On Thursday, the memory chipmaker reported an eye-popping 346% surge in quarterly revenue. As if that weren’t enough, it reported a profit of $28.2 billion for the quarter, almost 15 times as high as the same quarter last year. By all accounts, the company blew past analyst expectations, sending its stock soaring almost 16% in after-hours trading. The rally continued into Thursday evening, with Micron’s shares closing up 15% at $1,213 per share.
“This shows the memory and chip trade is well-intact and still in the early stages of playing out with the AI Revolution still in the third inning,” wrote tech bull and Wedbush analyst Dan Ives in a Wednesday note.
And that’s evident in its business model. While Micron’s business is anything but flashy, it makes something essential: the memory chips that store and move data at blistering speeds inside AI servers.
High-bandwidth memory chips like those made by Micron help AI processors function by stacking multiple layers of memory directly on top of each other and connecting them with microscopic wiring that allows data to flow at extremely high speeds.
Because training a large language model requires moving billions of data points per second between processors and memory, the chips Micron makes need to be fast enough to keep up.
The company, which today has more than 50,000 employees, was founded in 1978 in the basement of a Boise dental office. Twin brothers Ward and Joe Parkinson were two of the four founders. Joe was a lawyer in Boise, while Ward, after earning a master’s degree in computer design from Stanford, worked at the Dallas-based Mostek Corp., which at the time was one of the dominant players in dynamic random-access memory, or DRAM.
The fledgling company’s first contract was for Mostek, the very company three of the four cofounders had left. Mostek wanted Micron to help it design a 64K memory chip, which at the time was cutting-edge technology—but today would hold only enough memory for a single short email.
Micron took on the challenge and improved on the existing design of the 64K DRAM chip to make it the smallest in the world, setting Micron on a course to mass-produce its own improved 64K chips starting in 1981—arguably its first step to success.
It was this manufacturing facility, which still stands today, that helped Micron mass-produce its 64K memory chip product that was then used in many early computers including the Commodore 64 home computer.
As the company grew, it began producing improved chips with more memory, from 64K to 256K to its 1-megabit memory chip, which had about 16 times the memory of the company’s original 64K chip. By 1998, the company had purchased Texas Instruments’ worldwide memory operations, making it one of the biggest memory producers in the world.
But partly because of competition, for years memory chips were a commodity business with prices that swung wildly. Some years brought large profits, while other years nearly brought the company to an end as the market produced more memory chips than clients needed.
Micron went public on the Nasdaq in 1984 at $13 per share, far below today’s price of more than $1,200 per share. For decades, Micron’s boom-and-bust business made Wall Street view the company as a short-term trade rather than a long-term investment.
The rise of artificial intelligence has now taken the company’s memory business to a new level. Micron’s newfound dominance is partly the result of a new category of memory chip that the company has capitalized on. High-bandwidth memory, or HBM, is a specialized product that stacks memory chips on top of each other to move data at extraordinary speeds—exactly what AI processors need to function.
Yet only a handful of companies have the capacity to mass-produce these top-end products. Unlike traditional DRAM chips, HBM chips are extraordinarily complex to manufacture, and there is a shortage of supply worldwide.
Analysts at investment bank William Blair said it’s this dynamic that assures Micron will be an essential company in the AI race for years to come.
“Micron remains one of the key beneficiaries of the AI supercycle, which is consuming a growing share of the global memory capacity and driving record-high prices and profits for vendors,” the analysts wrote in a Wednesday note.
