Hironobu Yoshikawa had a respectable problem. He was sitting on startup boards, advising people who were doing something he had never done. He had worked in software. He understood venture capital. He could discuss a company's prospects. Running the company himself was the missing qualification.
Years later, he described the discomfort through baseball: coaching Major League players without having played the game. For someone already employed to exercise judgment, it was an awkward admission. The remedy would be considerably more expensive than buying a glove.
In 2011, he co-founded Treasure Data. Arm acquired it in 2018 in a deal reported at roughly $600 million. Today the business is called Treasure AI, and Yoshikawa is its chairman and co-founder, alongside his work as a general partner at Carbide Ventures. The return to investing completes a career loop. Somewhere along it, the advice acquired weight.
A literature student in the Linux classroom
Born in Tokyo in 1978 and raised on its outskirts, Yoshikawa liked novels, including Murakami, and the history of computers. At Waseda University, he studied literature. His first foothold in the software business came through an internship at Red Hat.
His assignment involved teaching people to build Linux servers. In 2001 and 2002, he trained engineers from companies including IBM, HP and Fujitsu. There is a pleasing reversal here: the literature student was helping established technology professionals become comfortable with a different way of building software.
He stayed at Red Hat, moving into work that joined technical knowledge with business development and partnerships. Open source was still earning its place in enterprise computing. He watched an approach many businesses had dismissed win commercial acceptance. That experience gave his later career a practical question to keep returning to: how do you make a business around technology people can share?
The introduction before the company
In 2007, Yoshikawa joined Mitsui's venture investment business. His early work covered technology businesses in Japan, China and Korea; a 2009 assignment brought him to Menlo Park. He entered Silicon Valley with an investor's job and an increasingly founder-shaped curiosity.
Looking into a potential investment in Cloudera drew him deeper into the data infrastructure community. Discussions with Kazuki Ohta, who had founded the Hadoop User Group Japan, helped turn that curiosity into a company. Their experience fitted together: technical depth, commercial experience and connections gathered through years of work.
Ohta remembered meeting Hiro at a Hadoop conference in Japan. He brought experience with high-performance computing and the enormous quantities of data it produced. Alongside fellow co-founder Sadayuki Furuhashi, they built a business around collecting, storing and analyzing data. The founding team had encountered the problem from different directions. They could recognize it without needing a fashionable name for it.
That background matters to the story. Treasure Data emerged from working relationships and existing technical communities. Long before the fundraising presentation, people had been meeting, comparing ideas and discovering whose skills could be useful to whom. A network sounds rather bloodless until it becomes your founding team.
Ten customers, one useful question
The company began operations in December 2011 and launched its initial product in September 2012. Its early customers included Wish. The attraction was straightforward: an internet business needed its engineers working on recommendation algorithms and experiments, while somebody else handled the machinery for collecting and analyzing the data.
Yoshikawa described working with ten early customers, mainly internet startups with technical buyers. All ten became paying customers. First-year revenue was below $1 million; learning from those users was his priority. Ten is a modest number beside the later acquisition price. It is also a number small enough for a founder to know what each customer actually needs.
The product's value could be expressed in an engineer's calendar. Time spent building infrastructure was time unavailable for the applications that made the customer's business different. Treasure Data offered to absorb that work. For its early buyers, the decision was about where to put scarce people.
What another investor saw
From the other side of the table, Sierra Ventures saw a complementary pair: Ohta could build software that handled substantial scale with a small team; Yoshikawa could turn conversations into customers. An introduction brought them to Sierra's Tim Guleri. A week separated the first meeting from a term sheet.
Sierra led the $7.7 million Series A in 2013. More rounds followed. Yoshikawa's ability to move between the technical explanation and the commercial case had become useful in a setting where neither could stand alone. A product needs somebody to buy it, and a persuasive founder eventually needs the product to work.
In 2018, Arm acquired Treasure Data as it pursued opportunities around connected devices and their data. Yoshikawa took a vice president and general manager role in Arm's data business. The founder who had wanted operating experience now had experience inside a much larger organization as well. His apprenticeship had acquired several departments.
Reported acquisition value when Arm bought Treasure Data.
A policy that leaves room for other people
Ask Yoshikawa about building an organization and the language becomes less technological. His recruitment rule was simple: “hire people better than you.” He connected that rule to practical choices about pay, stock options, professional recognition and the conditions in which people worked.
He advocated handling criticism privately and giving praise publicly. Contributions to open source deserved recognition. A company brand could give employees a sense of pride, while everyday appreciation helped people feel their expertise was respected. None of this requires a complicated management diagram. It does require the person in charge to remember it on an inconvenient afternoon.
“hire people better than you”
Hironobu Yoshikawa, on recruitment
Humility, in his account, followed from recognizing how much depended on others. That gives the baseball comparison an extra turn. Learning to play did not make the rest of the team dispensable. It gave him more reason to value what they could do.
Those principles also fit the way he had entered the business. At Red Hat, colleagues and customers needed a shared technical language. At Treasure Data, specialists needed their work to connect with a customer’s problem. A founder who hires for expertise has to accept that useful answers will come from elsewhere in the room.
The chair changes; the relationships remain
On October 1, 2020, Treasure Data announced that Pankaj Tibrewal would succeed Yoshikawa as CEO. Yoshikawa continued on the board, supporting the company's direction. The change marked a different relationship with the business he had helped start.
Carbide Ventures followed in 2021. Its team includes people whose careers overlap with Treasure Data, among them Dan Weirich and Tibrewal; early Treasure Data backer Bill Tai is a special limited partner. The names recur because the relationships continued beyond a single job or transaction.
The firm invests at pre-seed, seed and Series A, with a preference for business software and companies serving large enterprises. Its stated work includes helping businesses enter Japan and helping Japanese companies expand abroad. Yoshikawa's own route between those markets gives that proposition a personal history.
Carbide describes its team’s experience in terms of several functions: product, technology, finance and bringing a product to market. That is a useful way to read Yoshikawa’s own route. He had taught the technology, worked on partnerships, evaluated investments and then taken responsibility for customers and a company. Each job exposed a different part of the same enterprise. Returning to investment meant having more of those parts in view when another founder explained what needed to happen next.
The geography is familiar, but the practical work runs in both directions. Selling into a new market means finding people who understand its customers, expectations and working habits. Carbide makes that connection part of its investment proposition. Yoshikawa had already spent years translating between the Japanese software community and Silicon Valley’s commercial networks.
In February 2025, Carbide and NEC X announced a partnership connecting investment networks with the venture studio's startup programs. The arrangement included mentorship on fundraising and company growth. Here was a concrete setting for passing experience along: founders working through decisions with investors who had made some of those decisions themselves.
“Building a company is never easy,” Yoshikawa said in the announcement. A familiar sentence, perhaps, but one whose usefulness depends on who says it. From him, it comes after software classrooms, customer meetings, fundraising rounds, an acquisition and a change of chair.
A new question in Kyoto
Two developments in 2026 put the continuing story in different settings. On April 20, Treasure Data became Treasure AI. CEO and co-founder Ohta described a platform in which AI agents would help carry customer intelligence through to marketing execution, with people setting direction and constraints. The name changed; the old problem of turning scattered data into something useful remained recognizable.
Then, in September, Yoshikawa joined the Artificial Life Institute as a councilor. Based in Kyoto, the institute brings researchers together around questions about life, computation and the possibilities beyond present-day AI. Its governance includes people from research, entrepreneurship and finance.
Yoshikawa described wanting Japan to become a place that attracts international talent, connecting researchers and entrepreneurs across borders. He expressed an interest in futures that arise outside anyone's plan. For an investor, whose working life involves plans presented with considerable conviction, that is an appealing allowance for surprise.
The literature student who taught Linux has spent his career moving between people with different kinds of knowledge. First engineers and businesses, then founders and investors, now researchers and entrepreneurs. He wanted to earn the right to offer advice. Having played, he has found several ways to stay involved in the game.