Last week, Micron (Nasdaq: MU) reported one of the most remarkable quarters of the AI era.
Revenue soared 346% from a year ago. Gross margins reached an astonishing 85%. And management’s forecast for next quarter blew past Wall Street’s already lofty expectations.
The stock has now climbed around 280% since the start of the year.
Image: Google
So what’s going on?
The simple answer is that artificial intelligence has become incredibly hungry for memory.
And that appetite is only getting bigger.
Why AI Needs More Memory
Over the years, I’ve found that some of the market’s biggest winners emerge when a critical technology becomes a bottleneck.
We’ve seen it happen recently with AI chips, power infrastructure and data centers.
Now another bottleneck is becoming impossible to ignore.
Memory.
Because every major leap in artificial intelligence requires dramatically more of it.
Larger models need to store more information. Longer context windows require more data. AI agents have to remember previous interactions. And reasoning models perform far more calculations before producing an answer.
In other words, today’s AI systems aren’t simply becoming smarter. They’re becoming far more memory-intensive.
And that’s a major change for the memory industry.
For decades, memory chips were treated like a boom-and-bust commodity.
PC demand or smartphone demand would rise, so manufacturers would build more capacity. Then inventories would pile up and prices would fall.
That cycle repeated over and over again.
But AI is changing the equation.
You see, a modern AI data center doesn’t just need processors. It also needs huge amounts of high-bandwidth memory, or HBM, to keep those processors fed with data.
Without enough memory, even the most powerful AI chip can’t run at full speed.
Think of it like building the world’s fastest race car, then starving it of fuel.
That’s why memory has become so valuable.
Nvidia’s latest GB300 NVL72 rack includes 37 terabytes of fast memory, including 20 terabytes of GPU memory.
That’s not a normal server. That’s like an AI factory in a box. And companies are racing to build thousands of them.
That demand is reshaping the entire memory market.
Some estimates project the high-bandwidth memory market will grow from roughly $4 billion this year to more than $12 billion by 2031.
Other forecasts are even more aggressive, projecting HBM demand could reach more than $30 billion by 2030.
Either way, it’s clear where this is heading.
AI is turning memory from a background component into one of the most important parts of the entire AI buildout.
And Micron’s latest earnings report gives us a real-time look at what that means.
The company’s business has become a window into one of the fastest-growing bottlenecks in the AI ecosystem.
Management once again raised its outlook as demand for AI memory continued to exceed expectations.
And Micron has signed long-term strategic agreements with 16 customers worth roughly $22 billion in deposits and commitments.
That tells us they’re trying to lock up memory supply years in advance.
Micron also expects tight market conditions to persist beyond 2027.
That’s one reason I recommended Micron to Strategic Fortunes subscribers back in February 2024, as this memory cycle was just beginning.
At the time, most investors were still focused almost entirely on the companies making AI processors.
But I believed the bigger opportunity was hiding in the bottlenecks around those processors.
Because every new GPU cluster needed more power, more cooling, more networking equipment and more memory.
Since then, Micron has gained over 1,200% in our model portfolio.
That doesn’t happen because of a single good quarter. It happens when a company sits directly in front of a demand wave that Wall Street underestimated.
And I don’t think that wave is over.
Because the next wave of AI won’t live exclusively inside warehouses full of servers.
It’ll move into robots.
And that’s where the memory story gets really interesting.
Because a robot has to be able to see the world around it. It has to listen, to balance and to understand where its limbs are. It has to recognize objects, avoid people, follow instructions and make decisions in real time.
That requires cameras, sensors, storage and constant on-board processing.
In other words, a robot needs memory everywhere.
Micron estimates that a humanoid robot could require roughly 10X as much memory as today’s average electric vehicle.
That sounds crazy once you realize that today’s advanced vehicles already use far more memory than older cars. They need it for cameras, radar, driver assistance systems, infotainment, mapping and safety features.
But imagine putting an AI system into a human-shaped machine that has to operate inside factories, warehouses, hospitals, stores and eventually homes.
It’ll need an even greater amount of memory to make sense of the physical world.
And if humanoid robots scale the way I expect over the next decade…
Then they could create an entirely new source of memory demand that barely exists today.
Here’s My Take
Micron’s latest quarter is a reminder that the biggest AI winners won’t always be the companies making the most headlines.
Sometimes they’re the companies supplying the parts the entire boom can’t move forward without.
That’s why I’m always anticipating the next bottleneck for my readers.
Because I know that the biggest fortunes are often made by solving tomorrow’s problems before the rest of the market even sees them coming.
Regards,
Ian KingChief Strategist, Banyan Hill Publishing
Editor’s Note: We’d love to hear from you!
If you want to share your thoughts or suggestions about the Daily Disruptor, or if there are any specific topics you’d like us to cover, just send an email to [email protected].
Don’t worry, we won’t reveal your full name in the event we publish a response. So feel free to comment away!
