2022 China Graph Computing Technology and Application Development Research Report Overview

With the rapid development of cloud computing, big data, and artificial intelligence, industries are increasingly digitized, leading to more complex business environments and intertwined data relationships.

Traditional relational databases often struggle with the efficiency required for deep mining of complex data relationships.

Graph technology, based on graph theory rather than images, uses nodes and edges to naturally represent data connections, providing a more intuitive abstraction for analysis and computation.

CB Insights China released the "2022 China Graph Computing Technology and Application Development Research Report," analyzing the field from technology, talent, research, and industry perspectives, outlining background, evolution, principles, advantages, ecosystem, and future trends.

Graph databases handle CRUD operations on graph data, while graph computing systems (engines) perform deep analytical processing of graph data.

Graph databases trace back to early tree and property‑graph models; Neo4j, founded in 2007, pioneered native graph storage, followed by CosmosDB, OrientDB, ArangoDB, and others exploring multimodal and distributed architectures.

Early graph computing relied on generic frameworks like MapReduce, which were suboptimal for graph workloads.

In 2010, Google introduced Pregel, a BSP‑based distributed graph engine offering better programming models and synchronization; later, CMU’s Select Lab released GraphLab based on the GAS model, influencing subsequent systems.

Until around 2015‑2016, the market was dominated by overseas vendors, after which Chinese academia and industry began intensifying efforts, increasing market interest.

In 2016, Tsinghua researchers Chen Wanguang and Zhu Xiaowei presented the Gemini distributed graph computing system at OSDI, marking a significant domestic achievement.

Major companies such as Ant Group, Alibaba, and Tencent, along with numerous vertical startups, have entered the graph computing arena, exploring commercial opportunities.

Funding data from CB Insights shows that in the past three years nearly 15 companies received over 20 financing rounds, with Neo4j securing $325 million in a Series F round in June 2021—the largest investment in the database sector.

According to DB‑Engines, graph databases have been the most popular database type since 2013, maintaining a leading position with growing interest.

Academic research on graph computing has surged, with paper counts doubling from 2014 to 2021, led primarily by scholars from the United States and China.

Leading Chinese institutions include Tsinghua University, Peking University, the Chinese Academy of Sciences, and Huazhong University of Science and Technology, contributing the majority of highly cited domestic papers.

The industry ecosystem features numerous participants: internet and public‑cloud providers (e.g., Ant Group, Tencent, AWS, Azure), vertical graph firms (Neo4j, TigerGraph, Chuanglin Tech, Oruo Data), and traditional database vendors (Oracle, IBM).

Downstream users are application developers and system integrators, such as knowledge‑graph providers, who build complete solutions on top of graph databases and engines.

Key end‑user sectors include finance, energy, government, social networks, search engines, and recommendation systems.

In financial risk control, graph technology uncovers hidden relationships to detect credit risk, anti‑money‑laundering, fraud, fund tracing, and potential fraud rings.

In the power sector, graphs enable efficient, real‑time equipment management; in social networks, community‑detection algorithms reveal user connections.

Ant Group applies graph technology to its financial risk‑control platform, constructing a "full‑graph" risk architecture that enhances detection of organized criminal activities and complex risk scenarios.

Beyond traditional multi‑hop queries, Ant incorporates pattern‑recognition and community‑detection algorithms to address more sophisticated cases.

The market remains in an early commercial stage, with challenges such as limited customer awareness of graph benefits, difficulties in database selection, and the need for deeper vendor understanding of specific business scenarios.

Business models vary between open‑source and closed‑source; many projects adopt an OpenCore approach, open‑sourcing core modules while charging for advanced features.

Prominent open‑source graph solutions include Neo4j, ArangoDB, GraphX, GraphScope, Plato, and Ant Group’s TuGraph, which was open‑sourced at the 2022 World AI Conference.

Graph computing offers a new paradigm for modeling complex relationships, driving research into scalable, low‑latency, reliable, and cost‑effective systems.

Researchers are optimizing deployment architectures, parallel processing models, and integrating high‑performance computing techniques.

Graph Neural Networks (GNNs) merge graph computing with machine learning, sparking a surge of AI research that blends symbolic and connectionist approaches.

Long‑term, the ecosystem faces challenges in standardization, talent cultivation, and building comprehensive tooling and application layers to broaden adoption.

Efforts are needed to raise customer understanding, foster academia‑industry collaboration, and develop standards for query languages and benchmarking.

Rapid Rise of Academic Research Over the Past Decade

From 2014 to 2021, the number of graph‑computing papers worldwide doubled, with the United States and China leading the surge.

Standardization of graph computing remains a work in progress, with ongoing efforts to define query languages and testing benchmarks.

Overall, graph technology is gaining attention as a powerful tool for representing and solving complex relational problems across multiple domains.