Table of Contents
TSMC’s October 15, 2026 results show a business strongly weighted toward advanced process technologies. That is valuable context for semiconductor investment and platform planning. It is not a lead-time quote for an established analog IC, controller or power-management device.
For a mature-node purchase plan, the key is to distinguish revenue mix from manufacturing utilization, then connect both to the specific product’s qualified production flow. Two companies’ reports can illuminate different parts of that problem without being directly comparable.
What does TSMC’s 79% advanced-node revenue share measure?
It measures the share of TSMC’s wafer revenue generated by 7-nanometer and more advanced technologies in the reported quarter. It does not measure the percentage of industry wafer capacity, the share of physical wafers shipped or the availability of legacy devices.
In its October 15 third-quarter release, TSMC reported the following wafer-revenue mix for the quarter ended September 30, 2026:
| Technology | Share of wafer revenue |
|---|---|
| 2 nm | 8% |
| 3 nm | 33% |
| 5 nm | 28% |
| 7 nm | 10% |
| These technologies combined | 79% |
The four reported shares sum to 79%. The remaining 21% is the rest of wafer revenue, not a statement that 21% of capacity is available for older products. Revenue depends on selling prices and mix as well as physical output.
The company also described strong demand for leading-edge technologies and issued a fourth-quarter outlook. Those statements should retain their time frame. A forward-looking management expectation is not a completed fourth-quarter result.
Why does UMC’s utilization signal need a separate reading?
UMC’s utilization measures production relative to its own capacity under its stated definition, for a different reporting period. It provides another foundry signal, but it cannot be subtracted from or averaged with TSMC’s node-revenue percentages.
UMC’s July 29 second-quarter report reported 85% utilization for the quarter ended June 30 and defined utilization as quarterly wafer output divided by quarterly capacity. It guided third-quarter utilization above 90%. The latter is the expectation published in July, not an actual Q3 result established by that document.
This distinction matters for mature-node planning: strengthening demand can also appear outside the most advanced nodes. At the same time, one foundry’s aggregate utilization does not disclose spare capacity for the exact process, site or device needed by an OEM.
Read the reports as two different indicators: TSMC shows where its wafer revenue is concentrated; UMC describes loading of its own manufacturing base. Neither provides a delivery date for a device on an OEM’s BOM. That date still has to come from the product supplier.
Trace the foundry signal to the part number
The connecting evidence is the device supplier’s product status, qualified manufacturing flow and accepted delivery schedule. Where that information is not public, record the uncertainty rather than assigning a node or a foundry based on the package marking.
Locate the constraint along the actual supply path. A wafer-fabrication constraint, an assembly or test constraint, and a distributor’s depleted stock can all appear to the buyer as a longer lead time. They call for different responses: a qualified process transfer may address the first, another approved backend route may address the second, and a different traceable channel may resolve the third.
Ask the device supplier which stage constrains the affected ordering code and what milestone changes the accepted schedule. Where a transfer is proposed, obtain the applicable product change notification, affected revisions or lots, and customer-approval requirement. Keep an undisclosed foundry or node marked unknown rather than guessing from the package or product age.
A fab-construction timeline describes why a building or equipment milestone does not instantly become finished-device output. The same reasoning applies when a buyer reads an aggregate utilization figure: capacity must be applicable and qualified before it can resolve a purchase-line exposure.
How should mature-node buyers plan into 2027?
Prioritize parts with a coverage gap and a long path to an approved alternative. For these parts, reconcile the accepted shipment schedule with production consumption and identify what must happen before another source becomes usable.
An offered fabrication transfer needs a qualification date and an affected-part list before it changes coverage. A promise of more assembly capacity needs an accepted finished-device schedule. A readily available alternate channel needs the exact ordering code, packaging and traceability review. Keeping those conditions distinct prevents an upstream announcement from being counted as a second finished-device source.
When accepted deliveries remain stable, continue the established replenishment policy while watching relevant change notices. If deliveries repeatedly slip, escalate the affected purchase lines and accelerate feasible alternatives. If customer builds fall, reconcile open commitments before adding inventory. These actions follow the buyer’s exposure; TSMC’s revenue mix helps explain the wider investment direction.
The industrial analog-demand review provides a complementary way to prioritize qualification time against available coverage. Use that comparison for the purchase plan, and retain the foundry figures as dated market context.
Frequently Asked Questions (FAQ)
Does a foundry’s unused capacity mean a legacy IC can be produced immediately?
No. The device needs an applicable qualified process, masks, manufacturing flow and downstream capacity. Aggregate unused capacity is not automatically transferable to a specific product.
Can buyers use a foundry’s guidance as a completed-quarter result?
No. Guidance is a dated expectation. Keep it labeled as such until the company publishes the actual result, and record any subsequent revision.
Should OEM buyers track the node of every component?
Track it where it is available and decision-relevant, but do not guess undisclosed processes. Exact product status, qualified sources, accepted deliveries and change notices are usually more actionable than an inferred node label.
