Author:Wang Jian
Lishi Business Review丨Produces
Another trillion-dollar giant is born. On the evening of May 26th, Micron Technology's stock price surged, pushing its total market capitalization past the $1000 billion mark.
Rooted in Boise, a small inland city in the US with no semiconductor industry foundation, Micron Technology, born in 1978, now firmly occupies a top-three position in the global memory chip market, sharing the DRAM market with Samsung and SK Hynix. Having weathered multiple industry cycles and shakeouts where Japan's memory industry nearly faded and American peers exited one by one, Micron alone survived and stood firm. Its path to survival is filled with controversy and mystery.
Throughout its development, Micron lacked policy protection and substantial capital backing. Yet, it repeatedly navigated industry crises by leveraging political and legal tools: filing dumping complaints against Japanese firms early on, acting as a whistleblower to escape antitrust charges, and persistently lobbying to influence industry competition, earning it the label of "political opportunist." Political leverage merely bought it breathing room. Extreme cost control in manufacturing and decades of engineering accumulation allowed it to produce smaller chip die sizes and higher wafer yields, enabling it to withstand the impact of industry price cycles.
However, strategic misjudgments sowed hidden dangers. The acquisition of Elpida caused it to miss the golden decade of HBM development, leaving it significantly behind in the high-end AI era. Currently, Micron is caught in a triple squeeze: a vast gap in HBM market share, its mid-to-low-end market being eroded by Chinese manufacturers, and a sharp drop in its core market share in China. While desperately catching up to repay its technical debt, it also faces a new round of industry competition. Whether this chip giant, which has relied on unique strategies and hardcore manufacturing to stand its ground, can navigate future cycles and maintain its industry position is closely watched by the market. Below, Enjoy:
Micron Technology is one of the world's top three memory chip manufacturers, alongside Samsung and SK Hynix, holding roughly one-fifth of the global DRAM market.
This is actually quite surprising.
Founded in 1978, Micron Technology is based in Boise, Idaho—an inland small city in the US with no semiconductor industry foundation. Throughout its growth, it lacked the protective government industrial policies of its competitors, lacked massive capital support, and didn't even possess a sufficiently deep technological moat.
Yet, as the global memory industry experienced cyclical crashes one after another, with former American peers exiting the scene and Japan's memory industry nearly cleared out, Micron Technology managed to survive time and again.
Why is that?
The answer might lie in an unsavory detail: At its three most dangerous junctures, Micron's first reaction was not to accelerate technological investment, but to pick up the phone and call Washington for help.
This isn't to say Micron lacks genuine technical capability; its manufacturing cost control has long been among the most competitive in the industry. But its ultimate survival and longevity are underpinned by a survival logic rarely discussed head-on. And the limits of this logic are being re-examined in this era.
01 Unintentionally "Feeding" a Rival
By early 1985, Micron was the last remaining US-based DRAM company still standing.
DRAM (Dynamic Random-Access Memory) is like the "scratch paper" for electronic devices, a temporary storage space for CPU data. Without it, even the strongest CPU can't function. At the time, six Japanese electronics giants, backed by government industrial policies, were dumping products below cost, squeezing American competitors out of the market one by one.
Micron's situation was simple: either find another path or become the next casualty. However, Micron's choice was: pick up the phone and call Washington.
In June 1985, Micron officially filed a complaint with the U.S. Department of Commerce, accusing Japanese firms of dumping DRAM. As the only domestic DRAM company, the US government could not sit idly by and pressured Japan. In 1986, the U.S.-Japan Semiconductor Agreement was signed, forcing Japanese companies to accept export price controls. Reports indicate that in the following years, Micron's DRAM sales increased tenfold.
But this victory planted an unexpected consequence: While the agreement temporarily held back Japan, it ceded market space to a player no one paid much attention to at the time—South Korea's Samsung.
At the time, Samsung's DRAM technology was just starting, struggling to compete directly with Japan. Ironically, Micron's dispute with Japanese manufacturers presented a rare development opportunity. Even more ironically, Samsung's entry point into the DRAM race was a 64K DRAM technology license obtained from Micron. In its early days, to earn a hefty licensing fee, Micron had handed over a production license to Samsung.
In fact, when Samsung obtained this license, it was far smaller than Micron, with almost zero brand recognition. But backed by systemic support from the South Korean government and the chaebol system, it was willing to continuously invest despite losses, weathering cyclical downturns with a capital patience that Micron could not replicate.
By the mid-1990s, Samsung's DRAM capacity had surpassed Micron's; by the 2000s, it was firmly the world's largest memory chip manufacturer, a position it holds to this day. It can be said that Micron hand-fed its most formidable competitor for decades to come.
Nevertheless, Micron managed to recover by "filing complaints." However, it played the same survival card again in 2002.
That year, the U.S. Department of Justice launched an antitrust investigation into the DRAM industry, accusing several manufacturers of colluding to fix memory prices. Samsung, SK Hynix, and Germany's Infineon were fined over $600 million combined. At the time, Micron was also within the scope of the investigation.
However, instead of waiting for the investigation to proceed, Micron proactively contacted the Justice Department after the case had formally begun, while itself being a potential defendant, offering internal evidence to implicate its peers in exchange for immunity.
Reporting peers to gain protection, acting as a "whistleblower," is standard practice under US antitrust law. But in an industry highly dependent on multilateral relationships, this move by Micron was not a good look. In the end, Samsung, Hynix, and Infineon were fined, while Micron walked away unscathed.
In both crises, Micron extricated itself through somewhat unsavory political means, earning it the moniker "political opportunist" within the industry. In the fiercely competitive business world, lacking any structural advantages, Micron found a way to survive, which in itself is a kind of capability.
However, "all the gifts of fate come with a price tag attached in secret." Micron would have to pay a price. And Micron's price is hidden in its 2013 acquisition.
02 Missing the Golden Decade of HBM
In February 2012, Steve Appleton, the CEO who had led Micron through its long fluctuations, died unexpectedly in a private plane crash. His successor, Mark Durcan, admitted the first thing upon taking office was an ongoing acquisition negotiation.
In July 2013, Micron completed the acquisition of Elpida Memory for approximately $2.5 billion. Elpida was the last legacy of Japan's memory industry, formed from the merger of Hitachi and NEC's memory divisions, which had filed for bankruptcy in 2012 due to overwhelming debt.
On the surface, this seemed like a victory. But the technological legacy Elpida left behind was weaker than imagined. Yukio Sakamoto, Elpida's last president, stated at the bankruptcy press conference that "Elpida's technical level is very high." That wasn't wrong, but that technical level referred to a different path.
Before its bankruptcy, Elpida bet on mobile DRAM, following the smartphone market. The HBM (High Bandwidth Memory) technology path was almost non-existent on its strategic map.
What is HBM?
If DRAM is the computer's "temporary scratch paper," then HBM is its "premium 3D version." It's like stacking multiple layers of DRAM chips vertically like a sandwich, connected via thousands of tiny channels, offering bandwidth up to 10 times faster than regular memory. Regular DRAM is like a "single-story bungalow," while HBM is a "multi-story parking garage." Although made of the same material, HBM is specifically designed for AI chips (like NVIDIA GPUs), costs 5-10 times more, and determines the ceiling of AI computing power.
What Micron inherited wasn't just Elpida's 16,000 engineers but an entirely different manufacturing process system. Reports indicate that in 2014, the acquired Elpida factories contributed 54% of Micron's global DRAM output. However, over a year after the merger was completed, more than half of the company's capacity was still running on two separate process systems due to incompatibilities in processes, equipment, and parameters between the Hiroshima and Boise factories, causing significant waste.
In fact, Micron clearly listed risk factors in subsequent annual reports, explicitly acknowledging issues including "integration problems with products and process technologies."
Meanwhile, in 2013, the very year Micron completed the acquisition, what was then renamed SK Hynix (formerly Hyundai Electronics) released the world's first HBM chip. This HBM vertically stacked multiple layers of memory chips, connected through tiny vias about 10 micrometers in diameter and 100 micrometers deep (thousands per layer), directly linked to the GPU, potentially increasing data throughput several to tens of times.
Regrettably, for the first few years after SK Hynix's product launch, there was almost no commercial market. But on the HBM track, the value of time had already been quantified into an insurmountable market barrier.
At the end of 2022, the emergence of ChatGPT instantly ignited demand for AI computing power, pushing memory bandwidth to become the core bottleneck of the entire system. Silicon Valley engineers noted that when training GPT-4, about 90% of the time was spent on data transfer, not actual computation, with HBM being the key to unlocking this bottleneck.
Thus, SK Hynix, having laid the groundwork a decade earlier, seized the opportunity, starting to supply HBM3 to NVIDIA from June 2022, while Micron only released its own HBM3 product in July 2023. A mere one-year gap was magnified into a huge chasm in the rapidly developing AI market.
At that time, for the urgently needed HBM3, SK Hynix held about 85% of the market share, while Micron, having missed a decade of golden development, held only about 3%. This precisely reflects a fundamental rule of the AI era: the time that cannot be bought with money is the real value in this competition.
However, for the party at a disadvantage in accumulated time, it resorted to its habitual move yet again.
03 The Repeated "Complaint" Play
In 2017, Micron's legal team acted again. The caliber of its opponents had shrunk, but its tactics were exactly the same—simple and brutal.
The first two times, opponents were established industrial giants: Japan's six major electronics conglomerates, Korea's Samsung, and the price cartel formed with SK Hynix. This time, Micron's target was a newly established, not-yet-mass-producing Chinese startup— Fujian Jinhua Integrated Circuit (JHICC).
Micron accused Fujian Jinhua and Taiwan's United Microelectronics Corporation (UMC) of conspiring to steal its DRAM trade secrets. This transnational lawsuit quickly escalated into political action.
In October 2018, the U.S. Department of Commerce placed Fujian Jinhua on its Entity List for export controls, cutting off its access to American equipment and technology. A Chinese memory company that had just built its wafer fab without achieving mass production was thus strangled at the starting line.
Throughout this process, Micron's playbook for dealing with competition was identical to before: lead with legal means, finish with government force, competitor out.
In the following years, Micron persistently pushed Washington to tighten controls on China's memory industry. According to public filings, from 2018 to 2022, Micron spent approximately $9.54 million on political lobbying in the U.S., with about 67% of that lobbying content related to China.
In 2022, Micron announced a $100 billion investment to build a new wafer fab in New York State, located precisely in the district of Senate Majority Leader Chuck Schumer—one of the main proponents of the CHIPS Act, from which Micron is also a beneficiary of subsidies.
The first two "complaints" were won using this strategy, but by 2023, the situation reversed.
That May, China's Cyberspace Administration announced the completion of a cybersecurity review of Micron products, determining they "posed relatively serious cybersecurity risks," and prohibited critical information infrastructure operators from purchasing Micron products.
Micron's CFO responded publicly, stating the ban's impact on company revenue was "only in the single digits." But that wasn't the case.
Having established its presence in China early on, Micron's revenue from China once accounted for a significant portion of its global total, making the losses severe. According to Micron's financial reports:
-
Fiscal Year 2023: Due to China's countermeasures, Micron's revenue share from China dropped to 14%.
-
Fiscal Year 2024: Further dropped to 12.1%.
-
Fiscal Year 2025: That figure had fallen to 7.1%.
By the end of 2025, Micron had to announce its exit from the Chinese data center server chip business. Faced with China's strong countermeasures, Micron did not escape unscathed this time. This setback was not an isolated incident but a concentrated eruption of the systemic dilemmas Micron has long faced.
04 Dilemma Under Triple Squeeze
In the semiconductor field, unable to enter the high-end, seeing the low-end eroded, and the window to the Chinese market already closed. These three events interlock within the same timeframe, forming a series of unavoidable serious problems for Micron.
-
First Squeeze: Inadequate High-End Catch-up
Micron was the second manufacturer to pass NVIDIA's certification for HBM3E, ahead of Samsung, truly getting on the starting line. But this "second place" came at a cost. By the time it got certification, SK Hynix had already begun its next-generation product ramp and was continuously optimizing yields for the generation after that, putting immense pressure on Micron. Industry analysis suggests that even in the nearly comparable HBM3E phase, Micron's market share remains under 20%, while SK Hynix's share has long stabilized above 60%.
-
Second Squeeze: Downstream Market Erosion
With ChangXin Memory Technologies (CXMT) aggressively expanding in mid-to-low-end DRAM at prices about one-third below market rates, its 2025 shipments grew approximately 50% year-over-year, rapidly expanding its market share from near zero to about 7%. Mid-to-low-end DRAM has long been Micron's most stable cash flow source. As pricing power in this segment shrinks, it severely impacts the revenue Micron relies on to fund high-end R&D. For Micron, failure to catch up in the high-end means difficulty expanding share in high-margin products; erosion in the low-end means the cash flow supporting R&D is narrowing.
-
Third Squeeze: Losing the Chinese Market
China's ban deprived Micron not just of orders but of an irreplaceable opportunity to participate. From 2023 to 2025 was precisely the concentrated explosion period for AI infrastructure construction by Chinese tech companies. This demand included large volumes of high-bandwidth memory and high-end DRAM—exactly what Micron wanted to sell—but it couldn't secure a single deal. Moreover, Chinese tech companies' AI server supply chains smoothly completed their builds without Micron, with SK Hynix and Samsung securing those certification spots.
These successive setbacks have led outsiders to label Micron a "political opportunist." But this only explains part of its survival strategy, not how it survived the brutal industry cycles. The underlying capability that truly supported Micron through storms is actually its unparalleled manufacturing cost control.
05 Technological Time Accumulation is Key
It's true that Micron survived using unsavory political means and even suppressed various competitors with them. But objectively, Micron only bought itself breathing room, temporarily suppressing opponents; it couldn't fight price wars or endure cyclical lows on its behalf. Competition is something it must face itself.
Samsung and SK Hynix have the chaebol system behind them, allowing continuous investment despite years of losses, enduring until the next cycle reversal. But Micron lacks this structure; it has no parent body for continuous transfusions, and each round of investment must be earned back after each price war itself. This cannot be achieved merely by "filing complaints."
This forced Micron to relentlessly focus on one thing: continuously improving technology to push manufacturing costs lower than competitors', enabling it to hold out a breath longer than others when prices crash. This capability is also the crucial foundation for Micron's survival to this day and its relatively good standing.
As publicly stated by Micron CEO Sanjay Mehrotra:
The cell area of Micron's DRAM chips is approximately 66.26 square millimeters, smaller than Samsung's 73.58 square millimeters and SK Hynix's 75.21 square millimeters.
This means: from the same wafer, Micron can yield more chips than its competitors, naturally resulting in lower unit costs.
This advantage wasn't gained through subsidies or chaebol infusions; it was earned through forty years of engineering accumulation. For Micron, political means are leverage, buying critical time windows, but excellent manufacturing efficiency is the true factor enabling it to stand firm in manufacturing. These two are not independent but an interlocking survival system; lacking either, Micron wouldn't be where it is today.
However, this combination also has its unavoidable limits. Political means and manufacturing efficiency are competitive capabilities on existing tracks; while they help Micron survive, they cannot substitute for the time needed to lay groundwork on new tracks. Micron used forty years of accumulated cost advantages to survive, but on the new HBM track, it felt the expensive cost behind the "time gap."
Now, Micron has secured an HBM3E certification spot, with production capacity struggling to ramp up, while the window for the next-generation HBM4 has already opened. Simultaneously, the company continues to increase R&D investment, deepen cooperation with NVIDIA, and leverage the CHIPS Act to deploy new product lines. The essence of all these efforts is to repay the time debt incurred in the past.
After all, certification is just an entry ticket; moving from entry to stable mass production and then to profitability remains a marathon that can only be won through accumulated time. But the opponents never stopped. While Micron struggles to fill HBM3E capacity gaps, leaders are already optimizing yield curves for the next-generation HBM4.
And when competition ultimately evolves into a contest of "patience," can a company adept at using political leverage to buy time and manufacturing efficiency to digest cycles win the next competition that requires the test of time?
The answer for Micron still lies within the yet-to-be-perfected HBM4 wafers, hidden in a long, patient wait that requires truly settling down.
