Chat with this content: Summary, Analysis, News...
Share on Reddit
Micron Technology: The Memory Wars and the AI Revolution
I. Introduction & Setting the Stage
December 2025. Micron Technology steps up and posts a jaw-dropping quarter: $13.64 billion in revenue, the highest in its history. Gross margins push up toward 57%, and the stock is up more than 175% over the past year. For a company that started in the basement of an Idaho dental office nearly fifty years ago, it doesn’t just feel like a good run. It feels like a comeback.
This was Micron’s fiscal Q1 2026, ending November 27, 2025: $13.64 billion in revenue, up from $11.32 billion the quarter before and $8.71 billion a year earlier. And it wasn’t one division getting lucky. Micron delivered record revenue and meaningful margin expansion across the company and in each business unit. In other words: this wasn’t noise. It was a signal.
Because Micron occupies a strange, consequential position in the modern economy. It’s America’s last remaining manufacturer of memory chips—the only U.S.-based memory maker still standing in a category that underpins basically everything: smartphones, PCs, cars, data centers, and now the AI buildout that’s rewriting the rules of computing. Memory doesn’t get the headlines like GPUs do, but AI systems don’t run on hype. They run on bandwidth, capacity, and power efficiency. And that makes memory a strategic choke point.
So here’s the question that drives this story: how did a startup from a Boise dental office basement become America’s last memory champion?
And the tougher follow-up: now that AI has made memory suddenly precious again, can Micron finally break free from the commodity curse that has defined this industry for decades?
By fiscal year 2025, Micron had put up record revenue of $37.4 billion, up 49% year over year. Gross margins expanded to 41%, up seventeen percentage points from fiscal 2024. And Micron projected that the market for HBM—High Bandwidth Memory, the specialized, ultra-fast memory that sits next to AI accelerators—would surge from roughly $35 billion in 2025 to about $100 billion by 2028.
The numbers are striking. But they’re not the story—they’re the payoff. To understand how Micron got here, you have to go back to the beginning, to Boise, to a basement, and to a handful of engineers who decided to pick a fight with the most unforgiving business in semiconductors.
II. The Basement Beginning: Four Engineers vs. The World (1978-1985)
In 1978, the semiconductor industry belonged to California. Intel, National Semiconductor, Fairchild—companies with talent, capital, and gravity. The idea that a world-class memory player could come out of Boise, Idaho—then a city of about 100,000 people, better known for potatoes than integrated circuits—sounded like a punchline.
And yet, that’s exactly where four engineers decided to try.
Micron was founded in Boise in 1978 by Ward Parkinson, Joe Parkinson, Dennis Wilson, and Doug Pitman. At first, it wasn’t a chip manufacturer at all—it was a semiconductor design consulting firm. Four people. One basement. Specifically, the basement of a Boise dental office.
Ward and Joe Parkinson, twins, were especially unlikely candidates to take on Silicon Valley and Japan. They’d grown up in southeastern Idaho doing farm work and delivering potatoes to Simplot warehouses. The story only gets stranger from there: the twins later married—then divorced—sisters. But they were sharp, and school became their exit ramp. They went all the way to Columbia University in New York.
Eventually, life pulled them back toward Idaho. Ward landed in engineering—doing contract work tied to Texas chipmaker Mostek. Joe, trained as a lawyer, found himself doing what he later described as dreary small-city legal work. Ward and his engineering colleagues wanted to build something of their own, and in 1978 they formed Micron. Their first real shot was a contract: design a 64K memory chip for Mostek.
It was supposed to be the runway.
Instead, it became the first crisis. The contract was canceled after an acquisition, and the tiny Boise startup suddenly had no paying customer—and a half-built future. Rather than folding, the founders pivoted hard. With Joe Parkinson stepping in alongside the engineers, they decided they’d pursue the 64K chip on their own.
That decision created the next, unavoidable problem: money.
Building chips isn’t like writing software. Fabs cost a fortune, and even the equipment to get started is brutal for a small team. Venture capital in Silicon Valley wasn’t interested. The DRAM market was already turning into a knife fight, and Japanese manufacturers looked unstoppable.
Then Micron found its patron. And he didn’t come from Sand Hill Road.
Enter J.R. Simplot: The Potato King Bets on Chips
J.R. Simplot was Boise royalty: founder of the J.R. Simplot Company, a giant in agricultural supply and potato products. He’d built one fortune supplying dehydrated potatoes to U.S. troops during World War II, and another after a 1967 handshake deal with Ray Kroc that helped make Simplot McDonald’s key french fry supplier. By 2007, he was estimated to be the 89th-richest person in the United States, worth about $3.6 billion.
The Parkinson brothers met Simplot at the Royal Café in downtown Boise. They pitched him on transistors, capacitors, and memory chips—concepts that didn’t exactly map to farming.
Simplot wasn’t a tech investor. He was, in many ways, the opposite. But he did understand commodities. And he recognized something the VCs didn’t: memory was becoming a commodity market, and that meant timing mattered. When everyone else was running away because prices were collapsing and competition was brutal, Simplot saw what he’d seen in agriculture again and again—sometimes the best time to buy into a commodity business is when the market is depressed and the smart money is liquidating.
In 1980, at age 71, Simplot made his bet. He put in $1 million for 40% of what would become Micron Technology. Over time, he poured in another $20 million, helping fund Micron’s first fabrication plant and giving the company oxygen to survive its earliest years.
Simplot wasn’t alone. Other local Idaho businessmen chipped in too: Tom Nicholson, Allen Noble, Rudolph Nelson, and Ron Yanke. Ward Parkinson had built a relationship with Noble in particular—a farmer and irrigation-system inventor—another reminder that Micron’s early backers didn’t look like the usual cast of semiconductor history.
That capital unlocked the leap from idea to reality. By 1980, Micron broke ground on its first fabrication plant. In 1981, it stopped being “just” a design consultancy and became a manufacturer, with its first wafer fab—Fab 1—focused on producing 64K DRAM chips.
And Micron didn’t just ship chips. It shipped impressive chips.
By 1982, Micron had shipped more than a million. That number alone didn’t stand out next to larger competitors, but the quality and design did. Micron’s chips used bigger, easier-to-read memory cells, which improved reliability. And they were remarkably small—about 40% smaller than Motorola’s chip and 15% smaller than Hitachi’s.
In early 1983, Micron squeezed chip size down to 22,000 square mils, a notable step that put it ahead of Japanese competitors and even Texas Instruments. The progress translated into momentum, and in 1984 Micron went public, raising capital to fund expansion and further research.
But even with new money and better chips, this industry still had only one rule: it can turn on you overnight.
By 1985, the floor dropped out. National Semiconductor suspended plans to market a 256K chip. Intel announced it was shutting down all RAM production that fall. United Technologies closed Mostek’s operations the same year. Meanwhile, chip prices collapsed—64K parts fell from roughly $4 to about 25 cents, and 256K chips slid from around $20 to $2.50. Micron had to lay off about half its workforce in the spring of 1985.
This was the real test: not whether Micron could invent, but whether it could endure.
And here’s where Boise—supposedly the company’s greatest weakness—quietly became an advantage. Cheap hydroelectric power. Low land costs. And a workforce that didn’t treat tech like a gold rush. Micron’s cost structure, and its culture, were built to live through cycles that wiped out better-funded peers.
The lesson from Micron’s earliest years is simple and brutal: in commodity markets, you don’t win with branding or charm. You win with engineering, efficiency, and the discipline to survive when the price chart looks like a waterfall.
Micron learned that in its first decade—because it had to.
III. The Memory Wars: Survival in the Commodity Trenches (1985-2000)
The late 1980s through the 1990s were the crucible that made modern Micron. If the early years proved the company could build chips, this era proved it could survive the kind of market that chews up good companies and spits out scrap.
Because DRAM is brutally interchangeable. A memory chip is a memory chip—whether it comes from Micron, Samsung, or Toshiba. When buyers can swap suppliers without changing anything else, the whole business collapses into one question: who can make it cheapest? And in the early ’80s, the answer looked like Japan.
Japanese electronics giants like Hitachi, NEC, and Fujitsu grabbed an early lead in 64K DRAM, the memory that powered computers, video games, and telecom gear. By 1981, Japanese companies controlled about 70% of the global 64K DRAM market, and U.S. makers weren’t mounting much of a defense.
Micron survived anyway, mostly by turning manufacturing into a weapon. Boise helped more than outsiders expected: a stable workforce that wasn’t constantly rotating jobs, and a culture that kept pushing every spare dollar back into process improvements. In memory, “better” doesn’t always mean flashier. It usually means tighter yields, fewer defects, and lower cost per bit—over and over, cycle after cycle.
The 1990s PC boom gave Micron a powerful tailwind, even if the ride was never smooth. In 1994, Micron made the Fortune 500. It was no longer the plucky Idaho upstart; it was becoming a real industry player—growing through technology advances, partnerships, and the first steps toward the dealmaking playbook it would lean on later.
That same year, the company also handed the keys to a new leader. Founder Joe Parkinson retired as CEO, and Steve Appleton became Chairman, President, and CEO.
Steve Appleton: The Daredevil Executive
Steve Appleton wasn’t the typical semiconductor CEO. Steven R. Appleton (March 31, 1960 – February 3, 2012) was born and raised in Southern California, came to Idaho for school, and played tennis at Boise State.
He joined Micron in 1983, straight out of college, working the night shift in production for less than five dollars an hour. He moved fast: wafer fab manager, production manager, director of manufacturing, vice president of manufacturing. In 1991 he became president and COO. In 1994, at just 34, he was CEO and chairman.
Appleton fit Micron because he was Micron—scrappy, intense, and comfortable living on the edge. He flew stunt planes and raced off-road vehicles in the Baja 1000. He brought that same adrenaline to memory, a business that rewards bold bets but punishes mistakes immediately. In January 1996, he was fired—and then rehired eight days later. It was corporate whiplash, and it captured both the volatility of the industry and the man running through it.
"This strategic acquisition will enhance Micron's position as the most cost-effective memory producer in the world," said Steve Appleton, chairman, CEO and president of Micron. "The additional global capabilities, including participation in a unique joint-venture manufacturing strategy, positions Micron to take advantage of future markets."
Micron also tried to diversify—sometimes out of ambition, sometimes out of self-defense. In 1991 it went after RISC processors with a chip called FRISC, aimed at embedded control and signal processing. The pitch was strong on paper: 80 MHz, described as “a 64-bit processor with fast context-switching time and high floating-point performance,” built for quick interrupt handling. A subsidiary was set up, and designs even started landing in graphics cards and accelerators. But by 1992, Micron decided it wouldn’t deliver the “best bang for the buck.” The company pulled the plug, reassigned engineers, and moved on.
A bigger swing came later with PCs. Micron got into the personal computer business, only to find the economics were as unforgiving as memory—just in a different way. By the time the company spun off the unit as MPC Corporation in 2002 and put it up for sale, it was the number 12 American computer maker with just 1.3% of the market—too small to matter, too exposed to compete on price, and far from Micron’s core strengths.
The takeaway from this era wasn’t subtle: focus matters. Micron’s edge was building memory better and cheaper than the next guy—not trying to win the consumer branding game.
For anyone watching long-term, the 1990s revealed Micron’s true identity. This was a company built to endure cycles through manufacturing excellence. And just as importantly, the detours taught leadership what not to do—lessons that would shape the much bigger bets Micron was about to place next.
IV. The Great Consolidation Play: Building Through Acquisitions (1998-2013)
By the late 1990s, the endgame for memory was coming into focus. This business was getting too capital-intensive, too cyclical, and too brutal for a crowded field. If you couldn’t keep funding new fabs and new process nodes—over and over—you didn’t just fall behind. You disappeared. Steve Appleton saw the math early: in memory, survival increasingly meant scale. And scale often meant buying it.
Micron’s first big swing came in 1998, with Texas Instruments’ memory business. It wasn’t a single factory or a tidy product line. It was an international footprint overnight: TI’s wholly owned fabs in Avezzano, Italy, and Richardson, Texas; joint venture interests in Japan and Singapore; plus an assembly and test operation in Singapore.
The price tag reflected how serious this was. TI received roughly 28.9 million shares of Micron stock valued at $881 million at closing, $740 million in notes convertible into another 12 million shares, and a $210 million subordinated note. Micron also received $550 million in proceeds from financing provided by TI. In other words, this wasn’t just Micron buying assets—it was a carefully engineered deal to get Micron more capacity without choking on the upfront cash burn.
And it happened at exactly the wrong time—at least if you were trying to impress anyone with short-term financials.
Micron was still fighting the same ugly pricing dynamics that had defined the decade. In one snapshot from the period: the company reported a net loss of $106 million, or 50 cents per share, for its third fiscal quarter ended May 28 on sales of $610 million. Even as unit volumes climbed, the average price per megabit sold fell around 30%. That’s the memory business in one sentence: you can sell more and still make less.
So why buy now?
Because Appleton and the team were betting on the cycle. They believed prices would rebound, and when they did, the company that owned more efficient capacity—more wafers, more output, more leverage—would be the one still standing. Buying into the teeth of a downturn wasn’t reckless. It was the point.
The 2000s kept that playbook rolling, but with a twist: Micron wasn’t only bulking up DRAM capacity anymore. It was also trying to make sure it had a seat at the table in flash memory, the other half of the modern storage-and-memory world.
In 2005, Micron and Intel formed a joint venture: IM Flash Technologies, based in Lehi, Utah. In 2011, they added another joint venture, IM Flash Singapore. For Micron, Intel brought technical partnership and shared investment in flash—an on-ramp to diversify beyond DRAM without trying to build everything from scratch.
Then came consumer-facing moves. In 2006, Micron acquired Lexar, known for digital media products. In February 2010, Micron agreed to buy flash-chip maker Numonyx for $1.27 billion in stock.
Deal by deal, Micron was building something larger than a Boise DRAM shop: a scaled, global memory platform, with flash in the mix and more manufacturing options around the world.
But the most transformative acquisition—the one that would change Micron’s position in the industry outright—was still ahead.
V. The Elpida Acquisition: Betting the Company (2012-2013)
Tragedy struck on February 3, 2012. Steve Appleton was killed while attempting an emergency landing at Boise Airport, moments after takeoff, in a Lancair IV-PT experimental-category, four-seat turboprop aircraft. Micron confirmed his death in a press release that praised his “passion and energy.” He was 51.
Soon after, Mark Durcan replaced Appleton as CEO, dropping his previous title of president.
Durcan inherited a company in mourning—and, at the same time, one of the most consequential opportunities the memory industry had seen in years: the chance to buy Elpida Memory.
Elpida wasn’t some random distressed asset. It was Japan’s DRAM champion. Founded in 1999, it developed, designed, manufactured, and sold DRAM products, and also operated as a semiconductor foundry. It began as NEC Hitachi Memory, created by merging the DRAM businesses of Hitachi and NEC, and in 2003 it absorbed Mitsubishi’s DRAM business as well.
The global financial crisis hit memory especially hard. In 2009, Elpida received 140 billion yen in financial aid and loans from the Japanese government and banks. But by early 2012, with prices collapsing and debt mounting, the support still wasn’t enough. Elpida was bankrupt, unable to service what it owed.
Micron stepped in.
Micron Technology, Inc. and Elpida’s trustees announced the closing of Micron’s acquisition of 100 percent of Elpida’s equity, under a Sponsor Agreement entered into on July 2, 2012, as part of Elpida’s corporate reorganization proceedings overseen by the Tokyo District Court.
Jones Day acted as Taiwan counsel to Micron in its acquisition of Elpida for 200 billion yen (US$2.5 billion), and in Micron’s US$335 million acquisition of an approximately 24 percent stake of Rexchip Electronics Corp. from Powerchip Technology Corporation.
The deal structure was the genius. Micron acquired Elpida for about $750 million in cash at closing, then agreed to pay roughly $1.75 billion to Elpida’s secured and unsecured creditors in annual installments through 2019. In plain English: Micron didn’t have to swallow the full cost upfront. It largely financed the acquisition using Elpida’s own future cash flows.
What did Micron get for it? Real, modern manufacturing scale—fast.
The acquisition gave Micron 100 percent ownership of Elpida, including a 300mm DRAM fab in Hiroshima, Japan, plus approximately a 65 percent ownership interest in Rexchip, which came with 100% of the capacity of Rexchip’s 300mm DRAM fab in Taiwan.
Together, Elpida and Rexchip could produce more than 185,000 300mm wafers per month. That increased Micron’s manufacturing capacity by about 45%. And with an estimated 28% market share after the deal, Micron jumped past Hynix to become the number-two DRAM player behind Samsung.
But the real strategic value wasn’t just more wafers. It was where those wafers went.
At the time, Elpida was supplying SDRAM for Apple’s iPhone 5—specifically for the A6 processor. Reuters had reported that Apple had booked about half the output of Elpida’s 12-inch Hiroshima plant. Chip analysis also showed the iPhone 5’s A6 package using 1GB of Elpida RAM.
Apple used 80% of Elpida’s mobile DRAM production capacity, which was supplying SDRAM for the iPhone 5.
That mattered because mobile DRAM—low-power memory built for smartphones—commanded higher margins than the commodity PC memory Micron had historically lived and died by. The Elpida deal handed Micron something it had never truly had before: a meaningful relationship with the world’s most valuable technology company, anchored in a premium segment of memory.
And for investors, Elpida became the purest example of Micron’s countercyclical DNA. In a business where the cycle kills the careless, Micron bought when others were forced to sell—picking up distressed, world-class capacity at a fraction of what it would cost to build from scratch.
VI. The Pivot to High-Value Memory (2014-2020)
The post-Elpida era needed a different kind of leadership. Micron had scale now. What it didn’t yet have was a clear escape route from the old trap: win share in a commodity market, then get crushed when the cycle turns.
In February 2017, Micron announced that CEO Mark Durcan would retire. The company tapped Sanjay Mehrotra as his successor. Mehrotra officially took over on May 8, 2017, and Durcan stayed on as an adviser until early August.
Mehrotra wasn’t a career Micron insider. He was the co-founder of SanDisk, and he’d led it as president and CEO from 2011 until Western Digital acquired it in 2016. Born in 1958 in Kanpur, India, he brought something Micron valued immediately: a builder’s mindset from flash memory’s formative years, plus serious technical credibility. He holds more than 70 patents, several tied to high-capacity flash memory, and in 2022 he was inducted into the National Academy of Engineering.
What Mehrotra did next was subtle in description but massive in implication: he began steering Micron away from pure commodity memory and toward higher-value products where performance, reliability, and customer intimacy mattered as much as cost.
The center of gravity shifted to the data center. Micron leaned harder into enterprise SSDs for hyperscalers, high-capacity DIMMs for servers, and, eventually, the specialized memory that would sit closest to the most demanding compute on earth.
That pivot also meant rewriting old partnerships. Micron and Intel had built the IM Flash joint venture to scale NAND, but Micron didn’t want to be defined by low-value capacity. In October 2021, Micron closed the sale of IM Flash’s Lehi, Utah fab to Texas Instruments for $900 million—an unmistakable signal that Micron was willing to shed assets that didn’t fit where the market was going.
On the technology front, 3D NAND was the inflection point that made the strategy possible. Planar NAND—shrinking cells side by side—was running into the hard physics of silicon. 3D NAND flipped the approach: stack the cells vertically. That single shift reopened the density roadmap, enabling far higher capacity than planar NAND and keeping flash’s curve alive.
None of this happened in a straight line. The memory market crashed again in 2018 and 2019, triggering layoffs and cost cuts. But through the downturn, Mehrotra kept investing in R&D—because in memory, the companies that stop innovating during the bad years don’t get to lead in the good ones.
And a new kind of “good year” was coming.
VII. The AI Revolution: HBM and the New Gold Rush (2020-Present)
The AI revolution changed everything.
When OpenAI released ChatGPT in late 2022, the rest of the world finally internalized what the deep-learning crowd already knew: big models don’t just need more compute. They need a firehose of memory bandwidth. And suddenly, Micron wasn’t just “a memory company” again. It was sitting on one of the hardest constraints in the entire AI supply chain.
HBM: The New Gold
HBM—High Bandwidth Memory—is the specialized memory that sits right next to AI chips inside data center accelerators. Compared to conventional DRAM like DDR4 or graphics memory like GDDR5, HBM uses a much wider interface. An HBM stack of four DRAM dies (4‑Hi) has two 128‑bit channels per die for a total of eight channels and a width of 1024 bits. Put four of those stacks on a GPU, and you’re looking at a 4096-bit-wide memory bus.
The trick is how HBM is built. Instead of laying memory chips side by side, you stack them like a vertical tower and connect them with microscopic pathways called TSVs—through-silicon vias. That architecture dramatically increases bandwidth, while using less power per bit moved. For modern AI accelerators, that’s everything. Compute keeps getting faster, but if data can’t get in and out fast enough, the whole system chokes.
Micron’s HBM3E became its ticket into that choke point. Its 8‑high and 12‑high HBM3E memory cubes deliver more than 1.2 TB/s per placement, with the company saying they do it at up to 30% lower power consumption than competing solutions. HBM3E has 1024 IO pins, and with pin speeds above 9.2Gbps, the bandwidth math starts to look like science fiction—until you realize it’s now shipping.
That mix of performance and power efficiency helped Micron earn “preferred supplier” status with the category’s biggest buyer: Nvidia.
And the business impact showed up fast. Micron’s HBM revenue crossed $1 billion in fiscal Q2 2025. By FY 2025, the data center business was 56% of Micron’s total revenue, and in Q4 FY 2025 HBM alone reached nearly $2 billion. Micron also said the combined revenue from HBM, high-capacity DIMMs, and LP server DRAM hit $10 billion—more than five times the prior fiscal year. The point wasn’t just growth. It was mix: the company was increasingly selling the kind of memory you win on engineering, not just price.
Nvidia and AMD: The Essential Partnerships
This is where the relationships matter. Micron’s 24GB 8‑high HBM3E shipped with NVIDIA H200 Tensor Core GPUs, and its production-capable 36GB 12‑high HBM3E was also available. Micron also announced integration of its 36GB 12‑high HBM3E into AMD’s upcoming Instinct MI350X Series GPUs and platform.
Over time, Micron became a deeper part of Nvidia’s supply chain—and a meaningful supplier to AMD as well. In Q4, Micron said it expanded its HBM customer base from four customers in Q3 to six. And in another sign of how strategically placed it was, Micron was the sole LPDRAM supplier in NVIDIA’s GB family, strengthening its position inside AI servers beyond HBM.
The scary part—if you’re an AI builder—is how quickly HBM requirements are exploding. Nvidia’s B300 reportedly carries 288GB of HBM3e, up 60% versus the B200 and more than 3.5 times the H100. In an eight-server configuration, that’s about 2.3TB of HBM. At the rack scale, the numbers get almost absurd: Nvidia’s GB200 NVL72 supports up to 13.4TB of HBM, while the GB300 supports up to 21.7TB.
In other words: the “memory as a side component” era is over. HBM is becoming a defining input to how far, and how fast, AI can scale.
The Competition: Samsung's Stumbles, SK Hynix's Lead
HBM is an elite club with only three real players: SK Hynix, Micron, and Samsung. SK Hynix led early. Samsung, despite its massive manufacturing footprint, spent much of 2025 struggling through qualification for its HBM3E parts on Nvidia platforms. That opened a lane—and Micron drove through it, leapfrogging Samsung to claim the number-two position in the global HBM market.
Samsung was expected to push for a comeback with HBM4 development in 2026. But in this moment, Micron was ahead where it counted: qualification, efficiency, and actually shipping into the hottest demand wave semiconductors have seen in years.
Record Results: December 2025
All of this snapped into focus with Micron’s fiscal Q1 2026 results, released just two weeks ago. The company reported record revenue of $13.64 billion, up 57% year over year. Non-GAAP gross margin jumped to 56.8%, up 11 percentage points from the prior quarter. That margin expansion wasn’t financial engineering—it was product mix: more HBM3E, more high-capacity server DRAM, and more of the parts customers can’t easily substitute.
Micron also said it hit records across the board: total company revenue, DRAM and NAND revenue, HBM and data center revenue, and revenue in each business unit. And it locked in something that matters enormously in a supply-constrained market: agreements on price and volume for its entire calendar 2026 HBM supply, including Micron’s industry-leading HBM4.
Then came the guide. For fiscal Q2 2026, Micron expected non-GAAP revenue of $18.7 billion (plus or minus $400 million) and gross margin around 68% (plus or minus 100 basis points).
If you’re an investor, that’s the headline: Micron was no longer living and dying as a pure commodity producer with whiplash margins. It was becoming a differentiated supplier to AI infrastructure—one selling the exact kind of high-performance, high-value memory that the market can’t build modern accelerators without.
VIII. Business Model & Market Dynamics
To really understand Micron, you have to understand the weird, punishing economics of memory. This is an industry where the product can look like a commodity, the manufacturing looks like heavy industry, and the winners are often decided by who can survive the longest through the down cycles—then show up with the right technology when the up cycle arrives.
DRAM vs. NAND: Different Products, Different Dynamics
Memory comes in two main flavors, and they behave very differently.
DRAM—dynamic random access memory—is the “working” memory inside devices. It’s fast, relatively expensive per bit, and it forgets everything the moment power is cut. Without DRAM, nothing computes. PCs, phones, servers, cars—every one of them needs it. Micron is the third-largest DRAM producer in the world, behind Samsung and SK Hynix.
NAND is storage. It’s non-volatile, meaning it keeps your data even when the device is off. It’s slower than DRAM, but far cheaper per bit, which is why it’s used in SSDs, USB drives, and the storage inside phones and laptops. Historically, NAND has also been a more crowded battlefield—with more competitors over time—so it tends to be even more competitive.
Micron plays in both. DRAM has typically been the majority of the business, with NAND making up the balance.
The Cyclical Nature of Memory
Then there’s the part that has haunted memory companies for decades: the cycle.
When demand is hot—PC upgrades, new phone launches, data center buildouts—memory prices can spike fast, because you can’t conjure fab capacity overnight. But when demand cools, prices collapse just as quickly, because fabs are fixed-cost machines that keep running whether the market wants the output or not.
That boom-bust rhythm has wrecked countless companies and frustrated investors for generations.
AI, though, introduces something different—more structural stickiness. HBM isn’t just “more DRAM.” It’s specialized manufacturing, and it isn’t trivial to repurpose. Plus, customers don’t just buy it; they qualify it. Getting into platforms like Nvidia can take 12 to 18 months, which creates a kind of lock-in that plain commodity DRAM doesn’t have. That doesn’t eliminate cycles, but it can change their shape.
Reorganization for AI
Micron has been reorganizing itself around that reality.
The company recently reshaped its business segments to better match where demand is going—especially in the data center. In the most recent quarter, Micron’s cloud memory business unit grew 213% year over year. The organization now maps more directly to major customer categories: data center customers like hyperscalers and AI companies, broader enterprise and consumer markets like PCs and mobile, and embedded and automotive.
It’s a sign that Micron isn’t just selling parts anymore—it’s aligning the whole company around a handful of end markets that matter most.
The Crucial Exit: Prioritizing AI Over Consumers
The clearest signal of that shift came with a decision that would’ve been unthinkable in earlier eras: Micron announced it would exit the Crucial consumer business, including the sale of Crucial consumer-branded products through key retailers, e-tailers, and distributors worldwide.
"Micron has made the difficult decision to exit the Crucial consumer business in order to improve supply and support for our larger, strategic customers in faster-growing segments," said Sumit Sadana, EVP and Chief Business Officer at Micron Technology. "Thanks to a passionate community of consumers, the Crucial brand has become synonymous with technical leadership, quality and reliability of leading-edge memory and storage products."
Micron framed it plainly: “The AI-driven growth in the data center has led to a surge in demand for memory and storage,” and it wanted its supply pointed at the biggest, fastest-growing customers.
There’s a certain symmetry here. This is the same company that once tried to diversify into consumer PCs—only to learn how unforgiving consumer economics can be. Now it’s doing the opposite: walking away from consumers altogether. The message is hard to miss. Micron is staking its future on AI and the data center.
CHIPS Act: Government Support
As Micron leans into AI, it’s also leaning into a broader national priority: bringing advanced semiconductor manufacturing back to the U.S.
Micron Technology and the Trump Administration announced Micron's plans to expand its U.S. investments to approximately $150 billion in domestic memory manufacturing and $50 billion in R&D, creating an estimated 90,000 direct and indirect jobs. As part of today's announcement, Micron plans to invest an additional $30 billion beyond prior plans which includes building a second leading-edge memory fab in Boise, Idaho.
Micron anticipates that all of its U.S. investments will be eligible for the Advanced Manufacturing Investment Credit (AMIC). This includes up to $6.4 billion in CHIPS Act direct funding to support the construction of two Idaho fabs and two New York fabs, as well as the expansion and modernization of its Virginia fab.
Micron is the only U.S.-based manufacturer of advanced memory chips, and its DRAM technology powers everything from artificial intelligence and high-performance computing to automotive and next-generation wireless devices. Currently, 100% of leading-edge DRAM production occurs overseas, primarily in East Asia.
Step back, and the business model looks like it’s evolving in real time. Micron still competes in high-volume commodity memory, where cost and scale matter. But increasingly, it’s also becoming a differentiated supplier of high-value memory for AI systems, where performance, power efficiency, and qualification moats matter just as much.
And that mix shift—away from interchangeable bits and toward strategically irreplaceable ones—is the core driver behind Micron’s new profitability profile.
IX. Playbook: Lessons from the Memory Wars
Micron’s story isn’t just a history of chips. It’s a field guide for anyone trying to build—or invest in—a capital-intensive business where cycles are violent and mistakes are permanent.
Lesson 1: Location Can Be a Competitive Advantage
Micron’s Idaho origin wasn’t a quirky footnote; it was part of the strategy. A commodity memory company trying to survive on Silicon Valley cost structure and job-hopping culture would have been dead on arrival. Boise offered cheap hydroelectric power, affordable land, and a workforce that stayed put. If the game is making interchangeable bits cheaper than the next guy, you have to design your entire company around cost from day one.
Lesson 2: Survive the Cycles, Buy the Distress
Micron’s biggest leaps—Texas Instruments’ memory business, Elpida—came when the industry was down and sellers had no leverage. That only works if you manage boom times with restraint. Micron kept enough financial flexibility to act when others were forced to retreat. Countercyclical buying sounds obvious. In practice, it’s rare, uncomfortable, and often career-risky. It also built modern Micron.
Lesson 3: Technology Transitions Create Windows
Memory doesn’t stand still. Each transition—from 64K to 256K to 1M DRAM, from planar to 3D NAND, from DDR generations to HBM—reshuffled the deck. Micron’s recurring move was to keep investing through downturns so it could show up early when the market flipped. It hurt in the moment, but it’s how you earn the right to lead when the upcycle arrives.
Lesson 4: Know When to Exit
Micron has a surprisingly sharp kill switch. FRISC didn’t make sense, so it died. The PC business proved to be a brutal distraction, so it was spun out. And now even Crucial—an iconic consumer brand—was cut so supply could flow to bigger, faster-growing customers. Plenty of companies can start initiatives. Far fewer can admit a bet isn’t working and walk away before it becomes an anchor.
Lesson 5: Regulatory Relationships Matter
Micron’s position as America’s only advanced memory manufacturer makes it strategically important—and policy has become part of the playing field. CHIPS Act support didn’t materialize in a vacuum. Micron spent years building credibility and relationships in Washington so that when industrial policy finally arrived, it knew how to operate inside it.
Lesson 6: Customer Intimacy in B2B
The Nvidia relationship wasn’t a lucky break. It was earned through years of qualification work, tight engineering collaboration, and products built to exacting specs. In B2B semiconductors, that kind of partnership creates real switching costs—something commodity businesses usually don’t get. Micron didn’t just sell HBM. It integrated itself into the platform.
X. Bull vs Bear Case & Valuation
The Bull Case
The bull case is clean: AI is the biggest platform shift since the internet, and memory is one of the constraints. Micron expects the total addressable market for HBM to grow at roughly a 40% CAGR through calendar 2028—rising from about $35 billion in 2025 to roughly $100 billion in 2028. What’s telling is that Micron now thinks that $100 billion milestone arrives two years earlier than it did in its prior outlook. And even more telling: that 2028 HBM market would be larger than the entire DRAM market was in calendar 2024.
If AI demand keeps compounding—and hyperscaler capex plans suggest it will—Micron has a path to years of HBM growth at 40%-plus. The company also has real levers that matter in this fight: cost advantages through leading-edge DRAM nodes like 1-gamma, advanced packaging capability, and deepening relationships with Nvidia, AMD, and hyperscalers.
Then there’s margin structure. Micron’s fiscal Q2 guidance of roughly 68% gross margin would have sounded impossible in memory not that long ago. If this mix shift toward HBM and high-value data center products is durable—and not just a peak-cycle mirage—then Micron arguably deserves a fundamentally higher valuation multiple than the market has historically assigned to “a commodity memory company.”
The Bear Case
The bear case comes in three parts:
First, China risk. Micron has said that limits on selling to Chinese companies tied to key infrastructure projects could cost it a high single-digit percentage of annual revenue. Beijing announced sanctions in May 2023, and Micron planned to exit China’s data center semiconductor market after its business failed to recover from the 2023 ban restricting its products in critical infrastructure. China is both lost demand and a reminder that memory sits in the middle of an escalating tech war.
Second, cyclical risk. Memory has always been a boom-bust business. If AI capex slows, or if hyperscalers over-ordered and then digest inventory, pricing can roll over fast. Even with management sounding confident, investors have to watch the traditional cycle—especially in NAND and “plain” DRAM—because momentum in this industry can flip in a quarter.
Third, competition. Samsung may have stumbled on HBM3E qualification, but it has enormous resources and it won’t voluntarily give up the AI memory market. SK Hynix remains the leader in HBM. And Chinese players like CXMT are expanding aggressively with government support.
Porter's Five Forces Analysis
Supplier Power: Low. Equipment suppliers (ASML, Applied Materials) sell to many chipmakers and don’t have a clean ability to dictate terms to Micron.
Buyer Power: Moderate and rising. Hyperscalers (Google, Amazon, Microsoft, Meta) are huge, sophisticated buyers, though current HBM supply constraints still tilt leverage toward suppliers.
Threat of Substitutes: Low in the near term. There’s no practical substitute for HBM in large-scale AI training today. Alternatives like processing-in-memory are still years away.
Threat of New Entry: Low. Leading-edge fabs cost on the order of tens of billions of dollars and take years to build, and memory manufacturing depends on decades of accumulated process know-how.
Competitive Rivalry: High. Three major players—Samsung, SK Hynix, and Micron—are competing aggressively, even if HBM has meaningfully higher barriers than commodity memory.
Hamilton Helmer's 7 Powers Framework
Scale Economies: Strong. Memory manufacturing is a scale game, end to end.
Network Effects: Limited. This isn’t a network business.
Switching Costs: Moderate and rising. HBM is qualified over long cycles—often 12 to 18 months—and once qualified, customers don’t switch casually.
Cornered Resource: Moderate. Micron’s U.S. manufacturing base could become more strategically valuable as supply chain security concerns rise.
Branding: Largely irrelevant. Enterprise buyers purchase on specs, performance, and qualification—not consumer brand pull.
Counter-Positioning: Limited. Competitors can pursue similar strategies if they can execute.
Process Power: Strong. Manufacturing and yield know-how built over decades is hard to replicate, and it’s a real source of advantage in memory.
Key Metrics to Watch
If you’re tracking Micron from here, two KPIs matter most:
-
HBM Revenue and Market Share: This is the clearest read on whether Micron is truly escaping the commodity trap. Management’s 20%+ market share ambitions by 2026 should be followed closely, quarter by quarter.
-
DRAM Gross Margin: This is the pulse of both Micron’s execution and the industry’s supply-demand balance. If Micron can sustain DRAM margins above 40%, it suggests something structural is improving—not just a cycle peaking.
XI. Epilogue: The Future of Memory
As we write in January 2026, Micron stands at an inflection point. “In fiscal Q1, Micron delivered record revenue and significant margin expansion at the company level and also in each of our business units,” said Sanjay Mehrotra. “Our Q2 outlook reflects substantial records across revenue, gross margin, EPS and free cash flow, and we anticipate our business performance to continue strengthening through fiscal 2026.”
And the most important thing about that quote is what it implies: this isn’t just a lucky quarter. Micron is trying to convince the world it has earned a different future.
The technology roadmap extends far beyond today’s products. Micron shipped HBM4 samples with bandwidth above 2.8 TB/s and pin speeds exceeding 11 Gbps, positioning them as leaders in both performance and efficiency. In April 2025, JEDEC released the official HBM4 specification, supporting transfer speeds of up to 8 Gb/s across a 2048-bit interface, with total bandwidth up to 2 TB/s.
Beyond HBM4, the next shifts are about architecture, not just faster parts. CXL (Compute Express Link) promises to change how memory connects to processors, making memory feel less like a fixed, local resource and more like something you can pool and share across systems. Processing-in-memory—where computation happens inside the memory chips rather than on a separate processor—could push that idea even further, blurring the boundary between compute and memory entirely.
Micron is also building the physical capacity to meet what’s coming. Its HBM advanced packaging facility in Singapore is on track to contribute meaningfully to HBM supply in calendar 2027. As HBM becomes a larger part of Singapore’s manufacturing footprint, Micron expects opportunities for synergies between NAND and DRAM production. The company is also pleased with progress on its assembly and test facility in India, which has initiated pilot production and is expected to ramp in 2026.
Then there’s a second demand wave forming behind the data center: edge AI—models running on devices, not just in server farms. Smartphones, laptops, and autonomous vehicles all require more local AI processing, and that pushes memory requirements up with it. In the end, every AI story becomes a memory story.
Can Micron Finally Break the Commodity Curse?
The central question remains: is Micron’s HBM-driven transformation a temporary peak, or the start of a permanent shift in the industry’s economics?
Several forces point to “different this time”:
- HBM requires specialized manufacturing that can’t be quickly repurposed
- Customer qualifications create sticky relationships
- U.S. government support provides a cost advantage competitors can’t easily replicate
- AI demand appears structural rather than cyclical
But memory executives have been fooled before. The PC boom looked structural—until it wasn’t. The smartphone boom looked limitless—until saturation hit. AI could still follow the same pattern, just at a larger scale.
American Manufacturing Resilience
One of the most remarkable parts of Micron’s story is simply that it exists. Every other American memory manufacturer eventually walked away. Intel, Texas Instruments, IBM—they all exited memory to focus on logic or other businesses.
Micron survived for a messier set of reasons: stubborn Idaho founders who wouldn’t quit, a billionaire potato farmer who understood commodity cycles, and generations of employees who kept tightening yields and costs through downturns that wiped out “better” companies.
Whether Micron truly capitalizes on the AI revolution—breaking free of memory’s commodity past—still isn’t guaranteed. But after nearly fifty years of outlasting the field, betting against Micron has never been a comfortable position.
Micron’s record fiscal Q1 2026 results didn’t end the memory cycle. Nothing ends the memory cycle. But they did prove something meaningful: Micron can sit at the center of the most important computing buildout in a generation, selling products that customers can’t easily substitute and can’t easily delay.
From a basement in Boise to the beating heart of AI infrastructure—that’s the Micron story. And as of January 2026, its next chapter is the one the whole industry has been waiting for.
Spotify
Apple Podcasts
Amazon Music
Audible
YouTube
Castbox
Pocket Casts