Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/tsmc-price-hikes-end-the-era-of-cheap-transistors.23731/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2030770
            [XFI] => 1060170
        )

    [wordpress] => /var/www/html
)

TSMC Price Hikes End the Era of Cheap Transistors

Fred Chen

Moderator
By Pablo Valerio 10.01.2025

The global semiconductor industry is undergoing a profound economic transformation, one anchored by Taiwan Semiconductor Manufacturing Company (TSMC) that spells the end of an era defined by predictably declining costs of transistors.

At the center of this structural shift is TSMC’s decision to implement unprecedented price hikes for its most advanced logic chips, a move necessitated by astronomical capital expenditures, geopolitical mandates, and the sheer, unyielding physics of manufacturing at the angstrom scale.

TSMC, the world’s undisputed leader in advanced logic manufacturing and the holder of a commanding 70.2% share of all foundry revenue as of the second quarter of 2025, is leveraging its technological supremacy to fund the next generation of innovation. This strategy permanently raises the cost basis for the foundational components of the entire digital economy.

Decoupling Moore’s Law​

For decades, Moore’s Law promised that devices would become exponentially more powerful while simultaneously becoming more affordable due to declining cost-per-transistor. That principle has now reached an inflection point.

Soon, according to media reports, TSMC will implement price increases of 5-10% for its advanced nodes below 5nm starting in 2026. However, the most strategically significant adjustment will be for the generational leap to the 2-nanometer (2nm) node.

Wafers produced on the 2nm node would see prices surge by over 50% compared to their predecessors. The current cost of a 300mm wafer on the 3nm process is approximately $20,000. A 50% increase will push the price of a single 2nm wafer to an unprecedented $30,000 or more.

This dramatic escalation means that the cost of manufacturing is now rising faster than the economic benefits of density scaling alone can offset.

For the first time in a major node transition, the cost per transistor will rise. This structural shift signals to the entire industry that access to the pinnacle of semiconductor technology is no longer a commodity but a premium, non-negotiable service.

Price of sovereignty​

A primary catalyst for TSMC’s rising cost structure is the immense capital expenditure required for global diversification, a strategic pivot heavily influenced by geopolitical pressures. The company’s total planned expenditure in its Arizona, USA facilities has swelled to an astonishing $165 billion, representing the largest single foreign direct investment in U.S. history.

These overseas facilities, however, operate at a significant cost premium compared to TSMC’s optimized fabs in Taiwan.

AMD CEO Lisa Su has publicly confirmed that chips produced at the Arizona fab cost between 5% and 20% more than those made at their Taiwanese counterparts. Other industry reports suggest this premium could be as high as 30% for 4nm production in Arizona.

TSMC Transistors

Semiconductor Wafer Cost Evolution by Process Node (2020-2026E)

TSMC has acknowledged that its overseas fabs will initially dilute its consolidated gross margin by 2-3%. To protect its strategic financial target of maintaining a gross margin of 53% or higher, TSMC views price increases not as an option, but as an “unavoidable” necessity to offset these higher operational and geopolitical costs.

By implementing a clear price premium for its U.S.-made wafers, TSMC is effectively making its customers—and, by extension, U.S. consumers—co-investors in the strategic goal of diversifying its supply chain.

Pushing the atomic wall with GAA​

Beyond geopolitical factors, the second major driver of the price surge is the staggering technical complexity required to stay on the leading edge. The industry is pushing the boundaries of physics at the angstrom scale with technologies like Gate-All-Around (GAA) transistors.

The leap from the 3nm to the 2nm node requires one of the most significant changes in transistor design in over a decade: the transition from the older FinFET architecture to the GAA architecture.

TSMC Transistors

GAA transistors use stacked horizontal “nanosheets” of silicon surrounded by the gate, which is essential for scaling beyond 3nm. However, the fabrication of GAA transistors is “an order of magnitude more complex” than FinFETs, involving multi-step processes like the selective etching of sacrificial layers, which introduces numerous new potential failure modes and higher development costs.

The capital expenditure for the necessary tools has soared. A single state-of-the-art facility can cost between $15 billion and $20 billion. The most critical piece of equipment, the Extreme Ultraviolet (EUV) lithography scanner, costs approximately $350 million per unit.

Furthermore, the industry is battling stochastic defects—random, probabilistic errors inherent to EUV physics—which represent a fundamental barrier to scaling and ensure that future progress will come at a structurally higher cost. This ongoing battle against physics is now structurally embedded into the foundry business model.

Customer reactions​

TSMC’s pricing strategy is reshaping the technology landscape, forcing its largest customers to react based on their market position.

Nvidia, whose high-margin AI accelerators dominate the market, has publicly endorsed the price hikes. CEO Jensen Huang “fully supports” TSMC charging more for its services, arguing that the company’s value is not fully reflected in its current pricing. “The value of what TSMC produces is very high… a price increase is natural and consistent with the value they deliver,” Huang said.

He has even gone so far as to call TSMC “one of the greatest companies in the history of humanity”. For Nvidia, whose premium-priced chips mean the silicon cost is a relatively small component of the final selling price of a multi-thousand-dollar AI accelerator, securing priority access to 2nm and the future 1.6nm (A16) nodes is paramount, outweighing price sensitivity.

Apple, TSMC’s single largest customer, faces a complex challenge of rising wafer costs and geopolitical tariffs. “While we don’t comment on supplier discussions, securing access to the most advanced process technology is paramount for our product roadmap,” Apple said in a statement. As the largest customer, receives preferential treatment and likely secures more favorable terms than others.

In a strategic move to secure exemptions from additional semiconductor. tariffs—which are expected at rates as high as 100%—Apple has pledged a monumental $600 billion investment in U.S. manufacturing. Despite this proactive hedging, Apple has already revealed that it incurred $800 million in tariff-related costs in Q3 2025 and projected this figure would rise to $1.1 billion in the following quarter.

Meanwhile, Qualcomm and MediaTek, key players in the competitive Android smartphone ecosystem, face a direct margin squeeze. Both are seeing significant cost increases for their flagship chips on the N3P process, with MediaTek reportedly facing a 24% cost increase and Qualcomm a 16% rise.

Qualcomm CEO Cristiano Amon, while keen to diversify, has publicly stated that while the company would “like Intel to be an option,” the U.S. chipmaker’s foundry technology is “not an option today” for mobile chips. This lack of a credible, high-yield alternative to TSMC solidifies the foundry’s leverage.

Impact on consumer prices and data centers​

The new cost structure will have a trickle-down effect across the digital economy. For consumers, the margin squeeze faced by Android chip giants like Qualcomm and MediaTek will “inevitably translate to higher prices for flagship consumer devices” starting in 2026. The era of progressively cheaper smartphones and personal computers for top-tier products is over.

In the data center space, while Nvidia’s high margins insulate it, the $30,000+ cost of a 2nm wafer establishes a significantly higher price floor for all future AI and high-performance computing components.

The financial pressure is also accelerating the industry’s shift toward chiplet-based architectures. With the cost differential widening, companies like AMD are demonstrating that using older, more cost-effective processes for components while reserving the expensive 2nm process only for performance-critical logic is transitioning from a clever engineering choice to an economic necessity.

In conclusion, TSMC is leveraging its technological dominance and market power to implement a new pricing paradigm that ensures its financial stability and supports the challenging work of advancing technology.

This move fundamentally alters the business models of fabless semiconductor companies and their end customers, ushering in the era of premium, high-value silicon.

 
Last edited:
By Pablo Valerio 10.01.2025

The global semiconductor industry is undergoing a profound economic transformation, one anchored by Taiwan Semiconductor Manufacturing Company (TSMC) that spells the end of an era defined by predictably declining costs of transistors.

At the center of this structural shift is TSMC’s decision to implement unprecedented price hikes for its most advanced logic chips, a move necessitated by astronomical capital expenditures, geopolitical mandates, and the sheer, unyielding physics of manufacturing at the angstrom scale.

TSMC, the world’s undisputed leader in advanced logic manufacturing and the holder of a commanding 70.2% share of all foundry revenue as of the second quarter of 2025, is leveraging its technological supremacy to fund the next generation of innovation. This strategy permanently raises the cost basis for the foundational components of the entire digital economy.

Decoupling Moore’s Law​

For decades, Moore’s Law promised that devices would become exponentially more powerful while simultaneously becoming more affordable due to declining cost-per-transistor. That principle has now reached an inflection point.

Soon, according to media reports, TSMC will implement price increases of 5-10% for its advanced nodes below 5nm starting in 2026. However, the most strategically significant adjustment will be for the generational leap to the 2-nanometer (2nm) node.

Wafers produced on the 2nm node would see prices surge by over 50% compared to their predecessors. The current cost of a 300mm wafer on the 3nm process is approximately $20,000. A 50% increase will push the price of a single 2nm wafer to an unprecedented $30,000 or more.

This dramatic escalation means that the cost of manufacturing is now rising faster than the economic benefits of density scaling alone can offset.

For the first time in a major node transition, the cost per transistor will rise. This structural shift signals to the entire industry that access to the pinnacle of semiconductor technology is no longer a commodity but a premium, non-negotiable service.

Price of sovereignty​

A primary catalyst for TSMC’s rising cost structure is the immense capital expenditure required for global diversification, a strategic pivot heavily influenced by geopolitical pressures. The company’s total planned expenditure in its Arizona, USA facilities has swelled to an astonishing $165 billion, representing the largest single foreign direct investment in U.S. history.

These overseas facilities, however, operate at a significant cost premium compared to TSMC’s optimized fabs in Taiwan.

AMD CEO Lisa Su has publicly confirmed that chips produced at the Arizona fab cost between 5% and 20% more than those made at their Taiwanese counterparts. Other industry reports suggest this premium could be as high as 30% for 4nm production in Arizona.

TSMC Transistors

Semiconductor Wafer Cost Evolution by Process Node (2020-2026E)
TSMC has acknowledged that its overseas fabs will initially dilute its consolidated gross margin by 2-3%. To protect its strategic financial target of maintaining a gross margin of 53% or higher, TSMC views price increases not as an option, but as an “unavoidable” necessity to offset these higher operational and geopolitical costs.

By implementing a clear price premium for its U.S.-made wafers, TSMC is effectively making its customers—and, by extension, U.S. consumers—co-investors in the strategic goal of diversifying its supply chain.

Pushing the atomic wall with GAA​

Beyond geopolitical factors, the second major driver of the price surge is the staggering technical complexity required to stay on the leading edge. The industry is pushing the boundaries of physics at the angstrom scale with technologies like Gate-All-Around (GAA) transistors.

The leap from the 3nm to the 2nm node requires one of the most significant changes in transistor design in over a decade: the transition from the older FinFET architecture to the GAA architecture.

TSMC Transistors

GAA transistors use stacked horizontal “nanosheets” of silicon surrounded by the gate, which is essential for scaling beyond 3nm. However, the fabrication of GAA transistors is “an order of magnitude more complex” than FinFETs, involving multi-step processes like the selective etching of sacrificial layers, which introduces numerous new potential failure modes and higher development costs.

The capital expenditure for the necessary tools has soared. A single state-of-the-art facility can cost between $15 billion and $20 billion. The most critical piece of equipment, the Extreme Ultraviolet (EUV) lithography scanner, costs approximately $350 million per unit.

Furthermore, the industry is battling stochastic defects—random, probabilistic errors inherent to EUV physics—which represent a fundamental barrier to scaling and ensure that future progress will come at a structurally higher cost. This ongoing battle against physics is now structurally embedded into the foundry business model.

Customer reactions​

TSMC’s pricing strategy is reshaping the technology landscape, forcing its largest customers to react based on their market position.

Nvidia, whose high-margin AI accelerators dominate the market, has publicly endorsed the price hikes. CEO Jensen Huang “fully supports” TSMC charging more for its services, arguing that the company’s value is not fully reflected in its current pricing. “The value of what TSMC produces is very high… a price increase is natural and consistent with the value they deliver,” Huang said.

He has even gone so far as to call TSMC “one of the greatest companies in the history of humanity”. For Nvidia, whose premium-priced chips mean the silicon cost is a relatively small component of the final selling price of a multi-thousand-dollar AI accelerator, securing priority access to 2nm and the future 1.6nm (A16) nodes is paramount, outweighing price sensitivity.

Apple, TSMC’s single largest customer, faces a complex challenge of rising wafer costs and geopolitical tariffs. “While we don’t comment on supplier discussions, securing access to the most advanced process technology is paramount for our product roadmap,” Apple said in a statement. As the largest customer, receives preferential treatment and likely secures more favorable terms than others.

In a strategic move to secure exemptions from additional semiconductor. tariffs—which are expected at rates as high as 100%—Apple has pledged a monumental $600 billion investment in U.S. manufacturing. Despite this proactive hedging, Apple has already revealed that it incurred $800 million in tariff-related costs in Q3 2025 and projected this figure would rise to $1.1 billion in the following quarter.

Meanwhile, Qualcomm and MediaTek, key players in the competitive Android smartphone ecosystem, face a direct margin squeeze. Both are seeing significant cost increases for their flagship chips on the N3P process, with MediaTek reportedly facing a 24% cost increase and Qualcomm a 16% rise.

Qualcomm CEO Cristiano Amon, while keen to diversify, has publicly stated that while the company would “like Intel to be an option,” the U.S. chipmaker’s foundry technology is “not an option today” for mobile chips. This lack of a credible, high-yield alternative to TSMC solidifies the foundry’s leverage.

Impact on consumer prices and data centers​

The new cost structure will have a trickle-down effect across the digital economy. For consumers, the margin squeeze faced by Android chip giants like Qualcomm and MediaTek will “inevitably translate to higher prices for flagship consumer devices” starting in 2026. The era of progressively cheaper smartphones and personal computers for top-tier products is over.

In the data center space, while Nvidia’s high margins insulate it, the $30,000+ cost of a 2nm wafer establishes a significantly higher price floor for all future AI and high-performance computing components.

The financial pressure is also accelerating the industry’s shift toward chiplet-based architectures. With the cost differential widening, companies like AMD are demonstrating that using older, more cost-effective processes for components while reserving the expensive 2nm process only for performance-critical logic is transitioning from a clever engineering choice to an economic necessity.

In conclusion, TSMC is leveraging its technological dominance and market power to implement a new pricing paradigm that ensures its financial stability and supports the challenging work of advancing technology.

This move fundamentally alters the business models of fabless semiconductor companies and their end customers, ushering in the era of premium, high-value silicon.


It’s hard to compare the cost of two products from different nodes and generations because they vary in performance, design, memory, and features. But is there a way to estimate the average cost per transistor from various process nodes?

Are there any meaningful ways to measure the cost from different nodes?
 
It’s hard to compare the cost of two products from different nodes and generations because they vary in performance, design, memory, and features. But is there a way to estimate the average cost per transistor from various process nodes?

Are there any meaningful ways to measure the cost from different nodes?
Very rough estimates using the wafer price table in the posted article, and CGP x Cell height:
TSMC cost per transistor.png
 
Back
Top