TSMC — pure-play foundry model and process-node leadership

1. Episode summary

Taiwan Semiconductor Manufacturing Company was incorporated in February 1987 in Hsinchu Science Park as a joint venture among Taiwan's National Development Fund, Philips of the Netherlands, and a small group of private Taiwanese investors, capitalised at roughly US$220 million. Its founding president, Morris Chang, had been recruited from the United States by Li Kwoh-ting to lead the Industrial Technology Research Institute the prior year. At the time, the global semiconductor industry was organised almost entirely around vertically integrated device manufacturers — Intel, Texas Instruments, Motorola, NEC, Toshiba — which designed, manufactured, and sold their own branded chips. Chang proposed that TSMC do only the manufacturing step and sell capacity to any design house that would buy it, abandoning any branded-chip ambitions of its own. That choice defined what came to be called the "pure-play foundry" model. Over the next two decades TSMC listed on the Taiwan Stock Exchange (1994) and NYSE (1997), built out a sequence of 200 mm and then 300 mm fabs at Hsinchu and Tainan, shipped the industry's first commercial 0.13 µm all-copper/low-k logic process in 2003, led the foundry industry into 65 nm and 40 nm volume production, and reached approximately 45.5% share of the global foundry segment on US$13.3 billion of revenue in 2010, by which date the fabless design ecosystem it had enabled — Nvidia, Qualcomm, Broadcom, MediaTek, AMD after its 2008–09 fab spin-off — was itself a multi-tens-of-billions industry. The strategic question the episode turned on was whether manufacturing-only contract production could be made a first-class business model in an industry whose prevailing peer-group read was that only integrated device manufacturers could sustain the capital and process-technology investments required at the frontier.

2. Sources

Primary:

1. Taiwan Semiconductor Manufacturing Company Limited, 2010 Annual Report — Business Overview (Hsinchu: TSMC, 2011), sections on market share, process-node milestones, and revenue. Reports 45.5% foundry segment share, US$13.3 billion consolidated revenue, 72% of wafer revenue from 0.13 µm and below, and the November 2010 half-millionth 45/40 nm 300 mm wafer milestone.
2. Taiwan Semiconductor Manufacturing Company Limited, 2009 Annual Report (Hsinchu: TSMC, 2010). Reports 2009 consolidated revenue of NT$295.74 billion (US$9.0 billion), down 11.2% from 2008's NT$333.16 billion (US$10.61 billion), and the June 2009 reappointment of Morris Chang as CEO.
3. SEMI, Oral History Interview: Morris Chang (Milpitas, CA: SEMI, conducted 2007; published in the SEMI oral-history series). First-person account by the founder of the 1985 ITRI recruitment, the founding rationale for the foundry-only model, and the Philips joint-venture structure.
4. TSMC press release, "TSMC First Foundry to Announce 0.13-Micron Design Starts" (Hsinchu: TSMC Public Relations, 2001), and corporate-history entries on the 0.13 µm copper/low-k node launch at production in 2003.

Secondary (with justification):

1. Chris Miller, Chip War: The Fight for the World's Most Critical Technology (New York: Scribner, 2022), chapters on Morris Chang, the founding of TSMC, and the pure-play foundry model. Synthesises archival, interview, and government-record evidence on the 1985–87 founding window and the subsequent industry restructuring; used here as the standard secondary reference.
2. Stephen Cass, "Morris Chang: Foundry Father," IEEE Spectrum (profile article, 2011), covering Chang's Texas Instruments career, his 1985 move to ITRI, and the founding of TSMC. Secondary because it aggregates interview and archival material into a single profile.
3. Brian Potter, "Morris Chang and the Origins of TSMC," Construction Physics (Substack, 2023). Secondary synthesis of Chip War, Chang's public talks, and industry-history sources; used for the ITRI–Philips–government capital-stack detail.

Tertiary (flagged):

1. "TSMC" and "Morris Chang," Wikipedia (accessed 2026-04-23). Flagged tertiary — used only for date cross-checks (1994 Taiwan Stock Exchange listing; 1997 NYSE ADS listing; 2005 Rick Tsai succession and 2009 Chang return).

Additional sources identified during Phase 0 §4 generation:

1. EE Times, "TSMC picks Applied's low-k for copper ICs, boosting CVD approach" (January 2001) and "TSMC aims to jumpstart copper in foundry with 0.18-micron process" — trade-press contemporaneous coverage of TSMC's 0.13 µm CVD low-k dielectric selection and the contrast with IBM's spin-on approach; used for Capability and Processes evidence on the 0.13 µm node.
2. TSMC press release, "TSMC Selects Applied Materials' Black Diamond Low-K Dielectric for Copper Chip Production in 200mm and 300mm Fabs" (Hsinchu: TSMC Public Relations, 2001) — corporate announcement confirming the Applied Materials CVD selection for 0.13 µm production; used for Capability evidence.
3. TSMC press release, "Cadence and TSMC Collaborate to Distribute Design Kits for Baseband and RF Foundry Silicon" (Hsinchu: TSMC Public Relations, March 2001) — confirms PDK development and TSMC-Online customer portal as structural customer-access infrastructure from 0.18/0.25 µm nodes; used for Structure evidence.
4. Digitimes, "TSMC dossier (3): Tech leadership and aggressive investment" (July 2021) — industry trade coverage of TSMC's fixed R&D-spend model (~8% of revenue) and its origin in Chang's Texas Instruments experience; used for Processes evidence.
5. Digitimes, "40nm crisis: how TSMC shaped Nvidia's triumph, AMD's demise" (November 2024) — reconstructs the 2008–09 40 nm yield crisis, Chang's return as CEO, and the quality-attribution failure TSMC corrected; used for Processes and Culture evidence.
6. Digitimes, "TSMC 40nm yield issues resurface, CEO promises fix by year-end" (October 2009) — contemporaneous trade-press report on Chang's commitment to resolve 40 nm yield problems within the quarter; used for Processes evidence.
7. TSMC Form 6-K filed with SEC (June 2009) — regulatory filing confirming Morris Chang's reappointment as CEO effective June 12, 2009; used to verify the date of the leadership transition during the 40 nm crisis.
8. IESE Insight, "TSMC: lessons in strategy and operational excellence from the world's chipmaker" (IESE Business School) — synthesises TSMC's operational culture, employee behaviour during the 1999 Chi-Chi earthquake, and manufacturing-as-mission ethos; used for Capability and Culture evidence.
9. Acquired podcast, "TSMC Founder Morris Chang" (Acquired FM) — long-form interview with Morris Chang covering the R&D budget philosophy, the 40 nm crisis account, and the "Come up with Solutions!" management norm; used for Processes and Culture evidence.
10. techovedas, "Why Morris Chang Returned to TSMC 4 years After Resignation?" — synthesis of Chang's public account of the 40 nm quality-attribution failure that prompted his return; used for Culture evidence.
11. Quartr, "The Silicon Empire: TSMC's Revolution and Morris Chang's Legacy" — industry synthesis covering the no-own-product cultural commitment and TSMC's mission framing; used for Direction and Culture evidence.

3. OTA narrative

Observe. The observation at the heart of this episode was Chang's reading of the semiconductor industry's cost and structural trajectory in the mid-1980s, formed during his long career inside Texas Instruments. Chang observed that fabrication cost per node was rising steeply while chip design was separating, in practice, from the capital base required to manufacture; that a rising class of engineers was ready to design chips if someone else would make them; and that the integrated-device-manufacturer structure created a trust problem for any design house, because every IDM also competed with its potential foundry customers. None of the three observations was exotic in isolation; each was industry-available to any senior executive who had sat in a general-manager seat at a major IDM. What was non-trivial, relative to the IDM peer group of 1985–87, was the composite reading: that the separation of design from manufacturing was a durable structural shift that could anchor a standalone business, not a temporary cost-accounting artefact. The observation required reading the industry against the prevailing peer-group read rather than with it; that qualifies as Hard-Correct Observe. Observe is a root-cause phase in this episode — it carried the strategic insight from which the rest of the episode followed.

Think. The reasoning step converted the observation into a concrete company design: a manufacturing-only entity that would guarantee it would never compete with its customers, would price capacity on the merchant market, and would stake its long-run viability on closing and then leading the process-technology gap against the IDMs. Two specific reasoning moves deserve naming. First, Chang committed to the no-own-product rule as a strategic credibility device — the trust problem the observation had identified had to be solved by a structural commitment, not by contract terms — and held to it through subsequent decades. Second, Chang and his R&D leadership committed, around the turn of the 2000s, to self-developing the 0.13 µm all-copper/low-k node rather than licensing from IBM as peers did, betting that a pure-play foundry could out-execute an IDM on a frontier node. Both reasoning steps were tightly coupled to the observation and were executed by a small leadership group against contrary industry opinion. Think is a weight-bearing phase in this episode and, with Observe, constitutes a composite root cause — it was the decisive translation from a read of the industry into a company design that could actually capture the opportunity the read identified.

Act. Execution ran over twenty-three years. It included securing the Philips technology-transfer and IP-licensing arrangement in 1986–87, staffing the first Hsinchu fab, building out a succession of 200 mm and then 300 mm fabrication plants, the 1994 Taiwan Stock Exchange and 1997 NYSE listings that financed the capital deepening, the in-house development and 2003 production launch of the 0.13 µm copper/low-k process ahead of IBM's competing node, first-mover production at 65 nm and then 40 nm, and the 2008–09 recovery from the 40 nm yield crisis that prompted Chang's return to the CEO role in June 2009. Execution was competent at every stage, and at the 0.13 µm node it crossed from competent follow-through into capability-building that redefined what a foundry could be. But the decisive strategic choice had been made earlier, at Observe and Think; Act's role across the episode was to ratify that choice repeatedly over two decades of capital cycles. Act was not the root cause of the outcome — it was the transmission step that carried a correct observation and a correct interpretation into two decades of compounding operational delivery. Characterised against the peer group, the execution sits Hard-Correct on the task-difficulty axis, but no phase in this episode is decisive at Act alone.

4. Modality evidence

Direction. The foundational Direction event in this episode was Morris Chang's founding decision in 1986–87 to constitute TSMC as a pure-play foundry — an entity that would manufacture chips and only chips, hold no branded-product ambitions, and commit publicly to never competing with its customers (SEMI oral history, Morris Chang; Miller, Chip War). This was a specific, dated, attributable strategic choice: Chang was recruited by Li Kwoh-ting in 1985, developed the foundry concept at ITRI, and brought the proposal forward as the founding rationale for the joint venture that incorporated in February 1987. The no-own-product rule was not a passive by-product of resource constraints; it was an active structural commitment designed to resolve the trust problem Chang had identified — that any IDM offering foundry services would always be suspect as a potential competitor to the design houses it served (Potter, "Morris Chang and the Origins of TSMC"; SEMI oral history).

A second attributable Direction choice extended the first: around the turn of the 2000s, Chang and his R&D leadership committed to self-developing the 0.13 µm all-copper/low-k node rather than licensing a reference process, betting that a dedicated foundry could lead a frontier node against integrated device manufacturers. This choice was materially distinct from the founding direction — it was a decision to compete on process-technology leadership, not merely on service and trust — and it carried the company from a reliable second-tier manufacturer into a position from which it subsequently led the industry at 65 nm and 40 nm (TSMC 2010 Annual Report; TSMC press release, 2001; EE Times, January 2001).

Structure. TSMC's structural architecture encoded the foundational Direction commitment in formal governance from the outset. The 1987 joint-venture structure — Taiwan National Development Fund (approximately 48%), Philips (approximately 28%), and private investors — gave the founding entity both state-level capital backing and, through Philips' technology-transfer and IP-licensing contribution, a route to process-technology credibility that an independent start-up could not have self-financed (SEMI oral history; Potter, "Morris Chang and the Origins of TSMC"; Miller, Chip War). Critically, no business unit within TSMC was authorised to pursue end-product design; the no-own-product commitment was structural, not merely aspirational, because there was no internal division that held such authority or resources. This architectural choice made the trust commitment credible in a way that contractual assurances alone could not have achieved.

The customer-facing structural innovation — TSMC's development of process design kits (PDKs), a dedicated customer web portal (TSMC-Online™), and eventually the Open Innovation Platform introduced formally in 2008 — converted the foundry's manufacturing capacity into a usable design environment for fabless companies that lacked internal process expertise. The March 2001 Cadence-TSMC PDK collaboration for 0.18 µm and 0.25 µm processes illustrates the structural investment in customer-access infrastructure (TSMC-Cadence PDK press release, 2001). These arrangements lowered the structural barrier between a fabless design house and volume production, widening the addressable customer base and creating switching costs on both sides of the relationship.

Scoring note (zero-modality rationale): the structural arrangements described in this subsection are classified primarily under Direction in the scoring record on the rationale that the strategic value derived from a specific, datable strategic choice that the architecture happened to host rather than from a novel divisional architecture or governance design (TSMC retained a conventional reporting hierarchy across the episode). The dedicated structural elements are counted as the operational substrate of the Direction modality rather than as an independent Structure contribution. Categorisation under METHODOLOGY-ota-scoring-v4.md §5: classification boundary with an adjacent modality. This follows the S-006 (Cisco) precedent for Structure-as-Processes-substrate.

Processes. TSMC's process-technology development cadence was the operational machinery that converted the Direction commitment into durable competitive position. Chang institutionalised an R&D spending commitment of approximately 8% of revenue — a rate that provided technical teams with multi-year budget visibility rather than annual negotiation uncertainty, explicitly modelled against the constraint Chang had observed at Texas Instruments where R&D was capped at lower levels (Digitimes, "TSMC dossier: Tech leadership and aggressive investment," 2021; Acquired podcast, TSMC founder Morris Chang). This consistent investment rhythm, tied to a process-roadmap discipline based on 18-to-24-month node advancement cycles, created a compounding planning process that IDM peers and competing foundries could not easily replicate by intermittent investment.

The 40 nm yield crisis of 2008–09 provides direct evidence of the process-recovery mechanism. When low yield rates on 40 nm production surfaced in mid-2009, threatening Nvidia and AMD's product roadmaps, the board reappointed Chang as CEO effective June 12, 2009. Chang's post-return account identifies the failure as both a manufacturing problem and a quality problem — with TSMC's previous quality-assurance process having incorrectly attributed root cause to the customer rather than to TSMC's own fabrication — and he committed to resolving it within the quarter (Digitimes, "40nm crisis: how TSMC shaped Nvidia's triumph, AMD's demise," 2024; TSMC Form 6-K, June 2009). The recovery restored the Nvidia relationship and preceded TSMC's transition into the Apple supply chain, confirming that the process-correction loop, not merely the initial process development, was a load-bearing operational competence.

Scoring note (zero-modality rationale): the Processes contribution described in this subsection is classified at the boundary with Capability per the methodology §3 Processes / Capability replacement test ("if the current operating staff were replaced by new hires of comparable background, would the operational pattern survive?"). The §4 evidence applies the test explicitly and concludes that the strategic weight sits on the Capability side — the operational edge depends on the specific individuals and tacit judgement carrying it, not on documented routine. The Processes component is acknowledged in narrative but does not carry standalone weight; both modalities are evidenced and the boundary call is recorded in the audit trail. Categorisation under METHODOLOGY-ota-scoring-v4.md §5: classification boundary with an adjacent modality.

Capability. The distinctive capability TSMC accumulated over the episode was the ability to develop and yield-ramp a leading-edge semiconductor fabrication node in a contract-manufacturing context — an environment significantly more demanding than node development at an IDM, because TSMC's processes had to work across a wide and heterogeneous portfolio of customer designs rather than the one or two product families an IDM's fab would run. The 0.13 µm all-copper/low-k node, brought to volume production in 2003 using Applied Materials' Black Diamond CVD low-k dielectric, illustrated this: while IBM chose a spin-on polymer approach suited to a narrow product line, TSMC selected a CVD approach that could accommodate the process diversity its foundry customer base required (EE Times, January 2001; TSMC press release, 2001). This was a capability call, not just a process choice.

Chang's own background from 25 years at Texas Instruments — including senior roles in semiconductor manufacturing and general management — provided the founding institutional knowledge base from which TSMC's engineering and operations staff were recruited and trained (Morris Chang Wikipedia; Miller, Chip War). The 1999 Chi-Chi earthquake episode, in which TSMC engineers reportedly returned to the fabs within hours of a 7.3-magnitude seismic event to protect in-process wafers, reflects the depth of craft culture embedded in the manufacturing workforce, a stock of tacit operational knowledge that could not be transferred to a competitor through licensing alone (IESE Insight, "TSMC: lessons in strategy and operational excellence"). By 2010 TSMC had accumulated a technology-node track record — from 0.13 µm through 65 nm and 40 nm — that represented institutional process-development knowledge built over roughly a decade of frontier-node production (TSMC 2010 Annual Report).

Culture. Three cultural mechanisms were load-bearing in this episode. First, the no-own-product rule required cultural maintenance as well as structural enforcement: in an industry where the most profitable move for a well-capitalised manufacturer would often have been to leverage its process advantage into branded-chip margins, the commitment not to compete with customers had to be continuously upheld as a shared norm, not merely obeyed as a rule. Chang's framing of TSMC's mission — manufacturing as a first-class vocation, not a service function subordinate to design — provided the normative substrate for this restraint (IESE Insight; Quartr, "The Silicon Empire"). Second, Chang instilled an explicit accountability culture, captured in the 1997–98 distribution to all managers of acrylic plaques reading "Come up with Solutions!" — a concrete norm-setting act linking individual ownership to problem resolution (Acquired Briefing, Morris Chang interview). Third, the 40 nm crisis resolution demonstrated how the culture handled accountability failure: Chang's return in 2009 was driven in part by his assessment that TSMC's quality process had incorrectly deflected blame to customers, a normative failure he treated as requiring leadership-level correction rather than technical-only remediation (Digitimes, 2024; techovedas, "Why Morris Chang Returned to TSMC").

The workforce culture — described consistently across sources as treating semiconductor manufacturing with the urgency of a national mission, with very low employee turnover and voluntary crisis response — reflects behavioural defaults that a new management team could not have purchased by replacing individuals (IESE Insight; Quartr). This distinguishes the cultural contribution from individual capability: the edge was in the collective orientation, not in any particular engineer's skills.

---