Company history

In 1954 Ikuo Takeda, a thirty-year-old engineer at the Ministry of Communications’ Electrotechnical Laboratory, left to found Takeda Riken Kogyo in Toyohashi, Aichi. His strategy was deliberately contrarian: rather than fight electrical giants like Hitachi and Mitsubishi, he specialized in the electronic measuring instruments they would not touch, selling one-of-a-kind products with no substitute at premium prices. “I had no head for business,” he said later, “so technology was the only weapon I had.” The formula won a tiny firm high margins and an international name — a vice-president of America’s GE once travelled to the Toyohashi headquarters to see the work for himself.

Yet behind the engineering there was no management floor. The company had no cost accounting at all, so no one knew whether any given model made money or lost it. For more than a decade Takeda Riken grew as two halves that did not fit — world-class technology and a near-total absence of managerial discipline — and as long as its irreplaceable niche products kept selling, the void stayed hidden. It moved its head office to Tokyo in 1957 and, from the 1960s, pushed into minicomputer-based instruments, advanced work that pulled it toward electronics and toward the semiconductor-test business that would come to define it.

In 1972, on a government subsidy and after four years of development, Takeda Riken shipped the T320, Japan’s first IC test system, and with it stepped from instrument maker toward semiconductor-test maker. Then the ground gave way. The 1975 oil shock froze chip-industry capital spending; the company fell to a net loss on roughly $26.8M (¥8bn) of sales and carried $16.7M (¥5bn) of debt. Its main bank refused fresh credit and demanded the founder’s removal, and by the end of 1975 an internal coup pushed Takeda out of the company he had built over twenty years — the price, paid all at once, of two decades of technical brilliance without management.

To save the company, Takeda made an extraordinary appeal — to Hiroshi Seimiya, his former superior at the Ministry of Communications and now president of Fujitsu, gravely ill and in long-term care. Overriding strong opposition inside Fujitsu, Seimiya decided from his sickbed to take a rescue stake in Takeda Riken; he died that April. It was a rescue reasoned less from strategy than from the loyalty running between a former superior and his subordinate that secured the company’s survival.

In February 1976 Fujitsu sent in Toshimasa Umiwa as president. The first thing he found was that Takeda Riken had no cost accounting whatever — no one could say which models were profitable. He built the missing machinery piece by piece: model-by-model costing, a ban on discounting at the salesperson’s discretion, withdrawal from minicomputer instruments, and concentrated investment in IC testers. Fujitsu played three roles at once — shareholder, manager, and largest customer — and it was that combination, not capital alone, that made the rebuild work.

The results showed in the numbers: new products rose from 4% of sales in 1976 to 45% by 1981, and a 100 MHz test system co-developed with the state telephone monopoly lifted Advantest from fourth to second in the world in IC testers. In February 1983 the company listed on the Tokyo Stock Exchange, completing the rebuild — though at listing Fujitsu held 28% against the founder’s 5.68%. It renamed itself Advantest in 1985, pushed into Korea and Taiwan under an “Asia-first” policy from 1993, and by 1996 held roughly 40% of the world market for semiconductor test systems — the global leader in a business born in Japan.

World leadership did not bring stability. Advantest’s fortunes now rode the capital-spending cycles of the world’s chipmakers, and though its technology was irreplaceable it swung violently all the same: a $779.4M (¥73bn) net loss in the year to March 2009 after the financial crisis, then three straight loss-making years from 2012 to 2014 — extraordinary for the global market leader. Debt-free finances contained the damage, but the swings laid bare the structural fate of a technology specialist: the more chips shrink, the more testing matters, yet how much its customers make, and when, is decided beyond its reach.

The one great structural move of these years was the 2011 acquisition of Verigy for about $1.1B (¥91bn), which brought in the SoC (system-on-chip) test line where Advantest, dominant in memory testers, had never reached the front. Owning both halves — the volume game of memory and the function game of SoC — made it the outright world leader, even though the first year sank into a cyclical loss before the payoff came.

The capital tie also began to unwind. In 2005 Fujitsu sold part of its stake for $816.2M (¥90bn), stepping back from the position it had held since the 1976 rescue. Nearly three decades on, the bet Seimiya had made from his sickbed out of loyalty had become, for Fujitsu, a long-term investment that paid — and a relationship a colder era of capital efficiency would soon close out entirely.

In 2017 Fujitsu sold its remaining shares for $472.5M (¥53bn), ending some forty years of capital ties — $1.3B (¥143bn) recovered across the two sales, comfortably more than the 1976 rescue had ever cost. Advantest was now, in name and in fact, a company with no parent, carrying alone both the old problem of results dictated by chipmakers’ capex and the new freedom to decide for itself whether to reinvest its cash or hand it back to shareholders.

Then the demand it could never manufacture arrived from outside. The generative-AI build-out sent testing demand surging for AI SoCs and the high-bandwidth memory (HBM) that feeds them, and as world leader Advantest took in the wave in bulk — a record $1.2B (¥171bn) net profit in the year to March 2023, then fresh highs beyond it. It began sharing the proceeds: a four-for-one stock split in 2023 to widen its investor base, a raised dividend, and about $35.6M (¥5bn) of share buybacks — a company that under Fujitsu had always held cash back to reinvest now paying shareholders out.

Yet the way it earns has not changed since the Fujitsu years: it is still the specialist whose testers customers cannot do without, and still cannot set when they buy. Whether the AI supercycle is a break from that fate or merely its largest upswing yet — the highest peak in a cycle that has always come back down — is the open question the company now carries into its independence.