The Navy's $3 million flight simulator made a pretty good blackjack player...
In 1949, a team at the Massachusetts Institute of Technology ("MIT") was tasked with a first-of-its-kind challenge.
They had to develop a computer that could respond in real time.
A typical early computer could receive a problem, work through the calculations, and return an answer... later. But "later" didn't cut it for the U.S. Navy's training programs.
When a pilot moved the controls, the flight simulator needed to react immediately. It had to calculate the aircraft's new speed and direction, then display the result before the pilot moved again.
The MIT team got to work on their naval assignment – and came up with the "Whirlwind"...
Far beyond an ordinary research project, the Whirlwind was built for fast, real-time computing...
It occupied roughly 2,500 square feet inside MIT's Barta Building. Nearly 100 racks of equipment filled the room.
The Whirlwind's speed was impressive. But its storage was perhaps even more crucial.
Each time the simulated aircraft moved, the Whirlwind had to preserve new data. The next calculation depended on the last one. A delay could break the entire sequence.
MIT designers initially used electrostatic storage tubes for the Whirlwind's memory. These tubes held the data the computer needed to keep operating. But they couldn't store information long enough to offer continuity.
So the team turned to magnets. Specifically, they pioneered "magnetic-core memory." Instead of relying on electrostatic storage tubes, the Whirlwind used tiny magnetic components to hold information.
The new design made the Whirlwind much more reliable. It was easier to add more memory with more magnets... which helped the computer respond faster.
By 1952, the Whirlwind was running so well that most of its daytime hours were assigned to Navy and Air Force projects. As Cold War tensions boiled, its work shifted away from aircraft simulation and toward air defense.
And at night, it had a little fun...
When they weren't using the Whirlwind for military projects, the programmers got creative. They used the computer's lightning-fast calculations to create a digital blackjack game.
The Whirlwind displayed the card numbers. Players aimed a "light gun" at the glowing spots on the screen, choosing whether to hit, stand, or reset.
Even this silly computer game showcased the value of memory – the programmers left some of the controls unlabeled to preserve precious processing resources. Players simply had to learn what each glowing spot did.
The Whirlwind's magnetic-core setup dominated the computer industry for the next two decades.
But these days, a familiar problem is rearing its head once again...
There's not enough memory to go around.
The AI revolution relies on processors that can complete an extraordinary number of calculations in seconds. Those processors still need memory, though. It's what supplies the information behind every calculation.
And when the memory can't keep pace, the processor waits.
Memory is one of the biggest bottlenecks of the AI build-out. The three major manufacturers – Samsung Electronics, SK Hynix, and Micron Technology (MU) – are racing to expand production.
They're spending more than $1 trillion combined on new facilities and manufacturing lines over the next decade.
That's creating a bottleneck behind the bottleneck...
Memory companies can only increase supply when they have enough equipment. After all, someone has to actually make the chips.
Major chip-equipment makers are already seeing the results of this demand. In 2014, Lam Research (LRCX) generated $678 million in operating profit. That number ballooned to $8.2 billion in fiscal 2026, signaling more than 12 times growth.
It's a similar story for fellow equipment maker Applied Materials (AMAT). Operating profits were $432 million in 2013... and an incredible $8.3 billion in 2025.
Share prices have also followed suit. Lam Research is up more than 8,500% since the start of 2013. Applied Materials has soared more than 4,000%.
Last week, I revealed another set of stocks that's poised to benefit from the current AI bottleneck. The most powerful signal in our firm's history is firing on these companies. And it's the same signal that preceded the rise of Nvidia (NVDA), the rest of the Magnificent Seven, and 448 of today's S&P 500 Index stocks. Click here to learn their names and ticker symbols.
All told, there's a massive opportunity in the memory market today... and it won't last long.
Regards,
Joel Litman
September 2, 2026
