Weimo Liu can tell you how many seconds it takes to run a 10-hop query across half a billion edges. The answer is 2.26. For the co-founder and CEO of PuppyGraph, that number is the wedge into a much larger argument: that the next layer of enterprise AI infrastructure will be built on graph queries, and that those queries should run directly on the data you already have.
PuppyGraph is a graph query engine. Its core proposition is technical and specific. It connects to an enterprise’s existing data warehouse or lake,a Databricks Delta Lake, an Apache Doris cluster,and allows AI agents or analysts to query relationships across tables as if they were a native graph, all without moving or copying the underlying data [Perplexity Sonar Pro Brief]. The company claims it can map relationships across roughly 3,000 enterprise tables and return results in subseconds [Perplexity Sonar Pro Brief]. In November 2026, it announced the general availability of PuppyGraph 1.0 alongside a $10 million Series A round led by Particle Future [Axios Pro, November 2026].
The Wedge: No ETL, No New Database
The traditional path to graph analytics has been a heavy lift. Teams identify a use case,fraud detection, supply chain mapping, security analysis,extract the relevant data, transform it, and load it into a dedicated graph database. This ETL process creates latency, cost, and a separate system to maintain. PuppyGraph’s bet is that this step is now a bottleneck, especially for AI agents that need real-time, multi-hop reasoning across a company’s entire data estate.
The engine sits as a query layer atop existing storage. It uses what the company calls a "zero-ETL" architecture to dynamically infer and traverse relationships between tables, supporting what it reports as 5 to 10 petabytes in production today [PuppyGraph]. The commercial pitch is speed and simplicity: deploy in 10 minutes, point to your data, and start querying [PuppyGraph]. For buyers in security, data platform, and AI teams, the promise is to unlock graph analytics without the operational overhead of a second database.
The Team and the Traction
The founding team carries pedigree from some of the most demanding data infrastructure environments. Liu holds a computer-science Ph.D. and worked on Google’s F1 database team before a stint at graph database company TigerGraph [Perplexity Sonar Pro Brief]. Co-founder Danfeng Xu spent nine years on LinkedIn’s infrastructure team [Perplexity Sonar Pro Brief]. The third co-founder, Lei Huang, is a three-time Google Code Jam world finalist [Perplexity Sonar Pro Brief]. It is a team built to optimize petabyte-scale queries.
That technical credibility appears to have opened enterprise doors quickly. The company launched in March 2024 and announced a $5 million seed round led by defy.vc that November, noting production use at customers including Coinbase and Clarivate [Business Wire, November 2024]. By the Series A in November 2026, the customer roster had expanded to include names like AMD, Palo Alto Networks, Datadog, Netskope, and Trend Micro [Perplexity Sonar Pro Brief]. The company also reported that annual recurring revenue grew sevenfold over the prior year [Perplexity Sonar Pro Brief].
| Customer | Industry | Use Case (Inferred) |
|---|---|---|
| AMD | Semiconductors | Supply chain, product lifecycle analysis |
| Palo Alto Networks | Cybersecurity | Threat intelligence, entity relationship mapping |
| Coinbase | Financial Services | Fraud detection, transaction network analysis |
| Datadog | Observability | Service dependency mapping, AI ops |
| Baker Hughes | Industrial | Asset performance, maintenance networks |
| Table: A selection of named PuppyGraph production customers [Perplexity Sonar Pro Brief]. |
Where the Market Is Moving
The Series A was not an isolated bet. Particle Future led the round, with seed investor defy.vc participating, bringing total disclosed funding to $15 million [Axios Pro, November 2026]. The financing coincides with a broader push to make data lakes more interactive and intelligent. PuppyGraph has positioned itself at this intersection, becoming a design partner for the Databricks Unity Catalog to enable graph analytics directly on Delta Lake [VMblog, June 2025]. It also lists partnerships with streaming platforms like Confluent and Red Panda for real-time analytics [Insights from Analytics].
The product evolution is squarely aimed at the AI agent stack. The 1.0 release added an AI-assisted graph schema builder and Model Context Protocol (MCP) integrations, tools designed to help developers and agents understand and query complex data relationships more easily [InfoWorld]. Liu presented on "Context Trace/Graph: Observability for AI Agents" at Data Day Texas in 2026, framing the product as critical infrastructure for understanding agent behavior [Data Day Texas].
An Honest Counterfactual
The bet is not without its risks. The market for graph technologies, while growing, has been dominated by native graph databases like Neo4j and TigerGraph. PuppyGraph’s differentiation,avoiding ETL,is also its constraint; it is inherently tied to the performance and capabilities of the underlying data platform. If a warehouse introduces latency, the graph query layer will feel it.
Furthermore, while the customer list is impressive, the depth of deployment and the size of the average contract are not public. A sevenfold increase in ARR is a strong growth signal, but the base from which it grew is unknown. The company must now prove it can move from a point solution for specific analytics teams to a standardized layer adopted broadly across the enterprise.
PuppyGraph’s most plausible answer is performance and partnerships. By focusing on subsecond queries across billions of edges and embedding itself into platforms like Databricks, it aims to become the path of least resistance. The question for the next twelve months is whether that technical wedge is enough to create a durable commercial moat in a stack that is being rapidly redefined by AI.
The $10 million from Particle Future and defy.vc will fund engineering and enterprise go-to-market expansion [Axios Pro, November 2026]. For investors, the calculus is clear: they are betting that the need for real-time, multi-hop reasoning will make a graph query layer non-negotiable for sophisticated AI agents. The next check will likely hinge on PuppyGraph proving that its engine isn’t just faster, but essential. Can a zero-ETL graph become the default connector between the data lake and the AI brain?
Sources
- [Perplexity Sonar Pro Brief] PuppyGraph company and product overview
- [Axios Pro, November 2026] Exclusive: PuppyGraph raises $10M Series A | https://www.axios.com/pro/enterprise-software-deals/2026/11/10/puppygraph-query-defy-venture
- [Business Wire, November 2024] PuppyGraph Raises $5 Million in Seed Funding Led by Defy.vc | https://www.puppygraph.com/blog/puppygraph-raises-5-million-in-seed-funding-led-by-defy-vc
- [PuppyGraph] Company website and product claims
- [VMblog, June 2025] PuppyGraph as a Databricks Unity Catalog design partner
- [Insights from Analytics] Partnership details with Confluent, Red Panda
- [InfoWorld] Details on PuppyGraph 1.0 AI features
- [Data Day Texas] Weimo Liu presentation at Data Day Texas 2026
- [LinkedIn, November 2026] Customer traction post