Five years is a long time to ask a startup to keep its voice down. Yu Xu spent roughly that long developing TigerGraph before its public launch in 2017. While other founders were learning to explain themselves in a pitch, his team was building a database. The intended result was speed. The method required a considerable amount of waiting.
That tension runs through Xu’s career: impatience with slow answers, patience with the machinery needed to produce faster ones. He had seen businesses wait days or weeks for reports. He had worked inside Twitter, where relationships between people and interests helped shape the product. He wanted to bring that way of working with connected data to companies that could not afford to invent their own infrastructure.
TigerGraph began in 2012. Its founder’s interest began much earlier, with an office payroll system, a mother who used it, and a child intrigued by what a database could remember. There is something pleasingly unglamorous about that starting point. Before venture capital and graph analytics, there was the administrative business of getting people paid.
A payroll system, then the Pacific
Xu has described roots in a remote village in Hunan, China. His mother was the first person in her office to use an HR management and payroll system. Its reporting capability caught his attention. The software would look primitive beside today’s tools, but it showed him that information could be stored and retrieved in useful ways. A database was already doing something consequential.
He later moved to the United States to study at the University of California San Diego, earning a Ph.D. in Computer Science and Engineering. The attraction included the university’s database program and Southern California’s beaches. Even a life devoted to data management can make room for a coastline.
Around 2000, he worked at a San Diego startup founded by a former doctoral adviser. The dot-com collapse interrupted that early encounter with entrepreneurship, and he returned to complete his doctorate. His route to founding a company included an experience of what happens when a promising technical venture meets an unforgiving market.
His subsequent work at Teradata put him close to large enterprises and their data problems. As a Hadoop architect, he worked on big-data initiatives. This was an environment in which storing information was only part of the challenge. The harder question was how long a business had to wait to make sense of what it had stored.
“I love databases. I want to do databases.”
Yu Xu, 2022
The delay that became a business
At Teradata, Xu encountered customers including eBay, Walmart and Bank of America. Some business intelligence reports took days or weeks to finish. For someone whose professional life centered on databases, that delay suggested a technical problem worth pursuing. Companies had accumulated information; asking certain questions across it remained expensive.
Consider an illustrative question about a payment. An analyst might need to connect an account to a device, that device to other accounts, and those accounts to transactions. Each relationship adds context. A single transaction can appear ordinary while its surrounding pattern deserves attention. The value lies in following the connections far enough to see the pattern.
A graph represents entities as vertices and relationships as edges. It gives those connections an explicit place in the data model. Xu’s ambition was to make this approach useful on large, changing enterprise datasets, where an answer that arrives next week can be an answer to a question nobody is asking anymore.
Twitter sharpened the idea. Xu worked on its data infrastructure before founding TigerGraph. Internet companies were already using graphs for their own purposes: social relationships, recommendations, knowledge and interests. Their internal systems demonstrated what was possible, but those systems were designed for the companies that owned them. A bank or retailer needed something it could actually adopt.
Five years behind the curtain
The startup initially operated as GraphSQL. By September 19, 2017, it emerged publicly as TigerGraph, announcing general availability and $31 million in Series A funding. It also introduced GraphStudio, a visual development tool. The launch joined the technical system to ways people could begin working with it.
That date gives the quiet development period a clear ending. Founded in 2012, publicly launched in 2017: a substantial stretch of the company’s early life went into making its product ready. In later accounts, Xu described first-release development as taking five or six years. The calendar matters more than the tidy slogan. This was a sustained engineering effort.
The launch already named customers including Alipay, Visa, SoftBank, Wish, Elementum and State Grid Corporation of China. The team had been working with customers before addressing the wider market. Enterprise software has a useful way of puncturing a founder’s confidence: eventually somebody tries to run it on their own data.
Xu’s 2017 launch essay emphasized “deep link” analytics, following relationships through three or more hops. It also treated storage and computation as parts of the same design. Keeping a map of connections was one requirement; calculating across that map quickly was another. His argument depended on delivering both.
An old academic connection, still working
One relationship in Xu’s own career endured beyond graduate school. He has identified UC San Diego professor Alin Deutsch as a mentor and credited him with helping shape TigerGraph’s architecture and its query language, GSQL. In June 2018, TigerGraph appointed Deutsch chief scientist, giving that academic connection a formal place inside the company.
The collaboration appears in the technical record, too. Xu, Deutsch, Mingxi Wu and Victor Lee coauthored the 2019 paper TigerGraph: A Native MPP Graph Database. It described a system built for massively parallel queries and analytics, with a language designed to accommodate SQL-style programming as well as iterative graph algorithms.
The 2019 paper also reported scale-up and scale-out performance against a published benchmark, with code on GitHub for reproducibility. That gave other researchers a way to inspect the work and evaluate the performance claims for themselves.
There is a human consideration inside that language design. A database may be technically capable, yet still ask too much of the people expected to use it. Making familiar programming habits useful in a new system reduces one obstacle to adoption. Xu’s work had to travel from a research idea to a tool someone else could understand.
His professional biography also credits him with 26 patents in parallel data management and optimization. The number belongs alongside the doctorate and the papers: evidence of a career spent on the mechanics of information. TigerGraph’s founder came to entrepreneurship with a technical specialty already developed over years.
The customers had questions of their own
Performance could win attention. Adoption required more. Xu’s account of the company’s development describes an initially engineering-heavy organization that later brought in a separate product management team and customer advisers. The people building the system needed the perspective of people trying to use it across more teams and applications.
His 2020 advice to aspiring entrepreneurs was to give high-quality software enough time. He also described the changing difficulties of running TigerGraph: early conversations involved explaining why a new database was needed; later ones involved shortening enterprise sales cycles and navigating procurement. The bottleneck had moved from the machine to the organization.
By 2022, his ambitions included making graph tools easier to use through visual interfaces, alongside applications for supply chains and financial crime detection. Dragging and dropping could make a question accessible to someone who did not want to become an expert in graph algorithms. Ease of use was part of the product’s next job.
That same year, Citi Ventures included Xu in its Enterprise Tech Summit panel on fighting financial crimes with AI. The subject offered a concrete setting for his technical interests: finding relationships that help institutions evaluate suspicious activity. A diagram becomes useful when someone can make a better decision with it.
A funding round, and a change of chair
In February 2021, TigerGraph raised $105 million in a Series C round led by Tiger Global, bringing total funding above $170 million. The company described plans for product development, cloud capabilities and geographic expansion. Money could finance a wider effort; it also brought a larger organization for its founder to lead.
The years afterward were less straightforward. In 2023, Mingxi Wu took over as CEO, while Xu moved to CTO. Reporting at the time connected the change to the company’s cash burn and financial management. The technical work that had established TigerGraph and the operational work of sustaining it demanded different kinds of attention.
Xu’s career has since taken another turn. Databricks now identifies him as Director of Product Management. In June 2026, he coauthored posts on Liquid Clustering and full-text search indexes. The topics return to a familiar concern: how information is organized, and how quickly a system can find the useful part.
His name also appears among the authors of the 2025 TigerVector paper, which brings vector search and graph queries together for retrieval-augmented generation. The technology has moved into another set of applications. Xu’s enduring question remains practical: what must a database do so that the person waiting for an answer can get back to work?