'Will K-chips go the way of LCD?' — CXMT rides Apple ambitions to challenge Samsung, SK hynix

Welcome back to Chip Chip Fab Fab. This week, we need to talk about CXMT — China's largest memory chip manufacturer — again.

Until last year, there was little reason to cover CXMT so frequently. Things changed after the company listed on the Shanghai Stock Exchange in July this year, accelerating its growth to a pace that now poses a genuine threat to Korea's memory chipmakers.

The clearest example is LPDDR6 mobile DRAM. CXMT announced mass production of the chip before Samsung Electronics or SK hynix had made any official mass-production declaration of their own.

Some in the industry still argue that Chinese DRAM has a long way to go before it matches Korean DRAM on performance — and that US sanctions, which make it difficult to procure cutting-edge equipment such as EUV lithography tools, are buying Korean firms valuable time.

An executive at a semiconductor materials, components and equipment company I met recently echoed that view. "China tends to announce first," he said. "Korea announces when it can actually mass-produce; China announces before mass production even begins." He cautioned against excessive alarm.

But he immediately added a caveat.

"Semiconductors are ultimately a matter of time," he said. "Keep repeating the process and yields go up. Sooner or later, it will become a threat to Korea."

That is precisely what worries me most — the possibility that semiconductors could follow the same path as displays.

In the mid-2010s, when China was rapidly closing the gap with Korea in LCD panels, the conversation sounded remarkably similar to today's. The prevailing view was that Chinese products lagged Korea in technology and that a meaningful gap remained.

With good reason, at the time: LG Display held the top spot in global large-panel LCD shipments as recently as 2016. China's BOE was rapidly ramping up volume, but its product mix skewed toward low-cost, commodity panels, and analysts generally rated Korean manufacturers as technically superior.

The consensus then was that the Korea-China LCD technology gap stood at roughly one year, with Korea enjoying a two-to-three-year window to act.

Events moved faster than anyone expected. Within just two or three years, China had climbed to the top of global LCD shipment rankings. Samsung Display and LG Display subsequently exited the domestic LCD market as profitability collapsed — all within three years of analysts declaring that a one-year technology lead still existed.

And where does the display industry stand today?

Through the 2010s, displays were one of Korea's flagship export industries, sustaining the economy alongside semiconductors. Today, interest in the sector has faded sharply.

I recently attended a display conference organized by market research firm Omdia, where one presenter noted that attendance has been shrinking every year — down to roughly a third of what it once was. Omdia itself has cut its domestic display conference from twice a year to once.

Talent recruitment is another problem. I remember a job-seeker I met at the Korea Display Industry Exhibition in June. A student enrolled in a display-related program, he told me that eight out of every 10 of his friends were preparing to enter the semiconductor industry instead.

Sustaining a technology lead ultimately requires people — but young talent is gravitating toward semiconductors, not displays. One display industry executive sighed as he told me: "This is exactly the moment we need to be widening the technology gap, but finding the right people is incredibly hard."

That is why China's semiconductor ambitions deserve closer scrutiny right now. Korea cannot afford to keep repeating the pattern of taking comfort in a current technology lead, only to cede market dominance before anyone notices.

Enough preamble. Let me get to why I am focusing on CXMT again this week.

Start with data released Thursday by market research firm Counterpoint Research. Samsung Electronics led the global DRAM market in the second quarter of this year with a 38 percent share.

The more striking figure belongs to CXMT. The Chinese chipmaker held a 10 percent share in the second quarter, cementing its position in fourth place — up 2 percentage points from the previous quarter.

Notably, a significant portion of CXMT's gain appears to have come at SK hynix's expense. SK hynix's second-quarter share fell to 25 percent, down 4 percentage points from 29 percent the prior quarter. Over the same period, both Micron and CXMT each added 2 percentage points.

CXMT's ascent has been steep. As recently as 2023, the company held less than 1 percent of the global DRAM market. It climbed to 4 percent in the second quarter of last year, reached 8 percent in the first quarter of this year, and has now broken through 10 percent.

"CXMT expanded its share from 4 percent in the same period last year to 10 percent in the second quarter of this year, raising its profile in the global DRAM market," Counterpoint Research said.

Industry observers generally regard the 13-to-15 percent range as a tipping point — the threshold at which economies of scale kick in fully and growth can accelerate sharply.

CXMT is already ahead of schedule. Counterpoint Research and investment bank UBS had previously forecast that CXMT would reach a 10 percent shipment share around 2028. Instead, the company crossed that threshold two years early.

The technology gap is narrowing as well. Last week's Chip Chip Fab Fab reported that CXMT's LPDDR6 was set to appear in Xiaomi's latest flagship foldable smartphone. This week, both Xiaomi and CXMT confirmed that publicly — and CXMT went a step further, announcing it had begun mass production of LPDDR6. The announcement came before Samsung Electronics or SK hynix had made any official mass-production declaration for the same chip.

CXMT is closing in on high-bandwidth memory as well. According to The Information, a US technology publication, CXMT has already begun low-volume production of HBM3E, the fifth-generation HBM standard, and is conducting product tests with Chinese AI accelerator developers including Alibaba and Cambricon. The company plans to begin large-scale mass production as early as 2027.

The industry had originally expected CXMT to begin mass-producing HBM3 this year, but its development timeline has moved one to two years ahead of schedule, pushing the company directly into HBM3E — the generation after.

HBM3E is a product that gave even Samsung Electronics roughly two years of difficulty during development and customer qualification. Samsung began mass-producing HBM3E in April 2024, so if CXMT successfully achieves large-scale production in 2027, the gap in mass-production timing would narrow to roughly three years.

Until recently, the domestic consensus placed the Korea-China HBM technology gap at as much as five years. Measured by product generation alone, that gap has now shrunk to just one generation.

CXMT is also turning a profit. When Chinese display makers were chasing Korea, they absorbed heavy losses and fought a war of attrition to build market share.

CXMT's situation is different.

The company posted revenue of 150.31 billion yuan ($22.4 billion) in the first half of this year alone and recorded a net profit of 77.60 billion yuan, swinging to profitability. CXMT has moved beyond simply flooding the market on the back of government support — it is now generating its own earnings and building the financial capacity to invest further.

CXMT has also filed a lawsuit against the US government in a bid to re-enter the American market. The company is seeking to have its blacklisting by the US Defense Department overturned, arguing it has never supported the Chinese military and that its chips are designed and produced exclusively for civilian and commercial use.

In an official statement issued after filing the suit, CXMT said: "CXMT is not a military company and has no ties whatsoever to the Chinese military. Since being placed on the list in January 2025, we have suffered ongoing reputational damage and commercial harm. We have taken legal action to protect the company's legitimate reputation and business interests."

With CXMT on the blacklist, Apple has been unable to easily incorporate CXMT memory into its devices. Apple has nonetheless been lobbying for permission to use CXMT memory in China-market devices only, as a cost-cutting measure. Even the prospect of CXMT entering Apple's supply chain represents a significant threat to the three major memory makers: Samsung Electronics and SK hynix together are estimated to supply more than 70 percent of Apple's mobile DRAM.

Until now, US policy has effectively bought time for Korean memory chipmakers — but that dynamic is beginning to shift. President Donald Trump has expressed willingness to meet North Korean leader Kim Jong-un. In May this year, Trump visited China and held a summit with President Xi Jinping, and Xi is set to travel to Washington for another summit in September. In November, Trump will visit China again for the APEC summit in Shenzhen. The once-frozen US-China relationship is quietly changing.

A full lifting of US semiconductor sanctions on China through these meetings remains unlikely, but speculation persists that Washington could ease restrictions enough to allow CXMT memory in Apple devices sold in China.

Washington has also recently threatened Korea with targeted tariffs if Korean chipmakers do not produce memory on US soil, while using Taiwan as leverage — Taipei has been increasing direct investment, and TSMC is building factories in the United States, indirectly expanding the investment footprint. This looks like a moment when the Korean government needs to stand firmly behind its semiconductor industry.