Staking a claim in the data center sector, the Shaoxing-based unicorn has raised 2 billion yuan in financing.

The AI wave is unfolding not only in large models and robots, but also across high-voltage power lines spanning China and data centers in the remote western regions.

It is exclusively learned from ChinaVenture that Corelink Dynamics (hereinafter referred to as "the Company") has announced the completion of a Series C financing of nearly RMB 2 billion. This round of financing is led by China Mobile Chain Length Fund, with participation from investors including a state-owned power grid central enterprise, Inspur Information, Vision Capital, Sinopec Hydrogen Energy Fund, Northern Huachuang Novus Capital, Jiangfeng Electronics, Changzhou Xingyu, Cathay Capital, Puchao Capital, CCB Equity, CITIC Goldstone, CITIC Securities, Galaxy Yuanhui and other investors.

The consistent financing philosophy of Corelink Dynamics is "gather people around a shared mission". This round, the new "mission" that the company aims to carry out is to cut into AI infrastructure from the new energy vehicle track, and provide power solutions for the latter.

The change in the investor lineup reveals this intention. Three years ago, when Corelink Dynamics was founded, all founding shareholders were leading automotive and new energy industry players; this round, the shareholder lineup has expanded to the fields of communication, computing power, power grid and other sectors, and the entry of multiple industrial capitals in the AI infrastructure field is particularly notable.

Founded in 2023, Corelink Dynamics is a one-stop system solution provider for compound semiconductor manufacturing and module packaging, focusing on the design, R&D, production and sales of compound semiconductors such as silicon carbide (SiC) and gallium nitride (GaN) as well as related module packaging. According to the 2025 report from Yole, the company's silicon carbide business ranks first in China and top five globally.

New energy vehicles and AI data centers seem to be far apart, but silicon carbide is a critical component in high-voltage power systems for both, which gives Corelink Dynamics the foundation for capability migration. The difficulties are also obvious: in addition to launching adapted products as soon as possible, the application scenarios, technical architectures and industry players are all completely different. This also explains why the company needs to gather a group of new partners in this round of financing.

Nearly RMB 2 billion is a considerable large-scale financing. It is understood that this round of financing was not publicly released on a large scale, and only communicated with some investors in a targeted manner, but the news spread among industrial institutions one after another, and finally received more than RMB 4 billion of investment intentions. After the completion of this round of financing, the valuation of Corelink Dynamics exceeds 2 billion US dollars.

On the surface, this is another investment stake grab where institutions compete for quota. But "unavailable investment quota" is far from enough to explain this round of financing.

The investor list is essentially a list of "allies": most of the participants have clear business collaboration directions, and some cooperations have entered the stages of co-development of chips, product sampling, industrial investment or project incubation. Equity investment is not the starting point of cooperation, but more like the result of deepened cooperation between the two sides.

This is not the first time that Corelink Dynamics has "gathered people around a shared mission". When it was founded in 2023, Corelink Dynamics gathered automotive and new energy industry players including SAIC, XPeng, Bosch, CATL, Sungrow Power as founding shareholders. For details, please refer to "40 Billion Chip Giant Built a 'Social Circle' in Shaoxing".

The previous rounds of financing of Corelink Dynamics can be called the "comfort zone rounds" — most of the participants came from the familiar automotive and new energy industries. This round is different, the company needs to bring its existing product and manufacturing capabilities to make a new group of friends in AI infrastructure.

Open up the second growth curve in the upward period

The AI boom has directly led to a surge in computing power, and the number and scale of data centers have increased rapidly accordingly. ABI Research predicts that by 2030, the number of data centers put into operation worldwide will exceed 8,400.

Go further down the line, what ultimately restricts computing power? According to Jensen Huang's proposed "five-layer AI cake", the bottom layer is not the model, nor the chip, but energy. Sam Altman has also repeatedly regarded energy as a key condition for the continued expansion of AI.

The main players in the AI race are China and the United States. Tech giants in both countries are investing tens of billions of dollars to build data centers, but the problems they face are not exactly the same.

The constraint in the United States is "there are data centers, but no available power access". The expansion speed of large model companies and cloud manufacturers far exceeds the upgrading speed of the power grid, and data centers have begun to "chase power".

China has a more unified power grid system and stronger infrastructure construction capacity, and the practical problem is: the power is there, how can the computing power be matched to it.

China's computing power demand is mainly concentrated in the eastern region, while energy resources such as wind and solar power are more distributed in the western region. In the past, it was "power transmission from west to east", and now "computing power transmission from east to west" has further emerged. The key behind this is not just to increase power generation, but computing-power-power coordination — to plan computing power hubs, power supplies, power grids and energy storage as a whole.

In 2026, "computing-power-power coordination" was written into the government work report for the first time, and included in the outline of the 15th Five-Year Plan. The coordination between computing power and power has risen from market discussion to a national-level infrastructure project.

The above constitutes the premise of Corelink Dynamics' "gather people around a shared mission" at the "people alignment" level.

Another change is taking place inside the data center. Data centers are moving from the traditional tens of megawatts level to hundreds of megawatts or even gigawatt level. It is reported that Deepseek plans to build a 1GW-level data center in Ulanqab.

With the expansion of data center scale, the power of a single cabinet rises sharply, and the traditional power supply architecture begins to face a series of problems: losses caused by multi-level power conversion, more copper materials are needed to carry large current, power supply and heat dissipation equipment take up space in the machine room, and the rapid fluctuation of GPU load puts forward higher requirements for the stability of the power grid, and so on.

Therefore, the clear direction of the next-generation AI data center is to reduce power conversion levels and increase power supply voltage.

At present, the main technical routes are promoted in layers: the faster-to-implement one is 800V high-voltage DC power supply, namely HVDC; the medium and long-term direction is solid-state transformer SST, which directly converts medium-voltage AC to high-voltage DC. The two transform different levels of the power supply path, with the goal of shortening the power supply link, reducing loss, supporting higher-power computing power clusters in a limited space, and realizing cost reduction and efficiency increase.

This change brings new opportunities for power semiconductors. Compared with traditional silicon-based devices, silicon carbide can withstand higher voltage, temperature and power, while gallium nitride is suitable for high-frequency, high-power-density conversion scenarios. The two types of compound semiconductors are exactly applied in SST, HVDC, UPS and server power supplies respectively.

Among them, silicon carbide power semiconductor is regarded as one of the links with the highest value and the highest barrier in solid-state transformers. According to the data released by Infineon and solid-state transformer company DG Matrix, compared with traditional transformers, the volume of solid-state transformers can be reduced to 1/14 of the former, and the weight is reduced to about 1/40.

A few months ago, Nvidia just included 7 SST manufacturers including DG Matrix into the Nvidia MGX ecosystem, becoming its solid-state transformer suppliers for the 800V DC power supply architecture. The change of data center power supply architecture has moved from concept to products and supply chain.

This is exactly the "mission" that Corelink Dynamics has found: to provide power solutions composed of silicon carbide, gallium nitride and other devices around AI infrastructure.

It is worth noting that this second growth curve is a new direction that Corelink Dynamics actively opened up during its upward period. In the first half of 2026, Corelink Dynamics achieved a revenue of 1.182 billion yuan, a year-on-year increase of about 106%; the annual revenue is expected to exceed 3 billion yuan, and the automotive business is still in a period of rapid growth.

Corelink Dynamics chose to "repair the roof on a sunny day" during the upward period, and actively extended the manufacturing capabilities of automotive silicon carbide to AI power supply. The judgment is mainly based on three conditions: First, AI infrastructure itself is a sufficiently large incremental market. According to Yuan Feng's judgment, AI infrastructure may be one of the few 10-trillion-yuan markets after real estate and automobiles, in which power-related investment may account for about 10%.

Although the automotive market is large enough, it already has a mature industrial foundation, and the annual change is more of growth based on stock; AI infrastructure is different. Two years ago, there was no such large-scale demand, but now it has emerged rapidly in a short period of time.

Macro prediction is one aspect. According to previous media reports citing people familiar with the matter, China plans to invest about 2 trillion yuan in the next five years to build a national computing power network, and state-owned operators such as China Mobile and China Telecom may undertake the main operation work.

AI infrastructure is undergoing drastic technological changes and industrial chain restructuring, providing entry points for new entrants. The opportunities brought by new energy vehicles once came from the simultaneous changes in the power system, electronic architecture and supply chain driven by electrification and intelligence. The transition of data centers from traditional low-voltage power supply to HVDC and SST will also change product definition and supplier relationships.

Similar to the opportunities of new energy vehicles, the last wave of opportunities came from the "four modernizations" that drove simultaneous changes in the power system, electronic architecture and supply chain. The transition of data centers from traditional low-voltage power supply to HVDC and SST will also change product definition and supplier relationships. What Corelink Dynamics hopes to seize is exactly the window when the supply chain is rearranged.

Third, Corelink Dynamics' entry into AI power supply is not to create a completely unfamiliar capability from scratch. New energy vehicles and data centers are different scenarios, but the underlying layer needs to solve the same problem: how to convert and use electric energy more efficiently.

Three years ago, Corelink Dynamics built silicon carbide chip design, wafer manufacturing, module packaging and mass production capabilities around new energy vehicles. The automotive industry mainly uses 750V and 1200V products, and AI power supply needs to extend devices to higher voltages, but the underlying layer still revolves around power semiconductors.

At present, Corelink Dynamics has laid out high-voltage silicon carbide products for AI data centers, and some products have entered the sampling and verification stage.

Therefore, what industrial players really recognize is not a certain automotive chip of Corelink Dynamics, but the company's comprehensive capabilities in designing, manufacturing and mass-producing high-performance power devices.

What can be migrated are chip design, process manufacturing, module packaging, reliability verification and mass production capabilities; what cannot be directly copied are customer relationships, application experience and product definitions.

The automotive business has proved that Corelink Dynamics can produce chips and complete large-scale delivery; the AI business requires it to re-understand the power supply architecture of data centers, re-recognize power grids, cloud manufacturers, server companies and power enterprises, and co-define products with them.

Replicate the model of "gather people around a shared mission"

The new round of "making friends" starts from top-down sorting. Corelink Dynamics first sorted out the key links of the industrial chain, and the new partners of "computing-power-power coordination" cover computing power, communication, power grid, new energy and so on.

The most upstream is the power grid, the starting point of power supply; China Mobile is not only an operator, but also a builder of cloud and computing power infrastructure; Vision Capital can connect Alibaba Cloud and its AI infrastructure ecological resources, and Alibaba is also one of the domestic enterprises that have studied the power supply architecture of AI data centers from the system level earlier; Inspur Information is a giant in servers and AI infrastructure; Northern Huachuang and Jiangfeng Electronics are located in the equipment and material links of the semiconductor industrial chain.

This round of financing does not lack financial investment institutions, but it presents a very distinct industrial color as a whole. Even for financial institutions, what Corelink Dynamics values is not just capital, but the industrial resources that can be connected behind it. For example, Cathay Capital connects with the European energy industry behind it, and is a potential partner for future business globalization; CCB Equity and CITIC Securities correspond to resources at the financing and capital market levels.

So far, the key nodes of the AI power supply chain have been basically covered. It is understood that Corelink Dynamics has carried out substantial business cooperation with some of the above companies at the levels of chip co-development, industrial investment, project incubation and so on. At present, Corelink Dynamics has covered silicon carbide products in the full voltage range from 650V to 3300V in the AIDC field, among which 3300V SiC MOSFET has been sent to core customers for sampling and verification, and products of 6.5kV and above are in the R&D stage.

These cooperations were not discussed after financing, but business recognition comes first, and financing comes later. Before the equity investment, Corelink Dynamics has carried out business communication, technical exchange or product sampling with them. Therefore, the 2 billion yuan financing is not the starting point, but the result, which is just the endorsement that strengthens the cooperative relationship.

Using CVC as a path, or establishing a joint venture to achieve a certain business goal together is not a new thing, but several practices of Corelink Dynamics' "gather people around a shared mission" can hardly be regarded as typical CVC cases, with distinct industrial characteristics and personal experience colors.

Yuan Feng, the operator of Corelink Dynamics' financing, has a distinct CVC background. He joined Corelink Integration in 2023, and was previously the CEO of GAC Capital. He has long focused on the vehicle business, and solved problems such as automotive chip shortage, supply chain stability and new technology layout through investment.

Three years ago, the "mission" that Corelink Dynamics wanted to carry out was to promote domestic silicon carbide into new energy vehicles. When it was founded in 2023, the company gathered automotive and new energy industry players including SAIC, XPeng, Bosch, CATL and Sungrow Power.

At that time, China's new energy vehicles were growing rapidly, car companies had just experienced chip shortage, and 800V high-voltage platforms began to popularize, but the silicon carbide market was still mainly controlled by overseas enterprises. Relying only on traditional procurement relationships, it was difficult to solve the problems of product iteration and supply security. Therefore, potential customers entered the product definition and R&D links in advance, the Corelink system was responsible for chip manufacturing and module packaging, and then the relationship between the two sides was further fixed through investment.

The joint development of hybrid silicon carbide modules by XPeng and Corelink is a sample of this model. The two sides combine silicon-based IGBT and SiC MOSFET to re-find the balance between efficiency and cost, and the product was later installed in XPeng X9. This is no longer a simple substitution according to overseas standard products, but a vehicle company and a chip enterprise jointly define a new product around a specific car model.

Three years later, the company is expected to achieve revenue of more than 3 billion yuan in 2026. According to statistics from NE Times, in the first half of this year, Corelink Dynamics' SiC power module ranked first in China with a 16.9% domestic installation share.

Now Corelink Dynamics is trying to replicate this experience to AI infrastructure. The reason for doing this again is that AI adapts to the logic of the supply chain industry — every additional user and every inference consumes more Tokens, and adds more GPUs, optical modules, data centers and power. The expansion of AI companies no longer only depends on users and models, but also on whether they can obtain sufficient chips, energy and infrastructure.

The United States has taken the lead in giving samples. The main players of computing power centers are large manufacturers, such as AI giants including Nvidia, Microsoft, Amazon and OpenAI are actively making upstream layouts along the AI supply chain, and locking in chips, advanced packaging, data centers and power resources through investment.

The logic in the Chinese market is the same, and the player portrait is even more diverse. Power grids, operators, cloud manufacturers, servers and chip enterprises are all involved, each with part of the capabilities, but no one can independently complete the whole construction from power generation to GPU power supply, which brings the possibility of alliances just like the new energy vehicle industry.

The so-called "replication" is to repeat a set of methods to organize the industry: first find a major event in the middle of technological change, then organize the application side, system side, chip side and manufacturing side to participate from the product definition stage together.

From this perspective, Corelink Dynamics is far more than a simple invested enterprise, it is a connector of the Corelink system, and itself acts as a bond for industrial collaboration. This method has very distinct CVC characteristics, but it is not exactly the same as traditional supply chain investment.

In the past, supply chain investment usually means that an enterprise finds that it lacks a certain product or capability, and then invests in a corresponding supplier. The "gather people around a shared