TSMC plans to produce 3-nm chips in Japan from 2028- filing

TSMC plans to begin producing advanced 3-nanometer chips at its second Japanese factory by 2028, targeting 15,000 12-inch wafers per month with equipment installation and mass production slated to commence in 2028. The second unit is in Kumamoto prefecture and the expansion is aimed at shoring up capacity to meet outsized AI-driven demand while TSMC also expands overseas foundries in the U.S. and Japan.

Analysis

TSMC’s ongoing geographic expansion shifts the marginal supply shock from a Taiwan-centric problem to a multi-jurisdictional rollout that will change pricing dynamics for advanced nodes over a multi-year horizon. The immediate market reaction understates the elasticity of customer sourcing: large AI customers can flexibly allocate next-gen tape-outs across foundries within 12–24 months once qualification risks fall, which compresses spot premiums and reduces bargaining power attached to capacity-constrained suppliers.

Equipment and materials vendors are the non-obvious primary beneficiaries of incremental overseas capacity—orders for lithography, etch, deposition and substrates are backloaded but concentrated, creating a 9–30 month equipment cadence where revenues and margins can outpace actual wafer shipments. Conversely, smaller pure-play legacy-node foundries and OSATs face compressed utilization and pricing pressure as buyers reprice whole-supply chains to favor scale and geopolitical diversification.

Key risks are cadence and yields: a multi-quarter slip in yields or a clampdown on cross-border equipment transfers would re-concentrate scarcity and reflate prices quickly. Watchables that will flip the trade: (1) public customer allocation updates and reported order backlogs (weeks–months), (2) equipment OEM order intake and lead-times (quarterly), and (3) inventory signals at hyperscalers (quarterly to semi-annual)—any of which can amplify or reverse value across the chain within 3–12 months.

Key Decisions for Investors

• Buy TSM (TSM) equity or purchase 12–18 month LEAP calls (target +40–60%, stop -20% from entry). Entry on a <10% pullback or after a confirmed positive equipment order intake print; thesis: durable margin for advanced-node foundry share gains if customer allocation trends persist.
• Long semiconductor equipment exposure: buy LRCX or TEL (6–12 month horizon). Target +30–50% if OEM order books/lead-times extend; stop -18%. Rationale: outsized revenue leverage to new fab build cycles and faster ASP recovery for tool segments.
• Relative pair: long TSM / short INTC (12 months). Position size 1:1 notional; target asymmetric return ~+35% / -20% if Intel’s foundry cadence misses while TSM captures incremental AI tape-outs. Close or trim on signs of Intel winning major hyperscaler allocations.
• Income entry: sell TSM 6-month cash-secured puts 8–12% OTM to collect premium (max loss = assignment at strike minus premium). Use if willing to own TSM at a disciplined entry price—effective yield target 4–8% over 6 months with capped downside.