Summary of Key Points in Plain Language
The craze for AI graphics cards and GPUs over the past two years has spread all the way to the very upstream of chip manufacturing. All the global wafer foundries, whether producing the most advanced AI chips or the “older” chips used in home appliances and automobiles, are now raising prices, with the highest increases reaching 25%. This price surge is not a short-term speculation; it is a result of a mismatch in supply and demand across the entire AI industry chain, which is expected to continue at least until 2027. The long-standing logic of “excess capacity and price cuts to boost sales” in the semiconductor industry over the past decade has been completely overturned.
---
Detailed Explanation of Each Point
1. How explicit are these price increases?
The wafer foundries are no longer hiding the price increases; they are now pricing differently based on supply and demand:
- Samsung’s pricing policy is quite straightforward: for new orders using its 4nm advanced process, the price increase is a maximum of 15%. However, customers from the Chinese mainland have to pay an additional 10%-15%. Customers from Taiwan, being closer to TSMC and having alternative production capabilities, only pay a 5%-10% increase. In essence, those with fewer options bear the higher cost of the price hike.
- TSMC’s policy is even more “market-oriented”: all process types will see price increases in 2027, with a basic increase of 5%-10% for processes below 7nm. If you want to secure additional production capacity for AI chips beyond your existing long-term contracts, you have to pay an additional 10%-15% as a “queue-jumping fee.” The combined increase can reach up to 25%, similar to going to a popular restaurant where not only do the dishes cost more, but you also have to pay extra to skip the queue and get served immediately.
- Intel, which was previously marginalized by TSMC and Samsung, has become highly sought after. Companies like Nvidia, OpenAI, Apple, and Microsoft are all placing new orders with Intel. Even SK Hynix, which previously relied 100% on TSMC for memory production, is now splitting its orders with Intel to ensure it doesn’t run out of capacity and miss production deadlines.
2. Everyone was surprised: older, decades-old chip processes are in even greater demand than advanced ones
Many expected that only the latest 4nm and 3nm processes used in AI would see price increases, but it turns out that older processes like 28nm and 45nm, which were mass-produced over a decade ago, are experiencing even more significant price hikes. This is due to several reasons:
- First, production capacity has been depleted: Leading companies like TSMC and Samsung have converted their older production lines for older processes into advanced lines for AI chips, significantly reducing the capacity for those older processes. The remaining orders have shifted to second- and third-tier foundries, reversing the supply-demand situation.
- Second, the demand for AI chips is massive: An AI graphics card cannot function on its own; it requires a series of power and signal chips. A server cabinet that can hold 72 AI graphics cards requires over 16,000 power chips, all of which are made using older processes. Each additional AI chip sold drives demand for dozens of older chips.
- Third, it’s difficult to quickly expand production for older processes: These processes used in automotive and industrial chips cannot be easily scaled up. Customers need to undergo safety certifications before supplying to car manufacturers, and it takes at least two to three years to build and optimize new production lines, which is not enough to meet the current demand.
3. The biggest beneficiaries of the price increases: Wafer foundries are seeing profits soar
Global wafer foundry capacity is mostly booked for the second half of 2027, with leading companies reaping huge profits:
- SMIC’s revenue in the second quarter exceeded $3 billion for the first time, and its net profit increased by 261.7% year-over-year, more than tripling. Hua Hong Hong Li’s capacity utilization rate reached 102.8%. This means the factories are operating at full capacity, with workers working extra hours and using reserved maintenance capacity just to produce more silicon wafers. Previously, domestic wafer foundries were worried about insufficient orders; now, overseas customers are coming to them to secure production capacity for mature processes, pushing the industry’s prosperity to record levels.
4. The most affected: Chip design companies are caught in the middle
The pressure of the price increases falls on chip design companies, which cannot pass it on to their customers:
- Wafer foundries are openly raising their manufacturing fees, but downstream customers in the consumer electronics and automotive sectors are reluctant to accept these increases. Leading chip design companies like STMicroelectronics have already announced three price hikes in 2026. Domestic analog chip companies have seen price increases of 10%-20%, with some high-end products seeing increases of up to 40%. Even with these increases, many companies’ profits are still being squeezed because the rise in manufacturing costs is higher than they can pass on to their customers. Industry leaders predict that prices will not fall until 2027 at the earliest. Many small chip design companies that did not secure long-term production agreements with wafer foundries are now struggling to find production capacity and may be forced out of the market.
5. Who will bear the cost? Consumers will pay the price
The overall cost of these price increases will eventually be passed on to consumers:
- The prices of AI-powered PCs and flagship smartphones are likely to rise by several hundred yuan next year, as the cost of AI chips and related components increases. The price war in the new energy vehicle sector will also be affected, as the cost of automotive-grade chips is rising, making it harder for manufacturers to reduce prices. Even for small items like smart door locks and home appliances, a few more dollars for power chips will result in a several-dollar increase in the final product price. In essence, all consumers are indirectly paying for this global AI-driven demand boom.