Author: Junxi
On November 12, XILECT (Xixuan Chuangzhi), jointly initiated by MetaX and UBTECH, announced the completion of a CNY 500 million angel round. Two days prior, smart NIC manufacturer Maxio and SiC power module maker Leapers Semiconductor had also disclosed new financing. Computing chips, data interconnects, and power supply components—multiple segments surrounding the computing system—are attracting capital.
A week earlier, Houmo Intelligence was reported to have completed a multi-billion yuan financing round, while ECHINT Technology and Transtreams disclosed financings of several billion yuan and CNY 1 billion, respectively. From October 31 to November 6, disclosed financing in China's integrated circuit sector totaled approximately CNY 11.1 billion, accounting for more than half of all disclosed financing amounts that week. Statistics from Venture Capital Research Institute show that in Q3 this year, VC/PE investment in the semiconductor sector reached approximately CNY 109.2 billion, a year-on-year increase of 91%.
Multiple companies have set mass production milestones for new products or processes around 2027, while others that have just completed large financings do not plan to launch their first chip until 2028. Capital is both supporting existing products to scale deliveries and investing ahead in new architectures and processes that have not yet generated sales. What competitive changes will this wave of financing bring to domestic chips?
01. Capital Bets on Different Computing Routes
According to Venture Capital Research Institute statistics, VC/PE investment in the semiconductor sector in Q3 this year reached approximately CNY 109.2 billion, up 91% year-on-year, significantly higher than the 49.6% growth rate of the overall VC/PE market, with the number of investment events also increasing by over 50%. Since October, at least 10 AI chip-related companies have publicly disclosed financing, covering directions such as compute-in-memory, server CPUs, intelligent driving, and embodied intelligence.
On November 4, compute-in-memory chip company Houmo Intelligence was reported to have completed a Series B financing of several billion yuan. The funds will be used to expand M50 deliveries, develop the next-generation M6X and 3D compute-in-memory products, and continue investing in software toolchains. At its ecosystem partner conference in October, the company disclosed that less than a year after M50 mass production began, it secured over 150 product design wins and served more than 100 customers.
Houmo's compute-in-memory technology tightly integrates storage and computing units to reduce energy consumption caused by data movement. Its products primarily target local inference on computers, smart terminals, and edge devices. These devices are constrained by power, memory, and thermal dissipation limits, requiring chips to run models under whole-system conditions. Existing products must align with customer system development, and next-generation products require continuous investment.
Weina Hexin is also advancing compute-in-memory architectures but is at a different commercialization stage. On October 28, the company announced the completion of a Series C2 financing exceeding CNY 1 billion, bringing its cumulative financing to over CNY 2 billion in two months. Its three first-generation chips were recently fully taped out, with plans to deliver them to customers for testing and initiate mass supply in early 2027. New funds are being used not only for product iteration but also for supply chain mass production inventory preparation. The company is cooperating with memory manufacturers and advanced packaging partners to prepare for subsequent manufacturing.
Sterically integrating storage and computing units requires coordinating chip design, memory supply, and packaging processes simultaneously. Weina Hexin disclosed that end-side customers have participated in specification definition and algorithm adaptation, while cloud-side solution evaluations are ongoing. First-generation products have not yet achieved scale sales, but customer introduction and manufacturing preparation are proceeding in parallel.
Other companies are entering through clear applications. Transtreams, which disclosed a CNY 1 billion Series B financing on November 4, originated from Kuaishou's heterogeneous computing and chip business unit. According to investor Qiming Venture Partners, 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 product shifts to AI video generation, utilizing 3D stacking technology. It recently completed tape-out and is expected to achieve scale shipments in 2027.
The customer base from video processing provides Transtreams with entry points for testing and cooperation, but generative video corresponds to different computing tasks, requiring re-validation for new products. The company has launched its proprietary software platform, adapting to models such as Tongyi Wanxiang and Hunyuan Video. This financing will also support software and market expansion.
In early November, Moffett AI disclosed a Series C+ financing of nearly CNY 1 billion, continuing to invest in sparse inference chips and pre-research for next-generation products. Unlike compute-in-memory which reduces data movement, sparse calculation reduces unnecessary computations through hardware-software coordination. The company plans to launch SP200 cards by the end of the year and update its software stack to support large model deployment and cluster operations.
XILECT, which just secured financing, targets robotics. The company plans to integrate the computing required for perception, decision-making, and motion coordination into a single main control chip, using a general-purpose GPU as the core computing unit while developing software toolchains. Currently, the first chip is still in the architecture design phase, with plans to provide a development validation platform to industry partners in 2027 and launch the first SoC and accompanying solutions in 2028.
MetaX provides GPU architecture and software experience, while UBTECH brings power, interface, and deployment requirements from robot tasks into product definition. From end-side inference and video generation to robot control, different workloads leave opportunities for new architectures to enter the market, while also requiring chip vendors to develop software and validate systems around specific tasks.
02. From Packaging Lines to Power Components
High-computing chips increasingly adopt advanced packaging to integrate processors, memory, and other chips together, leading to increased investment in manufacturing support. Yole Group predicts that global advanced packaging revenue will exceed traditional packaging for the first time in 2026, with related capital expenditures approaching USD 28 billion. To gain performance and cost advantages, new chips require corresponding packaging processes, equipment, and material support, extending financing needs from the design side to the manufacturing segment.
On November 5, Chengdu Industrial Investment Group disclosed that advanced packaging company ECHINT Technology had completed a Series C financing of several billion yuan. This marks Chengdu Industrial Investment's fourth investment in the company. ECHINT uses large-size square substrates for chip processing, offering potential area utilization improvements compared to circular wafers.
Chairman Li Chaoliang disclosed in late October that the total investment in the company's first factory project was CNY 5.5 billion. Construction began at the end of 2021, and production line ramp-up was achieved in October 2024. Packaging for existing consumer chips has entered mass production, while 2.5D packaging for AI chips targets large-scale mass production in 2027, with the second factory still in planning. This financing supports the extension of existing businesses to more complex packaging processes, rather than simple replication of existing capacity.
Previously, Chengdu Industrial Investment's funds covered ECHINT's pilot lines, process verification, and mass production stages. In 2025, it converted convertible bonds into equity to help the company lower its debt-to-asset ratio. Investments in packaging companies often precede stable revenue, and yield fluctuations and process adjustments may delay deliveries. Continuous investment allows companies time to complete these调试 (debugging/tuning) phases, linking capital recovery progress closely with yield ramp-up and customer introduction.
According to data disclosed by Li Chaoliang, the localization rate of equipment in ECHINT's first factory is 70%, and the localization rate of materials is 60%. Local equipment and material suppliers, in addition to gaining procurement opportunities, can also improve processes jointly with customers, helping new packaging routes gradually form cost competitiveness.
Bangxin Semiconductor, which disclosed the completion of a Series B+ financing exceeding CNY 600 million in early November, built its production base near memory customers. The company provides etching, ashing, and thin-film deposition equipment. According to investor disclosures, over 100 units were delivered in 2025. The first phase of the Wuhan Central China base commenced production on October 28, handling equipment assembly, testing, process verification, and delivery. Lead investor Jiangxia Ke Tou provided funding along with standardized workshops and coordinated other capital participation.
Leapers Semiconductor, which disclosed a Series B financing of nearly CNY 100 million on November 11, is directing funds primarily toward R&D for AI data center products and global market promotion. The company's existing SiC power modules are used in electric heavy trucks, energy storage, and industrial power supplies. Now, it aims to enter the power conversion and protection segments of data centers. Several products have been sent to customers for sampling, with some entering reliability verification. Related businesses still need to wait for verification completion before scaling up.
Leapers is simultaneously building a Yangzhou factory, expected to commence production in April 2027, with an annual output of 1.5 million modules upon full capacity. Factory commissioning and new customer procurement are not naturally synchronized; financing must cover intermediate debugging, verification, and inventory preparation.
03. From Securing Capacity to Joint Investment
In research released in late October, TrendForce noted that cloud service provider capital expenditures continue to drive growth in AI server shipments and value, with increasing applications of ASICs and Chinese chips. Meanwhile, supplies of advanced packaging, HBM, and substrates remain tight, limiting rack-scale volume expansion. For new chip companies, market expansion offers opportunities to win customers, but after completing R&D, they must organize supply and adapt systems to deliver products to customers.
On October 28, South Korean inference chip company FuriosaAI announced the completion of a USD 620 million Series D financing. Funds will be used to expand production and deployment, and to develop next-generation products. Its RNGD inference accelerator entered mass production in January this year. The company plans to produce 20,000 accelerator cards this year and stated it has secured wafer manufacturing and memory capacity to support doubling production in 2027.
FuriosaAI is also developing third-generation products with partners like Broadcom. It has released four major software development kit updates this year, adding model support and deployment features. After mass production, software and next-generation R&D expenditures continue, requiring new financing to support both delivery and subsequent product development.
Samsung SDS incorporated RNGD-based computing services into its cloud platform in July, allowing customers to subscribe on demand without purchasing servers upfront. This service runs parallel to existing GPU services, enabling enterprises to assign suitable inference tasks to new chips, test costs and performance in real business scenarios, and then determine deployment scale. This is closer to sustained procurement than announcing partnerships or disclosing test results, though production targets do not equate to secured orders.
Compared to FuriosaAI's mass-produced RNGD, UK-based Fractile has not yet launched its first commercial chip but has secured large financing. On November 11, Fractile disclosed the completion of a USD 720 million 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 UK government hopes to strengthen domestic AI hardware capabilities and reduce dependence on overseas technology supplies through this, supporting the technical route ahead of commercial delivery.
Domestic companies are also organizing cooperation in advance through investment. Maxio, which disclosed a new financing round on November 10, received investments from Lihe Suanxin Fund and Bai'ao Software. Funds will be directed toward R&D for smart NICs and GPU interconnect chips. Initiators of the Lihe Suanxin Fund include computing chip company DF Compute, while Bai'ao Software provides firmware required for server booting, configuration, and management.
Maxio's N30-400G smart NIC has completed adaptations with multiple GPUs, and the company is jointly developing high-speed interconnect solutions with local GPU manufacturers. To enter computing clusters, chips must interface with GPUs and adapt to server management software. This round of investment links several types of enterprises, involving direct cooperation on interfaces, software, and system validation, rather than merely increasing R&D budgets for a single company.
Similar industrial cooperation extends to packaging. On November 6, BIWIN Storage Technology announced that its subsidiary Guangdong Xincheng Hanqi had received CNY 600 million in capital increases. Investors include Zhuhai Biren, a wholly-owned subsidiary of Biren Technology. The storage company maintains controlling interest, while the AI chip company participates in investing in the advanced packaging and testing platform, facilitating earlier alignment of processing and verification needs. Whether equity cooperation leads to stable procurement depends on specific project introductions.
After design companies secure funds, demands for testing, tape-outs, trial production inventory, and packaging verification are expected to increase ahead of terminal-scale deployments, allowing manufacturing companies to accumulate process expertise and customers. However, procurement during the R&D phase cannot directly serve as the basis for long-term capacity utilization rates; subsequent expansion must follow product verification and order progress. With multiple companies setting mass production targets around 2027, this wave of capital is giving more domestic solutions opportunities to enter customer supply chains. From chip R&D and manufacturing support to system adaptation, how much market share these investments ultimately convert into depends on whether companies can deliver stably and sustain customer procurement.