The domestic chip financing boom is surging in with tremendous momentum.
On November 12, Xixuan Chuangzhi, jointly initiated by MetaX and Ubtech, announced the completion of a 500 million-yuan angel round of financing. Two days earlier, smart network card manufacturer Muchuang and silicon carbide power module manufacturer Lipuzi also successively disclosed new financing. Computing chips, data interconnection, power supply devices, multiple links around the computing power system are attracting capital.
A week earlier, news came out that Houtu Intelligence had completed financing of several billion yuan, while Yicheng Technology and Lingchuan Technology disclosed financing of several billion yuan and 1 billion yuan respectively. From October 31 to November 6, the disclosed financing in China's integrated circuit sector reached about 111 billion yuan, accounting for more than half of the total disclosed financing amount in the week. Statistics from ChinaVenture Research Institute show that in the third quarter of this year, the scale of VC/PE investment in the semiconductor sector was about 1.092 trillion yuan, a year-on-year increase of 91%.
Many enterprises have set the mass production node for new products or new processes around 2027, and some enterprises that have just completed large-scale financing plan to launch their first chip as late as 2028. Capital is not only supporting existing products to expand delivery, but also investing in new architectures and new processes that have not yet generated sales in advance. What kind of competitive changes will this round of financing boom bring to domestic chips?
01
Capital is betting on different computing power routes
According to statistics from ChinaVenture Research Institute, the scale of VC/PE investment in the semiconductor sector in the third quarter of this year is about 1.092 trillion yuan, a year-on-year increase of 91%, significantly higher than the 49.6% growth rate of the entire VC/PE market, and the number of investment events has also increased by more than 50%. Since October, there have been at least 10 AI chip-related enterprises that have publicly disclosed financing, covering directions such as computing-storage integration, server CPU, intelligent driving, and embodied intelligence.
On November 4, news came out that Houtu Intelligence, a computing-storage integrated chip enterprise, had completed a multi-billion-yuan Series B financing. The funds will be used to expand the delivery of M50, develop the new generation M6X and 3D computing-storage integrated products, and continue to invest in the software tool chain. The company disclosed at its ecological partner conference in October that M50 had obtained more than 150 product design wins in less than one year after mass production, serving more than 100 customers.
Houtu's computing-storage integration technology tightly combines storage and computing units to reduce energy consumption caused by data transmission. Its products are mainly oriented to local inference on computers, intelligent terminals and edge devices. These devices are limited by power consumption, memory and heat dissipation space, so chips must run models under the conditions of the complete machine. Existing products need to cooperate with customers' complete machine development, and the next generation of products also requires continuous investment.
MicroNucleus Core is also advancing the computing-storage architecture, but at a different commercialization stage. On October 28, the company announced the completion of a Series C2 financing of over 1 billion yuan, with total financing exceeding 2 billion yuan in two months. All three chips of its first generation have just been taped out recently, and are expected to be sent to customers for testing and start batch supply in early 2027. The new funds will not only be used for product iteration, but also for mass production stocking in the supply chain. The company is cooperating with original storage manufacturers and advanced packaging partners to prepare for subsequent manufacturing.
Three-dimensionally integrating storage and computing units requires coordinating chip design, storage supply and packaging processes at the same time. MicroNucleus Core disclosed that end-side customers have participated in specification definition and algorithm adaptation, while the cloud side is advancing solution evaluation. The first-generation products have not yet formed large-scale sales, and customer onboarding and manufacturing preparation are being carried out in parallel.
Other enterprises cut in from clear application scenarios. Lingchuan Technology, which disclosed 1 billion yuan Series B financing on November 4, was formerly the heterogeneous computing and chip division of Kuaishou. According to Qiming Venture Partners, the investor, its SL200 intelligent video chip has sold nearly 100,000 units and entered customers such as Kuaishou, Alibaba Cloud and Baidu Cloud. The next generation of products turns to AI video generation, adopts 3D stacking technology, and has recently completed the return of the taped-out wafer, which is expected to be shipped on a large scale in 2027.
The customer base brought by video processing provides Lingchuan with access for testing and cooperation, but generative video corresponds to different computing tasks, and new products still need to be re-verified. The company has launched a self-developed software platform that adapts to models such as Tongyi Wanxiang and Hunyuan Video, and this financing will also support software and market expansion.
At the beginning of November, Mochip disclosed nearly 1 billion yuan Series C+ financing, and the funds will continue to be invested in sparse inference chips and pre-research of next-generation products. Different from computing-storage integration which reduces data transmission, sparse computing reduces unnecessary calculation through the cooperation of software and hardware. The company plans to launch the SP200 board card before the end of the year, and update the software stack that supports large model deployment and cluster operation and maintenance.
The newly financed Xixuan Chuangzhi sets its target on robots. The company plans to integrate the computing required for perception, decision-making and motion coordination into a main control chip, take general-purpose GPU as the core computing unit, and develop the software tool chain at the same time. At present, the first chip is still in the architecture design stage, and it plans to provide development and verification platforms to industrial partners in 2027, and launch the first SoC and supporting solutions in 2028.
MetaX provides GPU architecture and software experience, and Ubtech brings the power consumption, interface and deployment requirements in robot tasks into product definition. From end-side inference, video generation to robot control, different loads leave opportunities for new architectures to enter the market, and also require chip manufacturers to develop software and verify complete machines around specific tasks.
02
From packaging production lines to power supply devices
High-computing-power chips increasingly adopt advanced packaging to integrate processors, storage and other chips together, and the investment required for manufacturing support also increases accordingly. Yole Group predicts that global advanced packaging revenue will exceed traditional packaging for the first time in 2026, with related capital expenditure approaching 28 billion US dollars. New chips need corresponding packaging processes, equipment and materials support to gain performance and cost advantages, which also extends financing demand from the design end to the manufacturing link.
On November 5, Chengdu Industrial Investment Group disclosed that Yicheng Technology, an advanced packaging enterprise, has completed several billion yuan Series C financing. This is the fourth time Chengdu Industrial Investment has invested in this company. Yicheng uses large-size square substrates to process chips, which has the opportunity to improve area utilization compared with round wafers.
Chairman Li Chaoliang disclosed at the end of October that the total investment of the company's first factory project is 5.5 billion yuan, which started construction at the end of 2021 and realized the mass production and connection of the production line in October 2024. The existing consumer chip packaging has entered mass production, while the 2.5D packaging for AI chips aims at large-scale mass production in 2027, and the second factory is still under planning. This financing is to support the extension of existing business to more complex packaging processes, rather than simple replication of ready-made production capacity.
Previously, the funds of Chengdu Industrial Investment covered Yicheng's pilot line, process verification and mass production stages. In 2025, it also converted convertible bonds into equity to help the company reduce its asset-liability ratio. Investment in packaging enterprises often precedes stable revenue, and yield fluctuations and process adjustments may delay delivery. Continuous investment gives enterprises time to complete these adjustments, and also closely links the progress of capital recovery with yield ramp-up and customer onboarding.
According to the data disclosed by Li Chaoliang, the localization rate of equipment in Yicheng's first factory reaches 70%, and the localization rate of materials reaches 60%. In addition to obtaining procurement opportunities, local equipment and material manufacturers can also jointly improve processes with customers, so that new packaging routes can gradually form cost competitiveness.
Bangxin Semiconductor, which disclosed the completion of over 600 million yuan Series B+ financing in early November, built its production base near storage customers. The company provides equipment such as etching, photoresist stripping and thin film deposition. According to investors, it has delivered more than 100 units in 2025. The first phase of Wuhan Central China Base was put into operation on October 28, undertaking equipment assembly, testing, process verification and delivery. Jiangxia Science and Technology Investment, the lead investor, not only contributed capital, but also provided standardized factory buildings and coordinated other funds to participate.
Lipuzi, which disclosed nearly 100 million yuan Series B financing on November 11, will mainly invest the funds in R&D of products related to AI data centers and global market promotion. The company's original silicon carbide power modules are used in scenarios such as electric heavy trucks, energy storage and industrial power supplies, and now it hopes to enter the power conversion and protection links of data centers. A number of products have been sent to customers for sampling, and some have entered reliability verification. The related business still needs to wait for the completion of verification before volume expansion.
Lipuzi is also building a Yangzhou factory, which is expected to be put into operation in April 2027, with an annual output of 1.5 million modules after reaching full production. The commissioning of the factory and the procurement of new customers do not naturally synchronize, and financing also needs to cover the intermediate debugging, verification and stocking.
03
From locking production capacity to joint investment
TrendForce pointed out in its research at the end of October that the capital expenditure of cloud service providers continues to drive the growth of AI server shipment volume and output value, and the application of ASIC and Chinese chips is also increasing; at the same time, the supply of advanced packaging, HBM and substrates is tight, and the volume release of full racks is restricted. For new chip enterprises, market expansion provides opportunities to win customers, but after completing R&D, they also need to organize supply and adapt systems before delivering products to customers.
On October 28, South Korean inference chip enterprise FuriosaAI announced the completion of 620 million US dollars Series D financing. The funds will be used to expand production and deployment, and develop next-generation products. Its RNGD inference accelerator has entered mass production in January this year. The company plans to produce 20,000 accelerator cards this year, and said it has locked in wafer manufacturing and memory production capacity that supports more than doubling of output in 2027.
FuriosaAI is also developing third-generation products with partners such as Broadcom. It has released four major updates of the software development toolkit this year, adding model support and deployment functions. After mass production, software and next-generation R&D expenses are still ongoing, and new financing needs to support both delivery and subsequent product development at the same time.
In July, Samsung SDS included RNGD-based computing power services into its cloud platform, allowing customers to subscribe on demand without purchasing servers first. This service runs in parallel with the original GPU service, and enterprises can assign suitable inference tasks to new chips, test cost and performance in real business, and then decide the deployment scale. This is closer to continuous procurement than announcing cooperation and disclosing test results, but production targets are not equal to the orders that have been obtained.
Compared with FuriosaAI's mass-produced RNGD, British company Fractile has not yet launched its first commercial chip, but has also obtained large-scale financing. On November 11, Fractile disclosed the completion of 720 million US dollars Series C financing, with investors including CapitalG under Google, the UK sovereign AI fund and the British Business Bank. The company plans to launch its first commercial chip in 2027. The British government hopes to strengthen local AI hardware capabilities and reduce dependence on overseas technology supply through this, and the support for technology routes precedes commercial delivery.
Domestic enterprises are also organizing cooperation in advance through investment. Muchuang, which disclosed a new round of financing on November 10, received investment from Hesuan Core Fund and Baiao Software. The funds will be invested in R&D of smart network cards and GPU interconnection chips. The initiators of Hesuan Core Fund include computing power chip enterprise Oriental Hesuan, and Baiao Software provides firmware required for server startup, configuration and management.
Muchuang's N30-400G smart network card has completed adaptation with multiple GPUs, and the company is also jointly developing high-speed interconnection solutions with local GPU manufacturers. To enter the computing power cluster, chips need to not only connect to GPUs, but also adapt to the server's management software. This round of investment connects several types of enterprises, and the cooperation directly involves interfaces, software and system verification, rather than just increasing R&D budgets for a single enterprise.
The same industrial cooperation also extends to packaging. On November 6, Biway Storage announced that its subsidiary Guangdong Xincheng Hanqi has received 600 million yuan of capital increase, and investors include Zhuhai Biren, a wholly-owned subsidiary of Biren Technology. Storage enterprises remain holding, and AI chip enterprises participating in the investment of advanced packaging and testing platforms help to connect processing and verification demands earlier. Whether equity cooperation can form stable procurement still depends on the onboarding of specific projects.
After design companies get funds, demands for test tape-out, trial production stocking and packaging verification are expected to increase before the large-scale deployment of terminals, and manufacturing enterprises can also accumulate processes and customers through this. However, procurement in the R&D stage cannot be directly used as the basis for long-term capacity utilization, and subsequent capacity expansion still needs to follow the progress of product verification and orders. Many enterprises have set their mass production targets around 2027, and this round of funds is enabling more domestic solutions to obtain opportunities to enter the customer supply chain. From chip R&D, manufacturing support to system adaptation, how much market share these inputs can eventually convert into depends on whether enterprises can deliver stably and continuously obtain customer procurement.
This article is from the WeChat official account "Semiconductor Industry Review" (ID: ICViews), author: Jun Xi, published with authorization from 36Kr.