The capacity tightness driven by AI computing demand is steadily transmitting to the pricing of wafer foundry services.
According to Digitimes, citing supply chain sources, TSMC plans to raise its wafer foundry quotes starting from January 2027, with an overall price increase ranging from 3%–6%. Advanced nodes such as 2nm and 3nm will see higher increases; mature and specialty processes will not undergo a uniform price adjustment, but will be negotiated individually based on customers and capacity utilization rates.
Supply chain sources also mentioned that TSMC's 8-inch capacity utilization rate has currently exceeded 100%, with nodes below 45nm remaining fully loaded, and order visibility has already extended to 2030.
TSMC has clearly adopted a differentiated pricing approach rather than a blanket across-the-board increase. The core logic is straightforward: advanced nodes face the largest capacity gap and the strongest inelastic customer demand, thus bearing higher price hikes; mature nodes feature a more diversified customer structure with significantly varied bargaining power, allowing TSMC to retain flexibility for individual negotiations.
The underlying driving force supporting this round of price increase expectations stems from the ongoing spillover of AI demand. The past market perception was that AI only consumes high-computing-power chips such as GPUs and ASICs, but now the demand has expanded. In addition to the main computing chips, a single AI server requires a large number of PMICs (Power Management ICs), MCUs (Microcontroller Units), analog, and driver devices. Many of these chips are manufactured on mature nodes and specialty processes. The volume growth of AI servers simultaneously drives up wafer start demand on 8-inch and 12-inch mature production lines.
On one hand, advanced nodes are queuing up to secure capacity; on the other hand, the capacity utilization rate of mature nodes remains at maximum capacity. The simultaneous tightening of both supply and demand creates conditions for foundries to raise their quotes. The continued tightness in CoWoS advanced packaging capacity further amplifies the overall delivery constraints for advanced chips.
The most closely watched aspect of the price hike news is the extension of order visibility to 2030. This statement can be easily misinterpreted as demand not declining in the coming years, but it needs to be viewed rationally. Order visibility, in essence, refers to long-term intent orders where top-tier major customers lock in capacity in advance, rather than final purchase contracts with legally binding force. Major customers lock in capacity in advance to ensure chip supply over the next few years and reduce the risk of supply interruptions, which does not mean all intent orders will fully materialize. Once the pace of AI capital expenditure slows down in the future and downstream server manufacturers adjust their shipment expectations, long-term orders also face the possibility of downward adjustments. Long-term contract intents cannot be directly equated with locked-in guaranteed revenue.
Crucially, the cost pressure brought by the price hike will transmit down the industry chain layer by layer, but the absorptive capacity of different customers varies significantly. Top-tier customers holding high-computing-power chip projects and highly reliant on 2nm/3nm advanced nodes enjoy high premiums on their chip products and have the room to pass on part of the costs to downstream server manufacturers, thus facing relatively limited impact.
However, for many small and medium-sized IC design companies, especially those manufacturing analog and power chips, product gross margins are already low, and market supply is relatively abundant, making cost pass-through more difficult. These companies will face a dilemma: either absorb higher foundry costs and compress their own profits, or evaluate transferring some products to UMC, Vanguard International Semiconductor (VIS), or domestic mature node production lines for tape-out.
However, order transfer and substitution are not as easy as imagined. For chip projects already in fabrication at TSMC, IP migration, process transfer, and testing verification require substantial engineering investment, involving high costs and long cycles. It is not a matter of quickly switching production lines just because of a price hike. Many mature node customers, even in the face of price increases, can only stay on their original production lines, as the hidden costs of migration may even exceed the additional expenses brought by the price hike. This is also the core reason why TSMC possesses strong bargaining power.
The expectation of price hikes will also bring an opportunity window for other wafer foundries. UMC, Powerchip Semiconductor Manufacturing Corporation (PSMC), and domestic mature node production lines are expected to take over some orders diverted from customers. However, the objective reality is that substitute capacity does not emerge out of nowhere. Process matching, yield stability, and supporting ecosystems will all become strict prerequisites for taking over orders. In the short term, substitution will mostly occur in newly initiated projects, and the large-scale cross-fab migration of existing mature chips will be very slow.
From the perspective of the industry cycle, this round of price hikes is the result of the mismatch between AI computing capital expenditure and the long-term capacity expansion pace. TSMC's capacity expansion requires a long cycle; fab construction, equipment installation, and process ramp-up all take several years, and capacity release cannot keep up with the new chip demand brought by AI. Against the backdrop where supply is difficult to expand rapidly in the short term, price adjustment has become a means to balance supply and demand.
In the era of tight capacity equilibrium, making capacity plans in advance and having multiple supply chain contingencies have become critical imperatives that chip companies must prioritize. The computing power race is not just about chip performance; the stability of supply chain costs and capacity equally determines a company's long-term competitiveness.