Decision time: May 15, 2002 at 09:00 Taiwan time
Knowledge cutoff: May 14, 2002 at 23:59:59 Taiwan time
Decision
TSMC's board must decide whether to preserve the dedicated-foundry model and how aggressively to fund 300mm, advanced process technology and added capacity after the 2001 semiconductor downturn. The recommendation is gated pure-play investment: keep technology and yield work moving, preserve customer neutrality, and release capacity modules only when comparable utilization, diversified binding demand and stressed liquidity clear explicit thresholds. source · tsmc.cutoff.2001-20f
This is not a target-price report. Capitalization-complete valuation inputs and human approvals are absent, so no market price, return forecast or investment recommendation is inferred.
Reference class and industry economics
This is a fixed-cost, capital-intensive manufacturer in a cyclical industry where both premature capacity and technology delay can destroy value. The 2001 stress was severe: consolidated ROC-GAAP sales fell to NT$125.885 billion from NT$166.198 billion, net income to NT$14.483 billion from NT$65.106 billion, and average utilization to 51% from 106%. source · tsmc.cutoff.2001-20f
Management described the mechanism directly: excess capacity may require lower prices or operation well below full capacity, compressing margin. It also warned that delayed advanced processes could lose customer orders. [evidence.tsmc.cutoff.2001-20f.fixed-cost-risk; evidence.tsmc.cutoff.2001-20f.technology-risk] The decision therefore cannot be reduced to “spend” versus “do not spend”; it must distinguish continuous learning and yield work from separable capacity commitments.
Business model
TSMC manufactured customer or third-party designs and described itself as the largest dedicated foundry. The model was already broader than wafer fabrication: fabrication represented about 86% of 2001 sales, with design, mask, probing, testing and assembly services providing the balance. Its design tools and technology files helped customers implement designs against TSMC process rules. [evidence.tsmc.cutoff.2001-20f.business-model; evidence.tsmc.cutoff.2001-20f.revenue-scope; evidence.tsmc.cutoff.2001-20f.design-enablement]
Fabless companies accounted for about 66% of 2001 sales and IDMs about 33%. [evidence.tsmc.cutoff.2001-20f.customer-mix] Management said it competed on process technology, quality and service rather than price. [evidence.tsmc.cutoff.2001-20f.competition] These facts support preserving neutrality and deepening enablement, but they do not prove a quantified trust premium. Concentration is the counterweight: the top ten rose from 49% of 2001 sales to approximately 63% in Q1 2002, and NVIDIA alone represented approximately 17% in 2001. [evidence.tsmc.cutoff.2001-20f.customer-concentration]
Financial reconstruction
The frozen financial tables preserve reported scope rather than manufacturing comparability. The annual table is consolidated ROC GAAP. The Q1 update is an unconsolidated issuer presentation whose sales and wafer figures include affiliate-manufactured wafers and whose capacity measure is managed scope. Its reported 33.6% gross margin differs from the 36.7% margin for TSMC's own manufacturing. [evidence.tsmc.cutoff.2002-q1.scope; table.tsmc.cutoff.q1-2002]
Q1 2002 nevertheless showed a meaningful sequential signal: utilization reached 67%, net sales were NT$35.790 billion, operating income NT$8.182 billion and net income NT$6.588 billion. [evidence.tsmc.cutoff.2002-q1.results; table.tsmc.cutoff.q1-2002] It is evidence of improvement, not proof of a durable recovery. Contemporaneous reporting still described inventory overhang and the preceding collapse in utilization. source · tsmc.cutoff.forbes-foundry
Liquidity and financing
At December 31, 2001 TSMC reported NT$38.954 billion of cash, equivalents and short-term investments; NT$57.668 billion across short-term bank loans, current bonds and aggregate long-term debt; and NT$22.262 billion of unused credit lines. [evidence.tsmc.cutoff.2001-20f.liquidity; evidence.tsmc.cutoff.2001-20f.debt-credit; table.tsmc.cutoff.liquidity] Q1 operating cash flow was NT$21.0 billion, cash used in investing NT$10.8 billion, period-end cash NT$58.3 billion, and a NT$15 billion bond issue partly funded the reserve. [evidence.tsmc.cutoff.2002-q1.liquidity; table.tsmc.cutoff.q1-liquidity]
Customers had deposited about US$48.8 million for future capacity. [evidence.tsmc.cutoff.2001-20f.customer-deposits] That proves a limited mechanism for customer-backed investment, not coverage of the capital program. The approximately US$2.5 billion public 2002 plan is retained in its stated currency; the filing's US-dollar amounts were convenience translations at a period-specific rate and are not silently mixed with NT-dollar cash flow. [evidence.tsmc.cutoff.2001-20f.fx; evidence.tsmc.cutoff.2002-q1.capex]
Competitive position and alternatives
The alternatives are: execute the public plan at ordinary project cadence; retrench and harvest; enter proprietary branded chips; require far greater customer financing; or use gated pure-play investment. The branded-chip path lacks evidence of superior economics and risks conflict with customers whose designs TSMC manufactures. Retrenchment protects near-term cash but can sacrifice advanced-node learning. Customer financing improves demand evidence but can give concentrated customers too much control. The public plan maintains speed but insufficiently protects against forecast error.
The strongest balanced choice is to preserve the pure-play model while separating technology learning from capacity volume. Release each module against stable-scope utilization, binding and diversified commitments or prepayments, verified yield economics and a twelve-month stressed liquidity floor. [judgment.tsmc.cutoff.pure-play-focus; judgment.tsmc.cutoff.gated-capex; judgment.tsmc.cutoff.financial-flexibility]
Disconfirming evidence and unknowns
Evidence against faster investment includes the 2001 collapse, explicit fixed-cost risk, inventory correction, customer concentration and the rapid change in public capex expectations. In October 2001 Morris Chang expected 2002 spending below 2001; by spring 2002 management cited improving conditions in raising its plan to approximately US$2.5 billion. source · tsmc.cutoff.register-capex
Evidence against retrenchment includes Q1 improvement, customer deposits, fabless demand potential and the disclosed need to remain ahead technologically. [evidence.tsmc.cutoff.2002-q1.results; evidence.tsmc.cutoff.2001-20f.customer-deposits; evidence.tsmc.cutoff.forbes.fabless-demand; evidence.tsmc.cutoff.2001-20f.technology-risk] The public corpus lacks board minutes, node-level yield and contribution, binding order duration, cancellation rights, equipment penalties and a stressed cash forecast. Those gaps make project-level gating essential and cap recommendation confidence at moderate.
Source and measurement controls
The 2001 Form 20-F is the authoritative cutoff source for annual financials, strategy and risks, but it is interested issuer evidence. Quarterly issuer reports are unaudited and scope-sensitive. The Register and Forbes add independent contemporaneous context but also contain quotations, estimates and forecasts rather than audited facts. source · tsmc.cutoff.2001-20f source · tsmc.cutoff.2002-q1 source · tsmc.cutoff.register-capex source · tsmc.cutoff.forbes-foundry
The packet does not merge consolidated with unconsolidated values, TSMC-owned with affiliate-managed utilization, ordinary wafers with eight-inch-equivalent wafers, NT dollars with convenience US dollars, or capital expenditures with total investing cash flow. [claim.tsmc.cutoff.scope-controls]