Donald Trump's assertion that Taiwan took America's semiconductor industry has become a fixture of his trade rhetoric. A detailed counter-argument published this week by a veteran of the US chip industry suggests the accusation not only misreads history — it inverts it.
Donald Merino, a former Intel licensing executive who writes under the name The Old Salt on Substack, spent years inside the intellectual property machinery of the American semiconductor business. This is not a think-tank rebuttal. It is an insider account of how the industry actually worked — contracts, patent cross-licenses, royalty schedules, and all.
It Started With a Check to an American Company
The story begins in the mid-1970s, when Taiwan made a deliberate national bet on semiconductors at a time when it could barely afford to. In 1976, Taiwan's government-backed Industrial Technology Research Institute signed a technology licensing agreement with RCA, paying US$3.5 million for access to CMOS manufacturing knowhow. Taiwan's GDP per capita at the time stood at roughly US$1,150. This was not a casual purchase. It was a calculated commitment of national resources to an industry that might not pay off for a generation.
Under the agreement, ITRI sent engineers to RCA facilities across the United States for structured training. Taiwan then built its first IC production line and within months was reportedly achieving better manufacturing yields than the original RCA facility. The lesson was already visible in 1977: Taiwan had not stolen a capability. It had licensed a foundation and immediately begun improving it.
Decades of Institutional Groundwork
What followed was not a shortcut. ITRI spun off United Microelectronics Corporation in 1980. Hsinchu Science Park opened the same year. Through the mid-1980s, Taiwan launched advanced process development programs and recruited Morris Chang — a Texas Instruments veteran with a reputation built on manufacturing discipline — to lead the next phase.
TSMC was founded in 1987 with a shareholder structure that itself refutes the theft narrative. Taiwan's government development fund held a major stake. Philips took roughly 27.5%, contributing capital, technology, and IP coverage under its existing patent-license arrangements. Both Intel and Texas Instruments had passed on the opportunity before Philips stepped in.
A Company That Paid Its Licensing Bills
Merino draws on TSMC's own regulatory filings to make a point that tends to get lost in political debate: TSMC operated squarely within the global patent system throughout its rise. The company held cross-licenses, paid royalties, and disclosed future licensing obligations in its annual reports. As of end-2002, those disclosed future payments totaled the equivalent of roughly US$157 million — around 3.35% of that year's net sales.
Philips' equity stake also served a legal function. So long as Philips held a specified ownership threshold, TSMC operated under Philips' patent umbrella as a covered affiliate — protection that TSMC itself acknowledged in filings as material to its ability to function without incurring significant additional costs.
The Model America's Own Companies Depended On
Perhaps the most inconvenient fact for the theft narrative is that TSMC's foundry model actively enabled the rise of American chip companies. By committing never to design or sell its own branded chips, TSMC solved a structural trust problem no integrated manufacturer could: a fabless startup could hand over its designs without fear of its manufacturing partner becoming a competitor.
Qualcomm's transformation into a chip design powerhouse, Broadcom's networking dominance, and Nvidia's emergence as the defining force in AI computing all relied on that guarantee. TSMC did not take something from American industry. It provided a service American industry needed and that no American company had chosen to offer.
Out-Executed, Not Robbed
Merino's conclusion is pointed. The United States did not lose semiconductor manufacturing leadership because Taiwan outmaneuvered it. It lost ground because its capital markets increasingly rewarded software margins, asset-light models, and short-cycle financial performance over the grinding, expensive, long-horizon work of process manufacturing. Taiwan made different choices — and compounded them over fifty years.
For policymakers in Washington now trying to rebuild domestic chip capacity, that distinction matters. Industrial strategy built on a false premise tends to produce the wrong remedies.
（Related： Global Talent War | Ep. 2: Intern at TSMC, Graduate With a Job: Taiwan’s New Export Is Education ｜ Latest ）
