How TSMC Redrew the Semiconductor Map
When Morris Chang stepped down as CEO of TSMC on July 1, 2005, the company he founded had grown into the linchpin of a global industry. In 2004, TSMC posted a record US$7.65 billion in revenue, a 30% jump year-on-year, with more than 300 customers relying on its fabs to produce 5,000 different products. By that point, TSMC accounted for over 7% of global semiconductor output — an astounding feat for a firm that, just two decades earlier, embodied a radical new approach many believed could never work: the pure-play foundry model.
Chang’s decision to build fabs that exclusively manufactured chips for other companies was initially ridiculed in an industry dominated by vertically integrated titans like Intel, IBM, and Texas Instruments. Yet by 2005, the skeptics had been silenced. Chang’s logic was simple: as fabrication costs soared, it no longer made sense for every chip designer to finance its own multi-billion-dollar facility. By providing state-of-the-art manufacturing as a shared service, TSMC democratized access to Moore’s Law and liberated an entire generation of innovative fabless chip design companies.
Betting on Independence, Beating IBM
The scale of TSMC’s capital gamble was enormous. The cost of a leading-edge fab escalated from US$200 million in the early 1980s to US$3 billion by 2001. Few companies could absorb such sums. TSMC not only took on the burden but doubled down on developing its own proprietary process technology. Between 2000 and 2008, the company rolled out ten successive nodes, from 0.18 micron down to 40 nanometers. By 2010, its R&D budget had reached US$945 million, ranking it among the global top ten spenders despite being “just a foundry.”
This commitment to independence was perhaps most visible in 2000, when IBM approached TSMC to collaborate on 0.13-micron process technology. Chang declined. IBM had demanded that TSMC halt its own R&D and instead adopt IBM’s platform, a condition Chang recognized as a dead end. TSMC instead pushed forward with its own technology, achieving risk production in 2001. Rival UMC, which struck a partnership with IBM and Infineon, lagged behind badly. By 2010, TSMC’s revenue was 3.5 times UMC’s, a widening gulf that underscored the cost of misplaced bets.
Alliances That Mattered: The ASML Connection
Just as important as in-house development was TSMC’s strategic partnership with photolithography leader ASML. By adopting ASML’s Twinscan platform in 2001 and immersion lithography in 2004, TSMC gained a decisive edge in precision and yield. In 2012, it cemented the alliance by investing €838 million for a 5% stake in ASML as part of a joint program to accelerate extreme ultraviolet (EUV) lithography and 450mm wafer technology. This move guaranteed supply priority and positioned TSMC to lead the industry’s transition to ever-smaller nodes, eventually overtaking Intel and Samsung in the race to mass-produce 7nm and beyond.
Fueling the Fabless Revolution
TSMC’s foundry model catalyzed the explosion of fabless chip firms. By offloading the financial risk of manufacturing, companies from Nvidia and Broadcom in the U.S. to MediaTek and Novatek in Taiwan could thrive by focusing solely on design. Between 2000 and 2005, many of these firms doubled revenues, riding waves in wireless, gaming, and consumer multimedia. Taiwan’s own ecosystem flourished, with Realtek leading in audio codecs and Phison Electronics in flash controllers, all thanks to TSMC’s role as enabler.
Even industry incumbents that once dismissed the model were forced to adapt. By the mid-2000s, TI, STMicroelectronics, and Infineon embraced “fab-lite” strategies, outsourcing cutting-edge production while maintaining legacy fabs for older technologies. The gravitational pull of TSMC’s process leadership remade the industry landscape, dictating which firms advanced and which receded.
Intel, the PC Boom, and the Taiwan Paradox
While TSMC redefined manufacturing, Intel continued doubling down on vertical integration. Its Centrino platform in 2003 triggered a notebook PC boom, fueling shipments from 24 million units in 2000 to more than 200 million by 2010. Taiwan vendors Acer and Asus leveraged close ties to Intel and Microsoft to climb into the world’s top five PC brands. Yet even as they expanded massively, the economics were cruel: while Intel and Microsoft captured over 80% of the sector’s profits, Taiwan manufacturers and global PC makers squeezed by thin margins chased volume in a race to the bottom.
The paradox became clear with the Asus Eee PC in 2007, the netbook that briefly electrified the industry. Cheap, compact, and widely imitated, it epitomized innovation at the hardware level, yet delivered little profit. Consumers embraced bargain prices while PC vendors bled. The treadmill of contract manufacturing left Taiwan’s PC champions hustling for survival, while the real wealth accumulated in Santa Clara and Redmond.
The iPhone Shockwave
In the same year, the launch of Apple’s iPhone began dismantling this old order. The smartphone heralded a new era in which mobile computing displaced PCs as the center of gravity. Intel’s attempt to respond with its Atom processor fell flat, while Qualcomm, MediaTek, and Apple designed mobile SoCs whose fabrication relied on TSMC’s fast-moving process roadmap. For the first time, Intel faced a serious competitor not as a fellow integrated giant, but in the form of hundreds of smaller fabless firms powered by a Taiwan foundry.
From Challenger to Champion
By the late 2000s, TSMC’s unrelenting pace of node transitions, its alliances with ASML and toolmakers, and its commitment to process autonomy propelled it to undisputed leadership. More than a supplier, it became the backbone of the global economy’s future, from PCs and smartphones to game consoles and cloud datacenters. Taiwan, once relegated to contract assembly with compressed margins, could now claim an industry champion steering the most advanced technology platform on earth.
Chang’s vision had proved prescient: the company did not need to design chips to dominate the semiconductor age. It only had to build them better, faster, and more reliably than anyone else. Few CEOs can retire leaving behind an entirely new industry model. Morris Chang did (twice in fact), and the world continues to run on it.
