2026-11-15 12:06

Introduction: Capital is flowing from chip design into packaging, power delivery, and software integration. On November 12, Xixuan Chuangzhi—jointly initiated by Muxi and Ubtech—announced a $500 million pre-seed round of funding.

November 12, Xixuan Chuangzhi—co-founded by Muxi and Ubtech—announced a $500 million pre-seed funding round. Just two days prior, intelligent NIC vendor Mochuang and silicon carbide power module maker Lipsis separately disclosed new financing rounds. Computing chips, data interconnects, and power delivery components—all key elements in the compute infrastructure—are attracting substantial capital inflows.

One week earlier, Hemo Intelligent reported closing a multi-billion yuan round, while Yicheng Technology and Lingchuan Technology also revealed multi-billion and 1 billion yuan financing respectively. From October 31 to November 6, domestic IC sector disclosures totaled approximately 11.1 billion yuan in funding, accounting for over half of all disclosed deals that week. According to Chanzhong Research Institute, semiconductor venture capital and private equity investment reached ~109.2 billion yuan in Q3 this year, up 91% YoY.

Multiple companies have set their mass production milestones for 2027 or earlier, while some newly funded firms plan to launch their first chip only as late as 2028. Capital is simultaneously supporting scaling of existing products and early-stage investments in unproven architectures and novel processes. What competitive shifts will this funding wave bring to domestic Chinese chip development?

01

Funding Aligned with Divergent Compute Roadmaps

Chanzhong Research Institute reports that VC/PE investment in semiconductors reached ~109.2 billion yuan in Q3, up 91% YoY—significantly outpacing the overall VC/PE market growth of 49.6%. Investment events rose by over 50%. Since October, at least 10 AI chip-related companies have publicly disclosed fundraising, spanning heterogeneous computing, server CPUs, intelligent driving, and embodied intelligence.

On November 4, Hemo Intelligent, a monolithic compute-in-memory chip company, announced a multi-billion yuan Series B round. Funds will support expanding M50 volume deliveries, developing next-gen M6X and 3D compute-in-memory products, and further investing in software toolchains. At its October ecosystem partner summit, the company revealed that M50—produced less than one year ago—has secured over 150 product integrations and serves more than 100 clients.

Hemo’s compute-in-memory technology reduces energy consumption by tightly integrating memory and compute units, minimizing data movement. Its products target local inference on PCs, smart terminals, and edge devices—systems constrained by power, memory, and thermal headroom. Chips must operate within full-system contexts, requiring close collaboration with OEMs during system integration. Existing products demand ongoing co-development, and future generations require sustained investment.

Weina Core is also advancing compute-in-memory architectures but at a different commercial stage. On October 28, the company announced a C2 round exceeding 1 billion yuan, bringing cumulative fundraising past 2 billion yuan within two months. Its first-generation three chips were recently taped out, with customer testing and mass production expected to begin early 2027. New capital will fund product iteration and supply chain preparation, including partnerships with memory suppliers and advanced packaging vendors to enable subsequent manufacturing.

3D integration of memory and compute units requires synchronized coordination across chip design, memory supply, and packaging processes. Weina Core disclosed that edge-side customers are already involved in specification definition and algorithm adaptation, while cloud-side teams are conducting solution evaluations. Despite no large-scale sales yet, customer onboarding and manufacturing readiness are progressing in parallel.

Other firms are entering markets through clearly defined applications. Lingchuan Technology, which disclosed a 1 billion yuan Series B round on November 4, was originally the heterogeneous computing and chip division of Kuaishou. According to investor Qiming Venture Partners, its SL200 AI video chip has sold nearly 100,000 units and is deployed in clients including Kuaishou, Alibaba Cloud, and Baidu Cloud. The next-generation product targets AI video generation using 3D stacking technology, with recent tape-out completed and mass shipment expected in 2027.

The established video processing foundation provides Lingchuan with testbeds and partnership entry points, but generative video demands different computational workloads, requiring re-validation. The company has launched its proprietary software platform, compatible with models such as Tongyi Wanxiang and Hunyuan Video. This round’s funding will support software enhancement and market expansion.

Earlier in November, Moxin disclosed a near 1 billion yuan Series C+ round, with proceeds directed toward sparse inference chips and R&D for next-generation products. Unlike compute-in-memory, which minimizes data movement, sparse computing leverages software-hardware synergy to eliminate redundant computation. The company plans to release the SP200 board card before year-end and update its software stack to support large model deployment and cluster management.

Xixuan Chuangzhi, freshly funded, focuses on robotics. The company aims to integrate perception, decision-making, and motion control computation into a single SoC, centered around a general-purpose GPU. It is also developing a corresponding software toolchain. The first chip remains in architecture design phase, with plans to deliver a developer validation platform to industry partners in 2027 and launch its first SoC and associated solutions in 2028.

Muxi contributes GPU architecture and software expertise; Ubtech brings real-world robot power consumption, interface, and deployment requirements into product definition. From edge inference and video generation to robot control, diverse workloads create openings for new architectures—and demand that chipmakers develop task-specific software and validate full-system integration.

02

From Packaging Lines to Power Delivery Components

High-performance chips increasingly rely on advanced packaging to integrate processors, memory, and other dies. This drives rising investment in supporting infrastructure. Yole Group forecasts that global advanced packaging revenue will surpass traditional packaging for the first time in 2026, with capex approaching $28 billion. Achieving performance and cost advantages requires aligned packaging processes, equipment, and materials—extending funding needs beyond design into manufacturing.

On November 5, Chengdu Industrial Investment Group disclosed that advanced packaging firm Yicheng Technology completed a multi-billion yuan Series C round. This marks the fourth investment by Chengdu Industrial Investment in the company. Yicheng uses large-area square substrates, offering higher area utilization compared to traditional circular wafers.

Chairman Li Chaoliang revealed in late October that the first factory project had a total investment of 5.5 billion yuan, commenced construction in late 2021, and achieved full production line ramp-up only in October 2024. Consumer chip packaging is already in mass production, while 2.5D packaging for AI chips targets large-scale output by 2027. The second factory remains under planning. This funding supports extending existing operations into more complex packaging technologies—not mere replication of current capacity.

Previous funding from Chengdu Industrial Investment covered pilot lines, process validation, and mass production stages. In 2025, convertible bonds were converted into equity, helping reduce the company’s debt-to-equity ratio. Packaging firms often face long lead times before stable revenue, with yield fluctuations and process adjustments potentially delaying deliveries. Continuous investment allows time for debugging, aligning capital recovery with yield ramp-up and customer onboarding.

Per Li Chaoliang’s disclosure, equipment domestication rate at Yicheng’s first factory reaches 70%, and material domestication hits 60%. Domestic equipment and materials suppliers gain procurement opportunities and collaborate with customers to refine processes, gradually building cost competitiveness for new packaging routes.

Bangxin Semiconductor, which disclosed a >600 million yuan Series B+ round early November, built its production base near storage customers. The company supplies etching, de-bonding, and thin-film deposition equipment. Per investor disclosures, over 100 units were delivered by 2025. The Wuhan Huazhong base began operations on October 28, handling assembly, testing, process validation, and delivery. Lead investor Jiangxia Science & Technology Investment not only provided capital but also supplied standardized facilities and coordinated additional funding.

Lipsis, which disclosed a near 100 million yuan Series B round on November 11, is directing funds toward AI data center product development and global market expansion. Originally focused on SiC power modules for electric heavy trucks, energy storage, and industrial power supplies, Lipsis now aims to enter power conversion and protection segments in data centers. Multiple products have been sent for customer sampling, with some entering reliability validation. Full-scale ramp-up awaits completion of validation.

Lipsis is also constructing a Yangzhou factory, expected to go live in April 2027, with an annual capacity of 1.5 million modules post-completion. Factory startup and new customer procurement do not naturally align—funding must cover intermediate phases including debugging, validation, and inventory buildup.

03

From Capacity Lock-in to Joint Investments

TrendForce’s research released at the end of October noted that cloud service providers’ capex continues driving growth in AI server shipments and revenue. ASIC and domestic chip adoption are increasing. Meanwhile, advanced packaging, HBM, and substrate supply remain tight, limiting rack-level scalability. For new chip startups, market expansion offers client acquisition opportunities—but after R&D completion, they must organize supply chains and system integration before delivering to customers.

On October 28, South Korean inference chip firm FuriosaAI announced a $620 million Series D round. Funds will expand production and deployment capabilities and advance next-generation R&D. Its RNGD inference accelerator entered mass production in January this year. The company plans to produce 20,000 accelerators this year and confirmed it has locked in wafer and memory capacity sufficient to double output by 2027.

FuriosaAI is also co-developing third-generation products with partners like Broadcom. Four major software development kit updates were released this year, enhancing model support and deployment features. Post-production, software and next-gen R&D continue—this new funding supports both delivery and future product development.

In July, Samsung SDS integrated RNGD-based compute services into its cloud platform, enabling on-demand subscription without upfront server purchase. This service runs alongside existing GPU offerings. Enterprises can offload suitable inference tasks to the new chip, test cost and performance in real business scenarios, then decide scale. This represents closer alignment to continuous procurement than mere partnership announcements or test results—but production targets don’t equal committed orders.

Compared to FuriosaAI’s commercially available RNGD, UK-based Fractile has yet to launch its first commercial chip but still raised a major round. On November 11, Fractile disclosed a $720 million Series C round led by CapitalG (Google’s venture arm), the UK Sovereign AI Fund, and UK Commercial Bank. The company plans to launch its first commercial chip in 2027. The UK government seeks to strengthen domestic AI hardware capability and reduce reliance on overseas tech—supporting technical roadmaps ahead of commercial delivery.

Domestic firms are also using investments to proactively form collaborations. On November 10, Mochuang disclosed a new funding round, backed by Lihexin Calculation Fund and Bai’ao Software. Proceeds will fund R&D on intelligent NICs and GPU interconnect chips. Lihexin Calculation Fund includes Eastern Calculation Chip, a compute chip player; Bai’ao Software provides firmware for server boot, configuration, and management.

Mochuang’s N30–400G intelligent NIC has passed compatibility tests with multiple GPU platforms. The company is also co-developing high-speed interconnect solutions with domestic GPU vendors. To join compute clusters, chips must interface with GPUs and adapt to server management software. This round links several types of enterprises, with collaboration directly involving interfaces, software, and system validation—not just increasing R&D budgets for individual firms.

Similar industrial cooperation extends into packaging. On November 6, Biwei Storage announced that its subsidiary Guangdong Xinchen Hanqi received a 600 million yuan capital injection. Investors include Zhuhai Biren, wholly owned by Biren Technology. The storage firm maintains majority control, while AI chip firms participate in advanced packaging and testing platform investments—enabling earlier alignment with fabrication and validation needs. Whether equity partnerships translate into stable procurement depends on actual project onboarding.

After securing funding, design firms face growing demand for test tape-outs, prototyping inventory, and packaging validation—preceding large-scale terminal deployment. Manufacturing firms can leverage this to accumulate process know-how and customer relationships. However, R&D-phase procurement cannot directly reflect long-term capacity utilization; subsequent expansion must follow product validation and order momentum. With multiple companies targeting mass production by 2027, this funding wave enables more domestic chip solutions to access customer supply chains. From chip R&D and manufacturing support to system integration, these investments will ultimately determine how much market share firms can capture—depending on their ability to ensure stable delivery and sustain customer procurement.

Source: Semiconductor Industry Horizons