In April 2026, the most buzzing event in China's semiconductor industry was not the launch of any particular chip, but rather the news that a company, which had been suffering losses for three consecutive years, had just submitted an IPO application for its shares on the Hong Kong Stock Exchange.

Xinyuan Microelectronics officially submitted its listing application to the Hong Kong Stock Exchange on April 1, with a market capitalization of 118.5 billion yuan. On the same day, the A-share semiconductor sector saw a collective rise in prices.

On the other hand, Changxin Technology's 29.5 billion yuan IPO on the STAR Market is in the queue, and Huawei's Ascend 910 series has been directly named in the new regulations of the US Department of Commerce - the scope of sales ban is still expanding.

This is the current state of China's semiconductor industry in 2026: capital is in a frenzy, while blockades are intensifying. Breakthroughs are being made, but bottlenecks also persist.

What level have we reached?

Before discussing specific progress, let's first set the basic context. The semiconductor industry comprises a lengthy industrial chain, which can be roughly divided into three stages: chip design (creating chip circuit diagrams using software), chip manufacturing (transforming the diagrams into actual chips), and packaging and testing (protecting the manufactured chips and testing their functionality). Each stage involves a very high technical threshold.

China's situation in these three links is not the same. Design is relatively strong, manufacturing is moderately strong, and equipment and materials are the weakest areas.

So below, I have used three numbers to illustrate the real situation.

① 55%: Which chip-making equipment can produce chips and which ones still cannot?

To manufacture chips, dozens of equipment are required, such as lithography machines, etching machines, thin film deposition equipment, polishing machines, cleaning machines, and so on. Simply put, the lithography machine is responsible for "drawing", the etching machine for "etching", and the thin film deposition equipment for "plating". Each step is indispensable.

This track has a characteristic: where it can be improved, progress is rapid. Where it cannot be improved, it simply cannot be.

For example, in processes such as degumming and cleaning, the localization rate has long exceeded half. Etching machines are even more impressive. The dielectric etching machines from Microsemi have entered the 7nm and 5nm production lines of global top-tier wafer foundries, with a global market share of about 17%, making them leading players in their respective product categories. Domestic manufacturers also supply products for thin film deposition and chemical mechanical polishing (CMP).

However, in the field of lithography machines, the localization rate is still less than 1%.

The lithography machine is the most core equipment in chip manufacturing. Simply put, its function is to "draw" the chip design onto the silicon wafer. What is the required precision? The error of mechanical movement must be controlled to be less than one ten-thousandth of a human hair.

Let's make an imprecise analogy: imagine a chip enlarged to the size of a building, but the circuit lines on it are many times thinner than a human hair. A bacterium is roughly a few hundred nanometers in size, while a 5nm transistor is even smaller than many bacteria.

The task of a lithography machine is to precisely "draw" these circuits, which are so fine that they are invisible to the naked eye, onto a silicon wafer. Any speck of dust or any vibration can render the entire wafer useless.

Currently, there is only one company in the world that can manufacture the most advanced lithography machines (EUV lithography machines, used for 7nm and below processes), and that is ASML from the Netherlands. ASML's ability to produce these machines relies on the cooperation of hundreds of suppliers worldwide, and it has accumulated decades of experience and expertise. It is definitely not something that can be quickly achieved by simply throwing money at it.

Regarding concepts like "7nm" and "5nm", I'll briefly explain here that these numbers refer to the size of the transistors inside a chip. The smaller the number, the more transistors can be packed into a chip of the same size, resulting in stronger performance.

For example, a fingernail-sized mobile SoC chip may integrate over 10 billion transistors internally, relying on the ever-shrinking size of transistors. The flagship mobile processor in 2026 has already entered the 2nm stage.

So what you see is: the localization rate of equipment with 28nm process technology is already quite impressive. However, when it comes to 7nm, the coverage rate of domestically produced equipment drops to single digits. And when it comes to 5nm, it is almost zero.

The walls of mature manufacturing processes are almost breached, while the ceiling of advanced manufacturing processes remains high.

This is the current situation of domestic equipment: multiple breakthroughs but a bottleneck in lithography machines.

93.5%: The factory is operating at full capacity, but it does not mean that advanced manufacturing processes have been mastered

With the equipment in place, the next step is to build a factory and manufacture chips.

SMIC has just released its 2025 annual report, and there are several figures worth examining closely.

The annual revenue was US$9.327 billion, representing a year-on-year increase of 16.2%. The capacity utilization rate averaged 93.5% throughout the year, marking an 8 percentage point increase compared to the same period last year. The monthly production capacity, converted to 8-inch wafers, amounted to 1.059 million pieces, with 87.6% of the revenue coming from the China region.

What does 93.5% capacity utilization rate mean? It means that SMIC's factories are operating at almost full capacity.

Just think about it, the demand for consumer electronics in 2025 is actually quite sluggish, and global mobile phone shipments are still declining. So why can they still achieve full production?

There are essentially two reasons: firstly, the demand for mature process chips driven by AI has increased; secondly, after the US regulations, domestic customers are willing to shift their orders to domestic foundries.

However, high production capacity does not necessarily mean that advanced manufacturing processes are well executed.

SMIC currently primarily earns its profits from mature processes of 28nm and above. The revenue share from advanced processes of 14nm and below is estimated to be less than 20%. It is expanding its production and making progress, but compared to TSMC, it is probably two to three generations behind in terms of process technology. This is not meant to be derogatory; it is an objective reality in the industry.

In chip manufacturing, improving the manufacturing process cannot be achieved simply by expanding production. It requires lithography machines, EDA software, high-end materials, and a complete ecosystem to support. Each step takes time.

Here's another easy-to-remember comparison: TSMC alone currently accounts for over 70% of the global foundry market share.

③ How many Chinese companies can you name that are capable of designing flagship chips?

After discussing equipment and manufacturing, let's finally talk about chip design.

Chip design requires two things: EDA design software and ARM architecture.

EDA software, to put it simply, is the "drawing software" for chip designers - without it, even the most skilled engineers cannot design chips.

Currently, the global EDA market is monopolized by three companies: Synopsys, Cadence, and Siemens EDA (formerly known as Mentor). These three companies collectively account for over 70% of the global market share, with an even higher market share in China. After Huawei was sanctioned, Synopsys and Cadence temporarily cut off their supply to Huawei.

In terms of architecture, over 90% of the world's mobile phone chips utilize designs licensed from the British company ARM. This includes chips from Apple, Qualcomm, MediaTek, as well as Huawei HiSilicon's once flagship chips, all of which are designed based on the ARM architecture.

So the question arises, how many companies in China are capable of designing advanced chips?

If you count carefully, you can actually count them all.

Huawei HiSilicon is the most typical example. At its peak, HiSilicon's Kirin series chips were on par with Qualcomm's Snapdragon, and the technology level of Apple's A series chips was recognized as top-notch in the industry. This is the result of more than a decade of continuous investment and accumulation, which is difficult for other companies to replicate. After being sanctioned in 2019, HiSilicon had first-class design capabilities but could not mass produce due to the lack of foundries, and once relied on other domestic businesses for support. However, in 2023, Huawei's Mate 60 Pro returned with the Kirin 9000S, indicating that the manufacturing aspect has gradually been resolved.

The Hanguang 800, designed by Alibaba's Pingtouge, is an AI inference chip that garnered attention in the industry upon its release, but it is more targeted at specific AI scenarios.

Cambrian is one of the companies in China that have been developing AI chips for the longest time, with products primarily targeting server-side AI inference. However, its commercialization has always faced challenges.

Suiyuan Technology, Baidu, Haiguang Information, and Biren Technology each have their own expertise - some specialize in AI training, while others focus on high-performance computing. However, overall, they are still in the rapid catch-up phase and have not yet achieved the comprehensive strength of Huawei HiSilicon, which can compete with the world's strongest players.

As for Unicore, it mainly focuses on mobile phone baseband chips and low-end SoCs, and has a certain share in the mid-to-low-end market. However, it is not the same as competing head-on with Qualcomm and MediaTek in flagship phones.

Here's another somewhat counterintuitive fact: before HiSilicon, there was only one enterprise in China capable of designing flagship-level chips. It was only after the sanctions that everyone realized how scarce this capability is.

The good news is that domestic companies such as Huada Jiutian and Xinzhixiang have already been following up on EDA tools beyond 14nm. Although there is still a gap in the most advanced design software, at least it is not a complete blank.

④ What is this battle really about?

After all that has been said, you will notice a pattern: where we can, we make rapid progress, but where we cannot, we simply cannot.

On the surface, it seems to be about technology, equipment, and production capacity. But at the bottom, it's about racing against time.

Chip manufacturing has a "cask effect": the efficiency of the entire production line depends on the weakest link. If the lithography machine is stuck, it doesn't matter how powerful the etching machine is. It cannot be solved by breakthroughs in just one link. It requires dozens of equipment, hundreds of materials, and thousands of processes to reach a certain level simultaneously.

Shen Bo, the president of ASML in China, once said that mastering mature manufacturing processes is of great significance to both China's and the world's semiconductor industries.

This sounds like comforting words, but it also reveals the truth of this industry: over 70% of the global chip demand actually doesn't require the most advanced manufacturing processes. Cars, industrial equipment, IoT chips, and home appliance control chips are the major players. 28nm and 14nm are perfectly sufficient in these fields, and they have lower costs and higher reliability.

From this perspective, things are not black and white. Domestic substitution is advancing, but not all areas can achieve breakthroughs. Gaps also exist, but not all areas need to be closed.

By understanding this basic situation, we can at least be less blind in this wave.