In 1973, Taiwan was a humble middle-income country. Its GDP per capita, adjusted for purchasing power and inflation, was $8,000 — on par with Nicaragua or Bangladesh today. Its main exports were cheap consumer electronics and textiles, and “Made in Taiwan” branding was a symbol of low cost rather than technological sophistication. The country was trying to attract foreign investment in electronics manufacturing, but its industry was still concentrated in the low-value-added stage of the supply chain. But 1973 was the year that Taiwan set itself on a path to become something more. 1973 was the year that Taiwan established the Industrial Technology Research Institute.

ITRI is an applied R&D lab, founded to rapidly elevate Taiwan’s technological capabilities, particularly in electronics.¹ In addition to conducting research, ITRI acquired technology from abroad, trained Taiwanese firms to use it, and even provided services to firms directly — with the ultimate goal of making Taiwan a global leader in the electronics industry. 

It succeeded, to say the least.

ITRI started with a shoestring budget of NT$210 million ($16 million today) and 400 employees, of whom only 10 had PhDs.² It became one of the highest-return research initiatives in history. ITRI was responsible for creating multiple billion-dollar companies — including the legendary TSMC— and driving Taiwan to produce 60% of the world’s chips today. 

How did it do it?

In the early 1970s, Taiwan had a foothold on the bottom end of the value chain. Because of the country’s low wages and duty-free “export processing zones,” Western companies had already started to outsource the assembly of electronics to Taiwan. The result was a large industry of Taiwanese firms that were subcontractors for Western electronics companies, doing commodity manufacturing with no advantage other than low labor costs.

The vision for transforming Taiwan’s economy from low-tech to high-tech came from Sun Yun-suan 孫運璿, the Minister of Economic Affairs. Sun had trained as an engineer in the U.S. at the Tennessee Valley Authority, a New Deal-era project to modernize the Tennessee Valley through large-scale investments in infrastructure and electricity. There, he witnessed the TVA’s success in modernizing a poor and agrarian region — proof that economic and technological growth could come from the government taking on ambitious projects normally reserved for industry. When Sun became a government minister in Taiwan, he advocated for a public research institute to advance Taiwan’s industrial capabilities. In 1973, under his leadership, three existing research labs were merged to form ITRI.

In 1974, ITRI achieved its first major strategic success. One morning in a Taipei breakfast shop, seven men formulated a blueprint for Taiwan's semiconductor industry. One of them was Pan Wen-yuan 潘文淵, a Chinese-American engineer and former lab director at the Radio Corporation of America, then a leading chip manufacturer. Pan drew up a plan for ITRI to license RCA’s chip technology to develop Taiwan’s chip manufacturing capabilities. Sun agreed to spend $10 million on the deal — a large commitment for the nascent institute. As part of the agreement, ITRI sent a team of engineers to RCA’s facilities across the U.S. for a one-year training course on chip design, manufacturing processes, quality control, and other key elements. After returning to Taiwan, they set up a demonstration factory in 1977 to produce chips with RCA’s technology. Within six months, the factory was producing chips according to RCA’s specifications, with a 70% yield rate — even better than the 50% yield in RCA’s own factories in America.³ By 1978, Taiwan was producing over a million chips annually with RCA technology, mainly for electronic watches.

With the proof that Taiwanese factories could produce chips themselves, ITRI made plans to commercialize their new capacity. Scaling up the demonstration plant’s operations required private capital, as well as a dedicated corporate strategy — and ITRI was still just a research lab. So in 1980, ITRI spun out its first company, United Microelectronics Corporation. To kickstart UMC’s operations, ITRI gifted the demonstration plant to the new company, along with instructions for chipmaking and key engineers. With this running start, UMC became Taiwan’s most profitable company by 1985 (and is worth $20 billion company today). It was a remarkable success for an enterprise that was started directly by the government. This spinoff approach would go on to become a hallmark of ITRI’s strategy — the lab would spawn nine more companies over the next 20 years.⁴

By far the most important of these spinoff companies would be ITRI’s next child. While UMC produced older chips for low-power electronics, the global industry was racing to create more sophisticated Very Large Scale Integration chips for advanced computers. At first, Taiwan tried to keep up by launching its own VLSI development program. But ITRI’s then-president, Morris Chang 張忠謀, understood that Taiwan was unlikely to develop VLSI chips that could compete globally. Instead, Chang realized that it could lean into its manufacturing strengths and serve as a dedicated chip manufacturer for global firms. He transformed the VLSI program into what would become the crown jewel of Taiwanese industry: Taiwan Semiconductor Manufacturing Company. Like UMC before it, TSMC was founded as a joint venture. The Dutch company Philips Electronics provided capital and transferred its existing VLSI technology, while ITRI provided the workforce and training.⁵ Today, TSMC is the sole producer of the most advanced chips in existence, and consistently places among the world’s 10 most valuable companies. It is a singular company, surrounded by a burgeoning corporate mythology — but it started as a natural continuation of ITRI’s spinoff model.

ITRI continued to spin off successful companies after TSMC, including Taiwan Mask Corporation (1988) and Vanguard International Semiconductor (1994). But by the early 1990s, ITRI became entangled in conflicts with the very industry it had helped create. When ITRI launched its “submicron plan” in 1990 — to develop advanced 0.35μm chips — established firms objected strongly. UMC’s chairman argued that ITRI continuing to develop more advanced chips would create an unfair playing field, since its spinoffs were benefiting from government subsidies (an ironic argument, given UMC’s origin as an ITRI spinoff). In the 1970s, when ITRI was founded, no Taiwanese firms were capable of developing cutting-edge chips on their own. By the 90s, this was no longer true. Firms argued that ITRI was no longer necessary, since industry R&D had surpassed the institute’s own capabilities. The tension culminated in 1993, when the Ministry of Economic Affairs canceled the annual budget that had been guaranteed in ITRI’s founding terms. ITRI’s glory days were officially over.

ITRI continues to exist today, primarily as a service provider for firms, far from its origins as an R&D lab that developed advanced technology. But that is not exactly a tragedy: being supplanted by its children is the ideal outcome for an applied research institute. ITRI achieved its ambitious goal of making Taiwan the world leader in semiconductors. To do so, it relied on three basic principles for effective industrial policy. First, it started small and leveled up its capabilities incrementally. Second, it focused on building human capital in private industry. Third, it aligned its incentives with private firms.

ITRI started with a clear-eyed view of Taiwan’s strengths and weaknesses in developing chip technology. Its weaknesses were plentiful: it had no technologically advanced companies, no scientific research institutions, and a low-skill labor force not equipped to do advanced R&D. It had exactly one strength: its pre-existing electronics industry, which could be seeded with knowledge to become more technologically sophisticated. 

It is not unusual for developing countries to stumble by adopting overly ambitious scientific research agendas that do not produce tangible economic benefits. Of course, this can be a winning strategy — in rich countries. The National Science Foundation in the U.S. was chartered to focus primarily on basic research rather than applied research, based on forceful arguments that basic scientific capacity was a prerequisite for developing new technologies. Taiwan recognized that while becoming a science powerhouse might be the long-run ideal, it was not realistic given its low income and small scientific community in the 1970s. At the time, using science as a direct tool of industrial policy was a new idea. One of the ways Sun Yun-suan broke with precedent was by insisting that ITRI would report to the Ministry of Economic Affairs, instead of the Ministry of Science and Technology (then called the National Council on Science Development). This choice ensured that ITRI was held accountable for its economic value in the short-to-medium run, rather than pursuing scientific agendas of its own choosing. 

Another industrial policy failure mode is trying to jump directly to advanced technologies without building up a foundation of supporting industries and skilled workers. Around the same time as Taiwan established ITRI, France implemented the “Plan Calcul”, with the goal of creating a company that could compete with IBM in advanced computing. But the companies it supported never achieved competitive scale or leading technology, despite substantial subsidies. Their mistake was trying to produce cutting-edge technology, motivated by national prestige goals, despite lacking a preexisting electronics industry that would have the capacity to absorb that technology. 

Unlike France, Taiwan already had a large electronics industry. These firms were not exactly sophisticated, but they made the country better equipped to assimilate chip manufacturing technology. It is easier to transform a less advanced industry into a more advanced industry than to create an advanced industry from scratch.

Moreover, ITRI was not seduced by the idea of making computer chips, even though computers were the most exciting and advanced use case for chips. The more advanced computers of the era would have required 3-micron chips, representing the cutting edge of semiconductor technology. Producing these chips demanded specialized equipment, rigorous adherence to sophisticated manufacturing processes, and extremely clean environments, none of which Taiwan could reliably guarantee. Instead, ITRI started with electronic watches — a rapidly growing industry that used older 7.5-micron chips, making them easier to produce while still offering reasonable profit margins. This pragmatic approach allowed Taiwan to establish a foundation in semiconductor manufacturing without jumping too far ahead of its capabilities. 

Of course, an important premise of this approach is that when you start small, you can upgrade knowledge and capabilities to enable frontier technology development in the future. That is why the most important part of ITRI’s deal with RCA was the extensive training it provided to Taiwanese workers. ITRI sent teams of engineers to be trained in RCA facilities across the U.S. One group studied chip design in New Jersey, another learned about manufacturing process improvement in Ohio, a third group in California focused on quality verification, and a fourth group in Florida developed skills in managing equipment. Together, they built up a comprehensive understanding of the chip manufacturing process — not just the science of semiconductors, but also tacit knowledge of industrial processes and management techniques for the semiconductor industry. As a testament to the importance of this training, this initial corps of RCA-trained engineers went on to fill the top ranks of Taiwan’s industry, including future leaders of UMC, MediaTek, and TSMC. Even though the RCA deal was for old-generation technology, it formed the knowledge base needed to create more advanced chip technology later.

ITRI could have easily failed by being too ambitious. The first keystone of its success was to start close to Taiwan’s existing capabilities, prioritize capacity-building in its agreement with RCA, and build incrementally from there.

The foundation of ITRI’s first success — the licensing deal with RCA — relied on importing technology and training from abroad. But if ITRI wanted to ever start developing its own frontier technology, it needed to cultivate a strong R&D workforce in Taiwan that could then diffuse into private companies. Thus, ITRI’s second imperative was to accelerate human capital development in Taiwan’s electronics industry.

The most important way that ITRI grew the R&D workforce was through the creation of the Hsinchu Science-Based Industry Park in 1979. Hsinchu Science Park was an industrial cluster, created to enable domestic electronics firms to grow using low-interest loans, tax benefits, and special exemptions from tariffs on their inputs. But greater than the tax and administrative benefits Hsinchu Science Park firms received, the main draw was the workforce. The industrial park was set up close to Taiwan’s best technical universities, Chiao Tung and Tsing Hua, providing ready access to skilled engineers. Moreover, to attract diaspora talent from the U.S., it had attractive housing, medical and health services, and educational facilities — including Taiwan’s only public bilingual secondary school. Talented engineers and researchers came to Taiwan from abroad in a “reverse brain drain,” bringing their own skills and upgrading the workforce available to Taiwanese firms.

ITRI used its proximity to Hsinchu Science Park to share even more knowledge and expertise with companies, just as Silicon Valley firms have benefited from research at Stanford and Berkeley. It hosted regular technology seminars, training workshops, and demonstration programs for employees at Hsinchu firms. Their proximity facilitated knowledge flows from ITRI to companies, making it easier for the private sector to benefit from ITRI’s accumulated knowledge.

Another way in which ITRI developed the R&D workforce of Taiwan was by training its own employees and dispersing them throughout industry. Even though ITRI had few researchers at first, it had an advantage in recruiting top talent. Taiwan required military service from university graduates, which they could meet by working at ITRI for at least four years. As a result, ITRI reliably recruited 400 top graduates from Taiwan’s best engineering colleges each year, and kept them on for four years. Critically, it did not just skim the cream of existing talent, but also helped develop it further through internal training and sponsoring of further education. ITRI would go on to fund employees to pursue PhDs and MBAs at Stanford and MIT, at an average cost per employee of $400,000 — a large expense, but worthwhile to bring expertise to Taiwan. From an initial cohort of just 10 PhD researchers, ITRI grew to include 900 doctorates among its 6,000 employees by the late 1990s, 120 of whom the institute sponsored directly.

ITRI also aimed for 15% of its staff to turn over each year. This counterintuitive goal had two purposes: first, it helped Taiwan’s industrial ecosystem acquire highly-skilled workers, many of whom had been funded by ITRI to do their PhDs in the U.S. Second, it ensured that ITRI’s internal process knowledge would spread to firms. Knowledge transfer in the semiconductor industry requires the movement of people who carry hands-on expertise that cannot be fully captured in documentation. When engineers trained in ITRI’s specialized facilities moved to private companies, they carried with them not just technical know-how but also the softer problem-solving approaches and organizational practices crucial for innovation. 

Since ITRI was not a firm, it did not see departing employees as a loss — rather, it was yet another way for it to provide essential inputs to Taiwanese industry. About 85% of ITRI's departing staff joined Taiwanese firms, creating an ever-expanding network of shared knowledge and expertise. By the late 1990s, this pattern of movement meant that about 40% of managers in the firms within Hsinchu Science Park had previously worked at ITRI. 

ITRI could have easily failed by hoarding its talent and knowledge. If ITRI had kept all its experts for itself, it might have produced impressive research, but it would have failed in its mission to upgrade Taiwan’s industrial capabilities. It might have even been counterproductive, denying top talent to private industry. Instead, ITRI proactively sent knowledge and talent to industry, so that Taiwanese firms could eventually upgrade their own R&D capabilities, rather than depending on ITRI indefinitely. 

ITRI’s strategy was excellent, but ambitious planners have dreamed up many excellent industrial policies that subsequently failed. The real challenge of industrial policy is navigating the delicate political economy of government-industry relations. It is an ever-present risk that subsidies could be allocated to unproductive but politically-connected firms, rather than to efficient firms that could power industrial growth. Why didn’t ITRI make this mistake?

From the beginning, ITRI was focused on R&D support. And one key difference between R&D support and other industrial policies is that firms only benefit from R&D if they are willing to innovate. A firm that receives tariff protection or subsidies doesn’t have to do anything — the policy puts money in their pocket unconditionally. Of course, industrial policies can try to require that firms become more efficient to be eligible for subsidies, but enforcing those requirements is difficult when firms have political power. 

In contrast, a firm that receives the blueprints for a chip fab simply will not benefit from them unless it actually sets up that fab and starts producing chips. That is not a legal requirement that firms can lobby against: it is a fundamental difference between knowledge and money. In other words, R&D support incentivizes firms to actually invest in their own productivity. 

This does not mean that R&D support is a magic bullet. Instead, it is especially susceptible to a different problem — firms might view it as wasteful spending, or worse, attempts to create competition. If firms saw ITRI as a waste or a threat, the political coalition supporting its already-threadbare budget could fall apart (as ultimately happened in the mid-90s). That is why successful R&D initiatives must be framed as beneficial for influential stakeholders, such as the large electronics industry in Taiwan.

During its critical first two decades, ITRI expertly managed this imperative. A clear example comes from their photomask duplication services. A photomask is a template used to etch a silicon wafer with the intricate pattern that makes it into a chip — it is a delicate, specialized, and absolutely necessary piece of equipment. Firms need to be able to duplicate their photomasks to use them at multiple stages of the manufacturing process, and for redundancy in case one mask is damaged. But mask duplication is a complex process. Taiwanese firms did not know how to do it themselves, and paying for mask duplication services internationally would have been prohibitively expensive. In 1979, ITRI licensed mask duplication technology from International Material Research in the U.S. Using IMR’s technology, ITRI established a department to provide mask duplication services to Taiwan’s chip manufacturers whenever they were needed. Demand for ITRI’s mask duplication services grew so much that in 1988, ITRI spun out its mask department into yet another company, Taiwan Mask Corporation. If ITRI did not exist or its operations were impaired, firms would have to get their mask duplication services from abroad at a higher cost, making them less profitable. Thus, ITRI’s mask department did not just strengthen Taiwan’s chip supply chain. It established ITRI as an invaluable service provider, so that firms would support its continued funding and independence. 

This incentive alignment is not something that every country can achieve. It worked because Taiwan’s electronics industry occupied a developmental sweet spot. The industry was large and sophisticated enough to benefit from R&D services, and important enough to provide political backing for ITRI. At the same time, its technology level and profit margins were low enough that firms were still hungry for more capabilities that they could not acquire themselves. They correctly saw that ITRI could enable them to produce more advanced goods, and thus earn higher profits without taking on risk. The combination of a large and influential industry with low profits  is essential to making applied R&D initiatives politically sustainable. 

It also helped that ITRI solicited incumbent firms for capital to invest in UMC and TSMC. This financing structure ensured that if an ITRI spinoff made profits, incumbent firms benefited rather than being displaced. ITRI was creating profitable subsidiaries for them, not competitors. This common interest was strengthened by the fact that all the firms and ITRI were co-located in Hsinchu Science Park. When firms form an industrial cluster, research shows that a new entrant benefits incumbents through agglomeration effects. In other words, ITRI and its spinoffs provided knowledge that made incumbent firms more productive than they would have been otherwise.

ITRI could easily have been crippled — as, in the end, it was — by making enemies with incumbent firms. Alternatively, it could have been hijacked to provide subsidies to incumbents without increasing their productivity. Instead, it promised risk-free R&D and services to firms eager to innovate. Those firms supported ITRI’s continued operation and allowed its incrementalist plans to bear fruit over the long run. 

Taiwan was one of the few developing countries to become genuinely rich in the 20th century, and, in contrast to high-profile failures in Latin America, a genuine industrial policy success story. Its technological ascendance has prompted reams of theories about development policy. Yet the country’s success is difficult to disaggregate from regional trends mirrored in the other “Asian Tigers,” and even alone, the extent to which its growth can be attributed to ITRI is not immediately clear. Even granting ITRI’s centrality, if the lab’s achievements were contingent on factors that cannot be replicated, the case for treating it as a model weakens considerably. The question, then, is whether ITRI offers generalizable lessons for technology policy today.

Some of the factors that helped ITRI were certainly lucky or specific to Taiwan. For example, ITRI could access top talent because of Taiwan’s military conscription, something that public R&D initiatives in other countries today cannot rely on. And creating a pure chip manufacturer in TSMC was a risky bet that could easily have failed. But history always has some amount of contingency — that doesn’t mean we can never learn from case studies. It remains true that ITRI had a careful plan that they executed on perfectly. They started with achievable technology goals and ladder up to more sophisticated technologies. They cultivated outstanding human capital in industry to fuel R&D efforts. They managed the political economy of industrial policy carefully to ensure that firms were supportive of their efforts for 20 years. These are valuable insights for technology policy today, even though they cannot guarantee anything like ITRI’s runaway success.

When scholars and policymakers discuss models of successful science and technology policy, they invariably turn to the same American benchmarks: DARPA, Operation Warp Speed, the NSF, the NIH. Meanwhile, ITRI receives scant attention, even though it is a more relevant benchmark to most countries trying to develop in critical sectors. Most countries would benefit from starting with achievable goals that are close to their existing capabilities, from building up the R&D workforce in private industry, and from aligning the incentives of private firms with their continued R&D activities. Global tech policy would flourish if, for every ten people trying to build the next DARPA, there was one trying to build the next ITRI.