look for as an investor.
One company I’ve got my eye on is PDF Solutions Inc. (PDFS).
Unlike memory manufacturers themselves, PDF Solutions helps semiconductor companies produce more usable chips from every manufacturing run. Its software identifies defects, improves manufacturing yields, and helps chipmakers reduce costly failures. Those are capabilities that become dramatically more valuable when every additional AI memory chip commands a premium.
As manufacturers race to increase DRAM and HBM production, companies like PDF Solutions quietly become indispensable behind the scenes.
That’s one reason PDFS has become one of the most interesting memory-related stocks I’m watching.
It isn’t the only one.
Where the Next AI Memory Winners May Emerge
If you’ve followed my work for any length of time, you know I spend very little time chasing whatever Wall Street already loves.
Instead, I look for the bottlenecks.
Years ago, that meant identifying Nvidia before most investors appreciated how valuable AI compute would become.
Last year, it meant recognizing that the AI trade was entering a new phase, one where the biggest gains would increasingly come from the companies supplying what I call AI’s Golden Rivets. Those are the irreplaceable materials, energy, networking, and memory infrastructure every AI company depends upon.
Today, I believe we’re entering the next phase of that acceleration.
Much like the astronomers who assumed the universe’s expansion would gradually slow, I believe many investors are underestimating what’s happening today. They see AI continuing to grow, but they haven’t yet recognized how quickly demand for the infrastructure supporting that growth is accelerating.
And as we’ve already seen with SpaceX, those accelerating demands are beginning to expose entirely new bottlenecks… and entirely new investment opportunities.
In my Market Shock presentation, I explain that while the AI Revolution is still in its early innings, the next big winners are unlikely to be the same companies that dominated the last three years.
I’ll also share four memory-related stocks I believe are positioned to benefit from this bottleneck, including three additional names and tickers beyond PDF Solutions that I’m watching very closely.
If you’d like to see the complete framework, and why I believe SpaceX’s earnings call may have revealed far more than Wall Street realizes, you can watch that free presentation here.
