HomeArticle

Silicon Flow has completed a new round of financing exceeding 2 billion yuan

星连资本2026-06-16 15:00
Silicon Flow Completes Over RMB 2 Billion Series B Financing to Build Token Factory MaaS Platform

Recently, SiliconFlow has completed a Series B financing round exceeding 2 billion yuan. This round of financing was jointly invested by industrial capital, top financial institutions, and state-owned assets, including Ctrip Strategic Investment, JinkoSolar Holding, Kingdee International Software Group, Unicom Xinwo, United Innovation and Investment Fund under Unicom Capital, Shengyi Capital (Industrial Capital of Runze Group), Biren Strategic Investment, NIO Capital, SenseTime Strategic Investment, Giant Network, Guotai Junan Innovation Investment, GGV Capital, Huaxi Ruideng, Huakong Fund, China Development Financial, Beigong Investment, Zhongguancun Science City, etc. China Renaissance served as the exclusive financial advisor. So far, the company has received investments from leading enterprises across the entire industrial chain of energy, chips, computing power infrastructure, cloud services, large models, and scenario applications.

In the past year, the company has achieved explosive growth in the enterprise market: By providing efficient MaaS (Model as a Service) through the Token factory model, the daily average Token (token) call volume has reached trillions, serving over 10 million users and 10,000 enterprise customers. The revenue has increased by over 10 times year-on-year, and the monthly revenue in the overseas market has reached millions of US dollars.

The benchmark customers cover leading central state-owned enterprises in core industries such as energy, finance, and transportation, public utilities and national-level scientific research institutes, multiple telecommunications operators and large local intelligent computing centers, cloud service giants, leading large model enterprises, and internet giants, as well as hundreds of leading AI applications and development tools.

Multiple third-party data indicate that SiliconFlow's market influence has entered the first echelon of the industry and has become the most well-known third-party MaaS platform in China:

It ranks firmly in the first echelon in the domestic market. The latest "Global AI Cloud Market Guide" released by Gartner shows that SiliconFlow is one of the few Chinese companies selected as representative vendors. It is also the only startup company selected as a representative vendor in the "China AI Infrastructure Market Guide". The latest "China AI Software Market Semi-Annual Tracker, 2025H2" released by IDC shows that SiliconFlow is the only startup company among the top four in the market share of China's public cloud MaaS. Similarweb data indicates that SiliconFlow's website traffic is comparable to that of leading giants and is dozens of times that of similar startups.

Its global influence is rapidly rising. Data from the well-known Token distribution platform OpenRouter shows that among more than 70 Token suppliers, SiliconFlow's daily Token consumption has ranked first for several consecutive weeks, and key indicators such as the average model cache hit rate and output speed have long been among the top. Data from the well-known AI development platform Dify Marketplace shows that the number of installations of SiliconFlow's plugins has exceeded 550,000, ranking among the top five among more than 100 Token suppliers from around the world.

Token Factory: The Most Certain Business Opportunity in the AI Era

The wave of large models has triggered a productivity revolution as profound as the steam engine and electricity, which will reshape all industries. This industrial revolution has created two types of business opportunities: One type is various applications close to scenarios, which are diverse but face high uncertainty. The other type is the base with extremely high certainty on which all applications rely - regardless of how the upper-layer applications change, they all depend on the continuous supply of underlying capabilities. The Token factory is precisely this base with extremely high certainty.

The core technical elements of the AI base are chips, models, and system software. They do not exist in isolation but are deeply coupled, ultimately condensing into a new product form - the Token factory. It encapsulates computing power, algorithms, and system capabilities to output the atomic unit of "intelligence", the Token, on a large scale, at low cost, and with high reliability.

As the hub for the production and distribution of intelligence, the Token factory will become the underlying infrastructure of the future intelligent society, playing the role of "water, electricity, and coal" in the new era. In the past few decades, infrastructure such as roads, railways, power grids, and communication networks has formed the cornerstone of economic takeoff. In the future, the competitiveness of countries and enterprises will greatly depend on the ability to produce and transmit intelligence.

Every time the capabilities of large models take a step forward, a large category of applications is unlocked, and in turn, the applications increase the consumption of underlying Tokens. IDC predicts that the Token consumption in the Chinese market will reach 40,000 trillion in 2026, an increase of about 20 times compared to 2025.

Facing the explosive market demand, the stable supply of large-scale and cost-effective Tokens has become a key ability affecting the large-scale implementation of AI technology. However, the structural shortage, fragmentation, and low utilization rate of computing power have led to a serious shortage of high-quality Token supply. At the same time, as large models unlock more application scenarios, the high Token cost has become the core pain point restricting the large-scale implementation of enterprises.

SiliconFlow: The Earliest Practitioner of the Token Factory

Since its establishment, SiliconFlow has prospectively proposed the Token factory model, committed to improving Token production efficiency and achieving technology popularization through large-scale production. It has gradually turned its initial goal into reality through a series of milestones:

  • In August 2023, it started the research and development of a large model inference engine, which has become one of the few engines verified in a large-scale production environment in the world.
  • In May 2024, the public cloud MaaS was launched. Dozens of mainstream open-source models were put on the shelves, the link from "bare metal" to standardized Tokens was established, and the strategy of supporting diversified computing power was promoted.
  • In February 2025, the DeepSeek inference service based on Huawei Ascend was launched, achieving the ultra-large-scale Token production service based on domestic chips in the industry. SiliconFlow has now become a Token production line supporting domestic chips.
  • In September 2025, the private MaaS was launched. For state-owned enterprises, central enterprises, and financial institutions that have their own computing power and have extremely high requirements for data compliance, it supports the rapid establishment of exclusive Token factories in their private environments.
  • In April 2026, the new-generation computing power scheduling engine "Elastic GPU" was launched, realizing the elastic expansion and contraction of multi-heterogeneous computing power and supporting customers to efficiently deploy models in a self-service manner.

After three years of in-depth development, SiliconFlow has taken the lead in all aspects of the Token production line and has become a Token factory with global influence.

The World's Leading Token Production Line: Fully Meeting Industrial Needs

The SiliconFlow team has been deeply involved in the field of AI system software for more than a decade. With the craftsmanship spirit of never neglecting the slightest detail and persevering, they have dedicated themselves to technological breakthroughs and product refinement, building a world-leading fully self-developed Token production line.

"Atomic-level" engineering optimization: Maximizing computing power efficiency: The self-developed inference engine integrates leading technologies such as PD separation, KV cache management, expert parallelism, and pipeline parallelism. It supports mainstream models such as DeepSeek, Qwen, GLM, and Kimi and can stably provide commercial-grade services with high throughput and low latency on diverse chips such as NVIDIA, Ascend, MXIC, and Moore Threads.

Heterogeneous computing power management and scheduling: Solving the problem of supply-demand matching: It realizes unified intelligent scheduling and elastic scaling of heterogeneous clusters, supports cross-regional resource scheduling, significantly improves computing power utilization, and achieves extreme cost control, meeting the diversified business and Token mass production needs of enterprise-level customers. SiliconFlow has established strategic cooperation relationships with many domestic and foreign computing power suppliers.

Quick adaptation to over 160 models: Providing full-modal choices: It precipitates the adaptation capabilities of different models into reusable modules, enabling the rapid deployment of the latest open-source models. The number of models adapted by the platform far leads among third-party suppliers, covering full-modal tasks and providing sufficient choices for enterprises to deal with complex scenarios.

Based on the self-developed inference engine and heterogeneous computing power management and scheduling system, SiliconFlow's solution can transform any bare computing power into a standardized Token factory with one click, helping computing power holders comprehensively solve core pain points such as complex heterogeneous computing power adaptation, low utilization rate, and low return on assets, and maximizing the commercial value of computing power resources.

Joint Investment by Industrial Giants: Deep Collaboration in the Token Supply Chain

The investors in this round of financing cover the upstream and downstream of the AI industrial chain. This all-industrial-chain investment lineup not only brings financial support but also brings deep ecological collaboration in scenarios, computing power, models, and the market.

JinkoSolar Holding said: JinkoSolar provides "photovoltaic + energy storage" solutions to help AI intelligent computing centers provide green and low-cost electricity. SiliconFlow integrates diverse computing power through core capabilities such as cross-chip adaptation and inference optimization, improving Token output efficiency from the computing power side. This investment is a strong combination of the two parties at the capital and industrial levels, connecting the entire chain of "electricity - computing power - model - platform - application" to build a green Token factory and promote the sustainable development and large-scale implementation of AI computing and electricity collaborative infrastructure.

Kingdee International Software Group said: SiliconFlow's full-stack closed-loop capabilities in the field of AI infrastructure are highly consistent with our strategic direction of prioritizing AI and building an enterprise AI ecological layout. Kingdee has been deeply involved in enterprise management cloud services for more than 30 years, accumulating rich industry know-how and a large number of high-value business scenarios. We look forward to in-depth collaboration with SiliconFlow around AI infrastructure, model services, enterprise-level application scenarios, and ecological co-construction to help enterprises obtain and apply AI capabilities at a lower cost and higher efficiency, and promote the large-scale implementation and creation of commercial value of enterprise AI from technological innovation.

The investment platform under Unicom Venture Capital said: Unicom Venture Capital uses capital means to serve the main responsibilities and businesses of the group, promoting the innovation of computing network integration and the AI full-fusion plan, and promoting the integrated innovation and ecological construction of AI infrastructure, AI technology, and AI applications. This investment further improves the AI full-chain ecosystem, builds a full-stack capability system of "connection + computing power + AI service", promotes the collaboration of cross-regional heterogeneous computing power, promotes the formation of a domestic, low-cost, and one-stop computing network integration solution, and outputs inclusive AI computing power to government affairs, industry, and developers, helping to cultivate new productive forces in the digital economy.

Runze Group said: SiliconFlow's self-developed inference engine and heterogeneous computing power management and scheduling have a strong industrial synergy effect with the liquid-cooled intelligent computing cluster currently deployed by Runze Group. We are optimistic about its technological barriers and commercialization implementation capabilities, which have connected the model, chip, computing power, tool chain, and terminal application ecosystem, helping to build a cross-ecosystem collaboration system for win-win results among multiple parties. The two parties will work together to create a one-stop MaaS computing power service solution that meets the needs of customers in multiple industries and scenarios.

The investment person in charge of Biren Technology said: SiliconFlow is the first AI infrastructure enterprise to achieve a closed loop of "inference deployment - computing power scheduling - application implementation" on multiple domestic computing power chips. We highly recognize its profound technological accumulation and engineering implementation strength and look forward to in-depth cooperation at the levels of chip adaptation, inference acceleration optimization, and large-scale deployment and implementation of computing power clusters to jointly build a high-performance Token factory and provide more controllable, user-friendly, and efficient domestic AI computing power solutions for more enterprises and developers.

The investment person in charge of Guotai Junan Innovation Investment said: We have long been optimistic about the development potential of the AI Infra track, fully recognize the company's technological strength and development prospects, and are optimistic about the application value of the intermediate scheduling layer in the context of the open-source ecosystem and domestic chips. SiliconFlow has built prominent barriers in the core technology field, with rapid commercial growth and a clear path. Its development layout is highly consistent with the strategy of promoting the implementation of the AI industry.

GGV Capital said: We believe that companies that can truly connect computing power supply and model applications and make inclusive computing power accessible will have great long-term value. Dr. Yuan Jinhui has profound technological accumulation and forward-looking judgment in the fields of distributed systems and heterogeneous computing power. The SiliconFlow team he leads has a solid technical foundation and excellent engineering capabilities and occupies a unique and crucial position in the industrial ecosystem of domestic computing power and large model implementation. They are building a Token factory in the era of domestic computing power. We look forward to the company continuously expanding the boundaries of computing power efficiency and enabling efficient and inclusive intelligent computing power to truly serve all industries.

Zhongguancun Science City said: We chose to invest in SiliconFlow because we value its top-notch technical team, the differentiated computing power base route, and the long-term value of domestic computing power adaptation. Relying on the full-chain innovation ecosystem of the science city, we will fully empower SiliconFlow to implement in diverse scenarios such as government and enterprise, manufacturing, and culture and entertainment.

Conclusion

The explosion of the Token economy is the most certain business opportunity in the AI era. SiliconFlow, with its forward-looking investment and strategic determination, focuses on the research and development of core technologies for Token production, and its product and service quality have been widely recognized by the market. With the joint support of the capital of giants across the entire industrial chain in this round, SiliconFlow will accelerate its progress towards becoming "the world's best Token supply platform" and make its contribution to the popularization of AI technology for humanity.

Recent Updates

SiliconFlow Receives Double Recognition from Gartner

Gaining 31K Stars: This Agent Understands You in Minutes

First Release with Limited Free Use | SiliconFlow Launches Nex-N2-Pro

Guizhou Mobile and SiliconFlow Accelerate the Construction of the "Token Factory"