How is the price of a chip defined?
Published time: 2019-12-20 11:15:10
The chip design industry is a typical high-input, high-yield industry, but it is also a very risky industry. In case the design of the chip fails to meet expectations, the huge investment will be overwhelmed. Let's talk about the cost structure of a chip and how the price of a chip is defined.
The cost of the chip includes the hardware cost of the chip and the design cost of the chip
Chip hardware cost
The chip hardware cost includes four parts: wafer cost + mask cost + packaging cost + test cost. A formula written is chip hardware cost = (wafer cost + mask cost + packaging cost + test cost) / final yield
Wafers are the raw materials for making chips. The cost of wafers can be understood as the cost of the materials (silicon wafers) used for each chip. When the output is large enough, in terms of 100 million units, the cost of wafers is the highest in hardware costs.
The mask cost is the cost of using different process technologies. For example, the 40 / 28nm process is very mature. The mask cost of the 40nm low-power process is $ 2 million; the 28nm SOI process is $ 4 million; the cost of 28nm HKMG is $ 6 million. But the most advanced process technology is sky-high. When the 14nm process was first put into production in 2014, the mask cost was 300 million U.S. dollars; and for the next 10nm process, according to Intel ’s official estimates, the mask cost required at least $ 1 billion.
The packaging cost is to stack the substrate, core, and heat sink together to make a chip that everyone commonly sees. The cost of this process is generally about 5% -25% of the cost of the hardware, but some of IBM's chip packaging costs account for about half of the total cost, it is said that the highest has reached 70%.
Test cost refers to testing the characteristics of each chip, such as maximum frequency, power consumption, heat generation, etc. And determine the cost of the chip level. For example, Intel divided a bunch of chips into categories: I5 4460, I5 4590, I5 4690, I5 4690K, and then offer different prices. However, the proportion of test costs is small, and if the chip output is large, it can even be ignored.
Chip design cost
The hardware cost is better and clearer, but the design cost is more complicated. This includes not only the salary of engineers, the cost of development tools such as EDA, equipment costs, site costs, etc., but also a large amount of intellectual property costs. The cost varies greatly. At present, the first few expenses are already at a stable level. The difference between the companies is not large. The major difference in the cost of chip design is intellectual property costs. The cost is much higher than Qualcomm.
The internationally common chip pricing strategy is the 8:20 pricing method, that is, when the hardware cost is 8, the pricing is 20, and Intel's general pricing strategy is 8:35. AMD's history has reached 8:50. A chip uses the 8:20 pricing method, with a production of 100,000, it sells for $ 305; with a production of 1 million, it sells for $ 75; with a production of 10 million , Which sells for $ 52.5. It can be seen that to reduce the price of the chip, production is crucial. If the chip is mass-produced in units of 100 million, such as Apple's chip, even if the mask cost is as high as 1 billion U.S. dollars, the cost will be 10 U.S. dollars. The mask cost of a chip is as high as $ 1,000, which is obviously not competitive.
Newer processes can bring lower chip prices and thus stronger market competitiveness. This is why giants such as Apple and Qualcomm use TSMC and Samsung's most advanced and expensive process technology, which can still make a lot of money. This is why IC design has the characteristics of winner-take-all.Tag: chip