Kazuki Ohta once left an investor meeting without understanding that he had just helped win an investment. Bill Tai had said yes. Ohta’s English was limited enough that the good news required a second delivery, this time from his co-founder Hironobu Yoshikawa. It is a pleasingly awkward beginning for a business devoted to making information useful.
Years later, Ohta would describe a company with more than $100 million in annual recurring revenue. But the meeting makes a better opening than the number. Here was an engineer who understood difficult systems, sitting in a room where a consequential piece of information had failed to arrive. Technical fluency and ordinary comprehension were, for the moment, conducting separate meetings.
He had come from Japan to build a software company in Silicon Valley. Treasure Data began in 2011 with Ohta, Yoshikawa and Sadayuki Furuhashi. Ohta became its founding chief technology officer. The subsequent story includes a sale to Arm, a return as CEO and, in April 2026, a new name: Treasure AI. Its more revealing episodes concern the work of choosing what to explain, and to whom.
Before the customer, the computer
Ohta’s education was in the machinery underneath the interface. He earned a bachelor’s degree at the University of Tokyo in 2008 and a master’s in 2010. His research concerned high-performance computing, including the movement of data through parallel systems. He also worked as a visiting researcher at Argonne National Laboratory, where his projects involved computing and storage at a scale far removed from the average laptop.
His technical bibliography preserves that earlier life. There are papers on parallel file systems, request scheduling and input/output forwarding. These titles have little of the breezy confidence of a startup slogan. They describe work in which the placement and ordering of requests can matter enormously, even if the person eventually using the result never thinks about them.
There were books, too. Ohta co-authored Japanese Hadoop guides published in 2011 and 2013. In 2020, the University of Tokyo’s School of Science recognized him as a distinguished alumnus. The route between those entries runs from studying how computers handle data to helping other people handle it.
He had already taken on commercial responsibility in Japan. At Preferred Infrastructure, he joined as the sixth member, later became CTO and worked on the distributed search engine Sedue. His career therefore combined research with the considerably less tidy business of putting software to work for an enterprise.
A meetup with consequences
Hadoop supplied another kind of education. Ohta recalled trying to use it on his laptop, making patches when it did not work, and helping organize a community in Japan. That community put him in conversation with developers dealing with growing volumes and varieties of data. He met Yoshikawa through the Hadoop conference world. The network around the software became part of the company’s beginnings.
In explaining the move to California, Ohta pointed to the people who knew how to sell and market software businesses internationally. Engineering talent was only part of the attraction. The company needed people who had already watched a commercial organization grow.
“Time is the only risk, right?”
Kazuki Ohta · 2017
He understood that subscription software would require years of commitment. The observation has an appealing lack of ceremony. A decade is a formidable thing to put into a company. Ohta discussed it as a practical condition of the job.
Ten million dollars, and still a question
The early Treasure Data could collect, warehouse and analyze information in the cloud. It reached $5 million in annual recurring revenue in its first two years, then $10 million by the third. Those figures looked encouraging. Selling still depended heavily on the founders.
Sales teams struggled, and the large cloud providers were making general infrastructure a difficult place to compete. The company examined what its customers actually did with the product. Marketing stood out. Treasure Data chose to build applications for that buyer, becoming a customer data platform.
The change had costs. Some employees left. Work on the underlying platform slowed temporarily, and the company had to change its marketing and sales approach. Eventually, recurring revenue rose from $25 million to $100 million over three years. The technical foundation remained useful; a specific business problem gave people a clearer reason to buy it.
Annual recurring revenue over three years after the shift toward customer data applications, described in Ohta’s 2021 talk.
There is a particularly uncomfortable lesson here for a technically accomplished founder. A system can be versatile enough to attract admiration while leaving the buyer with homework. Which department will use it? Who owns the budget? What happens on Monday after the contract is signed? A platform may answer many questions and still leave these unanswered.
Ohta’s story gives those questions some weight. The company had revenue before it found the focus that helped it grow further. Choosing a narrower commercial purpose required giving up some of the freedom that had made the original technology attractive. The customer supplied a discipline the architecture could not impose on itself.
The acquisition was a middle chapter
In August 2018, Arm announced that it had acquired Treasure Data. The chip technology company was assembling Pelion, an Internet of Things platform connecting device management, connectivity and data management. Treasure Data’s ability to bring together scattered information was part of that plan.
Arm said the company served more than 300 customers at the time. Its announcement described technology able to combine data from systems such as ecommerce, customer relationship management and connected devices. The data problem had followed Ohta into a different corporate setting: information arrived from many places, in forms that did not conveniently agree.
An acquisition also creates its own integration problem, with people and procedures taking the place of data formats. Ohta became Arm’s vice president of technology in its Data Services Business Unit. His account of the period includes work across culture, human resources, product, engineering, customer support and IT. The founder’s attention had to travel well beyond the code.
Treasure Data separated from Arm on June 25, 2021, becoming part of SoftBank Vision Fund 2’s portfolio. Ohta returned as CEO. Yoshikawa returned as executive chairman, and Dan Weirich returned as COO and CFO. The founding CTO was now responsible for the company’s daily business and product strategy.
There is a useful distinction between knowing a company’s origins and knowing how to operate its next stage. Ohta brought the first kind of knowledge with him. The new role demanded the second. The years of product work, customer contact and integration now sat behind a title with a wider set of obligations.
That November, Treasure Data announced $234 million in funding led by SoftBank Corp., with participation from Carbide Ventures. The company’s stated ambition extended the customer data platform across marketing, service and sales. The round gave that ambition capital; it also gave the returning leadership a larger undertaking to explain and deliver.
A founder on the other side of the table
Ohta’s investment preferences are unusually revealing about what he considers useful experience. On his personal website, he offers to share failures and pitfalls from building a global software business. He prefers teams with multiple founders, early companies and products capable of international distribution. He often invests alongside Yoshikawa.
The list of investments includes businesses working on data, software infrastructure and creative technology. His preference for helping before revenue arrives makes sense alongside that emphasis on mistakes. At that stage, a warning about the wrong customer or an avoidable organizational problem may arrive while there is still time to act on it.
His public description also has boundaries. He identifies business software, data and customer experience as areas he understands. That specificity is welcome from someone whose career has already demonstrated the inconvenience of trying to be useful to everybody.
Outside work, his personal site lists video games and sake. His media biography gets more precise: Fortnite, a sake collection and learning to prepare sashimi, with maguro o-toro a favorite. These are small details, but they give the executive biography some texture. There is a person here who has opinions about a piece of tuna, alongside opinions about a distributed system.
The next interface
On April 20, 2026, Treasure Data became Treasure AI. Ohta’s announcement described a business serving more than 400 enterprise customers and proposed a new way to work with its accumulated data infrastructure. AI agents would help run customer intelligence workflows, with people setting direction, boundaries and creative intent.
Treasure AI Studio puts a conversational workspace in front of tasks such as analysis, audience building and campaign work. The company presents the change as an extension of its data foundation. The old difficulty of assembling customer information remains underneath the newer interface.
For Ohta, this is another attempt to shorten the distance between a capable system and the person who needs something done. The research concerned moving data through computers. The customer data business concerned making that information usable for a particular buyer. The present ambition gives that buyer more ways to ask for action.
The early investor meeting remains a good scene to keep in mind. An important message was already in the room. Ohta needed help receiving it. Much of his working life since has concerned the practical business of helping information reach someone who can use it.