Changxin's battle is more difficult than that of BOE and NIO

The highly anticipated Changxin Technology has finally entered the final phase before its planned public listing.

In the early hours of July 9, 2026, Changxin Technology released its IPO prospectus for the Shanghai Stock Exchange Sci-Tech Innovation Board: the company plans to issue 6.688 billion shares, raise 29.5 billion yuan in funds, and open subscriptions on July 16.

The announcement quickly sparked widespread discussion across the market.

The most widely debated figures from the prospectus show that from 2022 to 2024, the company recorded a cumulative loss of 36.6 billion yuan; in the first quarter of 2026, it achieved quarterly revenue of 50.8 billion yuan and net profit of 33 billion yuan, representing a year-on-year increase of 1268%.

In other words, the profit generated in a single quarter alone surpassed the total losses accumulated over the previous three years.

In the history of China's A-share market, companies that have staged performance turnarounds are not uncommon, but a turnaround of this magnitude is extremely rare.

What exactly has this company gone through?

01. The Ambition Brought Back from Silicon Valley

To talk about Changxin Technology, we must start with a man named Zhu Yiming.

In 1994, Zhu Yiming graduated with a bachelor's degree in physics from Tsinghua University, then went to Stony Brook University, State University of New York, to pursue a degree in electrical engineering. After working in Silicon Valley for several years, he returned to China in 2005 with a portfolio of SRAM patents and $100,000 in seed funding, and founded GigaDevice.

Later, GigaDevice grew into a leading global supplier of NOR Flash memory chips, but Zhu Yiming's ambition extended far beyond that.

Compared to NOR Flash, DRAM represents a far larger market. After all, DRAM is the most critical and challenging segment of the global semiconductor industrial chain, and what Zhu Yiming truly wanted to do was to develop DRAM technology.

But DRAM is not a business that can be entered on a whim.

For more than two decades, the global DRAM market has long been monopolized by three giants: Samsung, SK Hynix, and Micron. The patent barriers accumulated over 40 years, capital investments running into tens of billions of dollars, and extremely complex manufacturing processes are each enough to deter the vast majority of entrepreneurs.

Zhu Yiming did not flinch.

In 2016, he partnered with the Hefei municipal government to launch a DRAM project codenamed "Project 506". This project later became the predecessor of Changxin Memory Technologies.

But the real challenge soon emerged: where would the technology come from?

The answer points to a defunct German company — Qimonda.

02. The "Foundation" of Changxin

Younger investors may never have heard the name Qimonda.

But in the history of DRAM development, it was once an unavoidable player. Qimonda was originally a subsidiary of Infineon (spun off from Siemens' semiconductor division in 1999) and was once the largest DRAM manufacturer in Europe. In January 2009, under the dual impact of the global financial crisis and a sharp downturn in the memory industry, Qimonda collapsed and declared bankruptcy.

But it left behind an extremely valuable legacy: approximately 7,000 DRAM patents, and a complete memory cell architecture independent of the Samsung-Hynix-Micron triangle — the BWL (Buried Wordline) technology.

In 2015, Canadian patent management firm Polaris Innovations purchased these patents from Infineon for approximately 30 million euros. In 2019, Polaris signed a licensing agreement with Changxin, granting Changxin access to around 2.8TB of Qimonda's technical documentation.

In other words, the technical foundation of Changxin's DRAM business was built upon the "legacy" of a bankrupt German company.

But patents and blueprints alone are far from enough. DRAM manufacturing is not simply assembling components according to a diagram; it contains countless implicit know-how that cannot be written into patent documents — which design choices to retain, which to discard, yield optimization tricks on the production line, and fine-tuning margins for process parameters.

Changxin was fortunate that it not only obtained the blueprints but also "inherited" key personnel.

After Qimonda closed down, Karl-Heinz Kuesters, the vice president of technology at its German factory, joined Changxin as a technical consultant. Kuesters is a veteran in the DRAM field, having worked continuously at Siemens, Infineon, and Qimonda for 24 years, leading DRAM development for many years. EETimes once directly referred to him as Changxin's "ace in the hole".

At the same time, Changxin recruited Dr. Ping Erxuan, a former technology executive from Micron, SanDisk, and Applied Materials, to serve as vice president in charge of processes and materials. Together with more than 400 Chinese engineers from Qimonda's Xi'an R&D center, and a large number of equipment and process specialists recruited from South Korea and Taiwan, China, Changxin assembled a R&D team capable of overcoming tough challenges.

At this point, the three key elements — technology, patents, and talent — were finally in place.

But there was still one critical problem: capital.

03. Hefei's "Ten-Year Long-Term Partnership"

DRAM manufacturing is a classic "capital-devouring" industry.

An advanced production line often requires investments in the tens of billions of yuan, and the first few years after production starts are almost guaranteed to be loss-making. The reason is simple: yield rates take time to ramp up, capacity utilization cannot reach full levels in the short term, and huge depreciation expenses from equipment investments will continue to weigh on profit statements.

Changxin Technology is no exception to this rule.

From 2022 to 2024, Changxin recorded a cumulative loss of 36.6 billion yuan. By the time it achieved its first annual profit in 2025, its accumulated book losses still stood at around 36.65 billion yuan.

Who can tolerate a company losing money for ten years? The answer is: the Hefei municipal government.

The first phase of the "Project 506" launched in 2016 had a total investment of 18 billion yuan, with Hefei state-owned capital covering about 80%, or 14.4 billion yuan. In subsequent multiple rounds of financing, Hefei's state-owned holdings were diluted, but it never reduced its shareholding or exited the investment.

Prior to the IPO, Qinghui Jidian, the largest state-owned shareholder in Hefei, still held 21.67% of Changxin's shares, with total state-owned ownership exceeding 30%.

This is clearly no ordinary financial investment.

From the very beginning, Hefei positioned its investment in Changxin not as a fund project to exit in three to five years, but as a strategic layout with a ten-year cycle. To bet on this company, Hefei not only invested capital but also built a complete local supply chain system around Changxin's wafer fabs.

Packaging and testing firms such as Pudong Technology and Xinfeng Technology were built right next to Changxin's factory; gas supplier Guanggang operates an on-site bulk gas plant; Zhiwei Semiconductor, a subsidiary of Zhichun Technology, provides wafer recycling capacity; and state-owned venture capital directly holds controlling stakes in semiconductor mold equipment manufacturer Wenyi Technology.

In other words, Hefei did not bet on an isolated DRAM enterprise, but attempted to build a complete memory industry cluster centered around Changxin.

Hefei has already executed this strategy twice before with BOE and NIO: using state-owned capital as an "anchor" to stabilize core enterprises, then attracting upstream and downstream industrial chains, and finally forming a full industrial cluster.

But this battle for Changxin is more difficult than the ones for BOE and NIO.

Because DRAM is one of the most fiercely competitive semiconductor tracks in the world, Changxin is not facing ordinary rivals, but the three giants Samsung, SK Hynix, and Micron.

04. Tearing a Crack in the Old DRAM Order

Facing these three giants, Changxin had no shortcuts. It could only accumulate advantages bit by bit through years of perseverance, relentless technical efforts, and industrial chain collaboration.

After ten years of persistence, Changxin finally encountered a favorable market trend.

In 2025, a new AI-driven DRAM super-cycle officially began. The demand for memory chips from large model training and inference far exceeded market expectations, with DDR5 and LPDDR5X remaining in short supply and prices continuously rising.

Changxin thus reached its performance inflection point.

For the full year 2025, Changxin achieved revenue of 61.799 billion yuan, a year-on-year increase of 156%, and recorded its first annual profit, with net profit attributable to shareholders reaching 1.875 billion yuan.

But this was just the "appetizer".

What truly excited the market was its performance in the first quarter of 2026.

In this quarter, Changxin's revenue reached 50.8 billion yuan, a year-on-year increase of 719%; net profit hit 33 billion yuan, a year-on-year growth of 1268%. In other words, the quarterly profit already exceeded the sum of the company's losses in the previous three years. At the same time, its gross margin soared from less than 20% in 2025 to nearly 65%.

Even more remarkably, this high growth is continuing.

According to the performance forecast disclosed in the July 9 prospectus, in the first half of 2026, Changxin expects to achieve revenue of 110 billion to 120 billion yuan, a year-on-year increase of 612% to 677%; net profit attributable to shareholders is expected to reach 50 billion to 57 billion yuan, a year-on-year growth of 2244% to 2544%.

A simple calculation shows: if it can earn more than 50 billion yuan in the first half of the year and maintain this momentum for the full year, a 100-billion-yuan annual profit is not impossible.

Of course, from a global perspective, Changxin still lags behind industry giants. In the first quarter of 2026, Samsung, the world's top DRAM maker, recorded memory business revenue of about $50.4 billion, the second-ranked SK Hynix around $36.2 billion, while Changxin's revenue was approximately $7 billion.

The gap still exists, but the speed of catch-up is already enough to make the three giants wary.

More critically, Changxin's global market share is rising rapidly. In the fourth quarter of 2025, Changxin held about 7.67% of the global DRAM market; by the first quarter of 2026, this figure had jumped to around 11%.

SemiAnalysis predicts that by the end of 2026, Changxin's monthly production capacity will reach 350,000 wafers, making it poised to overtake Micron and become the world's third-largest DRAM maker; by 2028, Changxin targets to further increase its monthly production capacity to 500,000 wafers, aiming for a 17% global market share.

This means that Changxin is no longer just a "follower" in China's DRAM industry, but has begun to become a real variable reshaping the global memory market landscape.

05. The Most Painful Weakness

However, it must be noted that Changxin's current performance boom largely benefits from the "cyclical dividend" brought by the DRAM market upswing, rather than "technological transcendence".

The DRAM industry is currently experiencing its most profitable super-cycle in history, where even the most outdated production capacity can generate profits. But when prices fall and supply and demand reverse, it is technological generation gaps that will determine who can survive.

In terms of manufacturing processes, Changxin's current mainstay is at the 16nm level (self-defined), with DDR5 speeds of 8000Mbps and a yield rate of around 80%. In contrast, Samsung and SK Hynix's main processes are already at 1b nm (12-13nm), their latest 1c nm (11-12nm) node entered mass production in 2025, and their DDR5 yield rates exceed 95%.

The gap is roughly 1.5 to 2 generations.

More critically, HBM (High Bandwidth Memory) — the most core DRAM product in the AI computing era and the category with the highest profit margins — remains a gap. SK Hynix's HBM revenue already accounts for more than 40% of its total DRAM revenue; Samsung's HBM4 achieved the world's first mass production and shipment in February 2026, and HBM4E will release samples in the second quarter.

But Changxin's HBM program is still in the early stages of research and development. The 9 billion yuan allocated for forward-looking technology R&D in the prospectus is largely intended to close this HBM gap.

There is a harsh reality here: Changxin can profit from general-purpose DRAM because the market cycle is favorable and prices are high; but it cannot earn a single cent from HBM revenue.

And HBM happens to be the most lucrative profit pool in the DRAM industry over the next decade.

06. Raising Capital Is Not Just About Money, But About Buying Time

With this background in mind, the purpose of Changxin's 29.5 billion yuan fundraising makes perfect logical sense.

The prospectus discloses that 13 billion yuan of the raised funds will be used for technical upgrades and renovations of memory wafer manufacturing production lines, 9 billion yuan for DRAM technology upgrades, and a portion for forward-looking technology R&D.

In other words, the funds will serve three things: expand production capacity, catch up in process technology, and fill the HBM gap.

For Changxin today, what it lacks most is not actually money. In the first quarter of 2026, the company's net profit reached 33 billion yuan; its 2025 operating cash flow was 5.3 billion yuan. In other words, amid the current prosperous market cycle, Changxin already has strong self-sustaining profitability.

What it truly lacks is time.

Catching up in process technology takes time, and HBM development from R&D to mass production also requires at least two to three years. During these two to three years, Samsung and SK Hynix will not stop and wait for Changxin. More harshly, there is a fundamental law in the semiconductor industry: once the technology generation gap widens to more than three generations, the difficulty for latecomers to catch up will rise exponentially.

This situation is very similar to Semiconductor Manufacturing International Corporation's (SMIC) struggle to catch up with TSMC in the past. SMIC spent many years trying to catch up at the 28nm node, while TSMC had already advanced all the way to 3nm. In the DRAM field, Changxin faces a similar dilemma — except that this time, it has caught a strong tailwind from the explosion of AI computing demand, giving it sufficient financial resources.

But the problem is that technological catch-up can never be achieved simply by throwing money at it.

Fortunately, Changxin still has two strong cards in its hand.

The first card is a 20-billion-yuan large order from Tencent.

At the end of June 2026, Tencent signed a long-term supply agreement for server DRAM chips worth over 20 billion yuan with Changxin. This marks the first large-scale procurement of domestic DRAM by a Chinese internet giant.

For Changxin, this is not just an order, but also a strong credibility endorsement. After all, server scenarios have extremely high requirements for chip stability. Tencent's willingness to integrate Changxin DRAM into its core server supply chain shows that its products have passed very strict verification.

The second card is the "ambiguous" contact with Apple.

In early July 2026, foreign media reported that Apple was in talks with Changxin to procure DRAM chips for products sold in the Chinese market. But later news emerged that since Changxin needs to prioritize supply to domestic manufacturers such as Xiaomi and Alibaba, Apple's procurement plan may face delays.

Citi commented that Apple's recognition of Changxin in itself is enough to elevate Changxin's status in the global memory industry.

But this also highlights the core problem.

Tencent's 20-billion-yuan large order will certainly help Changxin release short-term performance, and prove that domestic DRAM is capable of entering the supply chain of leading clients. But on the other hand, this also means that Changxin's current