The Chip War as a Case Interview: TSMC's Risk

The first thing you must establish in any concentration-risk case is where the system breaks if one node fails.

Think about a highway system where 90% of a city's freight runs through one tunnel. Traffic engineers do not just study the tunnel. They study what happens the moment that tunnel closes: which routes saturate first, which industries stop receiving inputs, how long before the whole economy slows. TSMC is that tunnel. The rest of the global technology supply chain sits on the other side of it.

In case interview terms, this is your single point of failure, and you name it in your opening structure. Every branch of your issue tree either describes the risk itself, quantifies its impact, or evaluates the mitigation options. Do not bury it in the middle of your framework. Put it at the top.

When you face this case, state it plainly: "The central risk is geographic and operational concentration. If TSMC's Taiwan fabs are disrupted, there is no near-term substitute for advanced logic chips. The case is about how a client should price that risk and what levers exist to reduce it."

Concentration risk has three distinct dimensions, and conflating them produces a muddled answer.

Imagine you are stress-testing a bridge. You test load capacity, material fatigue, and environmental exposure separately before combining them into an overall safety rating. Mixing the three into one undifferentiated "risk score" tells you nothing actionable. The same logic applies here.

The three dimensions for TSMC are: geographic concentration (one island, one political flashpoint), customer concentration (Apple, Nvidia, AMD each representing massive revenue share), and process concentration (only TSMC can manufacture at 3nm and below at volume today). Each dimension has different probability drivers and different mitigation timelines.

A strong candidate maps all three, assigns a magnitude to each, and then sequences the mitigations by feasibility and time horizon. Intel Foundry and Samsung are years behind on process nodes. TSMC Arizona adds capacity but not geopolitical safety in the near term. State both facts as constraints, not as disclaimers.

Identifying the concentration is table stakes. Partners reward candidates who quantify the cascade.

Practice this framework on a real case: the tsmc-geopolitics-2024 case on BoardroomIQ puts you in the room.

Nvidia reallocating chip capacity away from China is not an isolated business decision. It compresses TSMC's revenue from one customer segment while expanding it from another, changes TSMC's geopolitical exposure profile, and accelerates US government interest in subsidizing fab diversification. Each of those second-order effects has a dollar value or a timeline you can estimate. Candidates who trace the chain earn credit. Candidates who stop at "this is risky" do not.

Use a simple impact tree: revenue at risk times probability of disruption, set against the cost of mitigation. You will not have perfect numbers in the room. Use anchoring facts (TSMC's 2023 revenue was approximately $69 billion; advanced nodes represent more than half of that) and reason to an order-of-magnitude answer.

Your recommendation must be specific enough that the interviewer can challenge it.

"Diversify the supply chain" is not a recommendation. It is a category. A real recommendation sounds like: "Our client should dual-source any component that flows through TSMC's N3 node, accept a 15 to 20% unit cost premium as insurance, and use the 36-month window before TSMC Arizona reaches full capacity to qualify an alternative supplier for non-leading-edge production." That is arguable. The interviewer pushes back. You defend it with your impact quantification from the previous step.

Structure your recommendation around three things: the action, the cost of the action, and what triggers the action. If Taiwan Strait tensions escalate past a defined threshold, the client executes. If they do not, the client pays the option premium and waits. That conditional logic is what separates a partner-level answer from a solid associate answer.

Build the issue tree from a blank page. Set a five-minute timer. Draw three concentration dimensions, attach second-order effects to each, and identify the two highest-priority mitigation levers before time expires. Do this three times across different industries to build the pattern recognition, not just the TSMC answer.

Anchor your estimates with real numbers. Look up TSMC's revenue, its top five customers by share, and the current production capacity of TSMC Arizona. Memorize the order of magnitude, not the decimal. In the room, a candidate who says "TSMC generates roughly $70 billion in revenue and Arizona represents less than 10% of current capacity" immediately signals preparation.

Simulate the pushback moment. Record yourself delivering your recommendation in under 90 seconds. Then argue against it. If you cannot generate a credible objection, your recommendation is not specific enough. Sharpen it until the pushback is obvious, then prepare your counter.

The best way to practice concentration risk is under realistic pressure, with a case that fights back. Open tsmc-geopolitics-2024 on BoardroomIQ and find out where your structure breaks.