What Is an Integrated Circuit? IC vs Chip vs Die vs SoC
An integrated circuit (IC) is an electronic circuit with many transistors, wires and other parts built together into one small piece of silicon. Compared with connecting separate parts one by one, it is smaller, cheaper, faster and more reliable. "Chip" and "LSI" mean almost the same thing. An SoC is an IC that puts a device's main functions on one chip.
What you'll learn
- What "integration" brings together
- Why the IC was a breakthrough
- How IC, chip, die, LSI, SoC and package differ
- The types of ICs, and how ICs are designed
- How ICs connect to a device's circuit board
Prerequisites: none. It helps to know that a transistor is a small switch flipped by electricity ("What Is a Transistor?").
In everyday life: the black parts inside a phone
Inside a phone is a board that holds parts and connects them with wiring. It is called a circuit board. On it sit several square parts, black or silver. Most of them are ICs, or integrated circuits.
The repair site iFixit has examined the parts on the board of the iPhone 15 Pro Max (2023) and published the results. The board holds several ICs, each with a different job.
- The chip at the center of computing (Apple's A17 Pro)
- Memory that temporarily holds the data being worked on
- Storage memory that keeps photos and apps
- ICs that turn battery power into the right voltage for each part (there are several)
TVs, refrigerators and cars also contain many ICs.
What "integrated" means: making parts and wiring together
An electronic circuit is made by connecting many parts with wires. A typical part is the transistor, a small switch flipped by electricity. In the past, each part was made separately, placed on a board, and joined with solder (a metal that melts at a low temperature).
Other parts include resistors and capacitors. A resistor makes it harder for current to flow, and a capacitor stores a little electric charge. Both are still sold as separate parts.
An integrated circuit makes these parts and their wiring together, inside one piece of silicon. Transistors, resistors and capacitors are all shaped by stacking layers on silicon in the same process. The metal wiring that connects them is built in the same flow. "Integration" means gathering parts and wiring in one place and making them as one unit.
Billions of transistors are not made one by one, either. They are all made at once, by printing patterns with light. "How Are Chips Made?" explains the process.
Note that the figure is simplified. In a real IC, almost all the parts are transistors. Resistors and capacitors with large values take up a lot of space. So they are often placed outside the IC, as separate parts.
Why the IC was a breakthrough
Before the IC, a big headache was the number of parts. Making a computer more powerful meant more and more parts and joints. If even one joint came loose, the machine would not work correctly.
An extreme example is ENIAC, a US computer unveiled in 1946. It was built before transistors came into wide use. It used more than 17,000 vacuum tubes to act as switches. According to the University of Michigan, it had about 5 million soldered joints.
The integrated circuit solved this problem. Jack Kilby in 1958 and Robert Noyce in 1959 showed its basic forms. The timeline in "What Is a Transistor?" covers this history. Making everything together brought four main benefits.
- More reliable: there are fewer solder joints between parts, so fewer places can come loose.
- Cheaper: instead of being assembled, everything is made together. Hundreds of ICs come from one wafer (a round plate of silicon) at the same time.
- Faster: parts sit closer together. Shorter wires store less charge, so signals can switch faster.
- Smaller, and lower in power: the same circuit fits on a fingertip, and it uses less power.
In 1965, Gordon Moore, who later co-founded Intel, wrote a paper. It said that ICs were reliable and were lowering the cost of whole systems. The same paper noted that the number of components was roughly doubling every year. This later became known as Moore's law.
IC vs. chip vs. die vs. LSI vs. SoC
Many similar words surround the IC. Sometimes they are just different names for the same thing in different settings. Let's sort them out with a figure and a table.
Chip and die: the piece of silicon inside
The body of an IC is a piece of silicon cut from a wafer. Most are a few millimeters to about 2 cm on a side. This piece is called a chip. In manufacturing and design, the same thing is also called a die.
In everyday speech, though, "chip" often means the whole packaged part. News terms such as "AI chip" use this broader meaning.
Package: the part you can see
A die is very thin, breaks easily and is sensitive to moisture. So it is mounted on a base, given terminals that connect to the outside, and covered in resin or a similar material. The whole part, including this enclosure, is called a package. The black squares you see on a board are usually the resin of these packages.
LSI and VLSI: names based on size
ICs have been named by how many components they hold. A large IC with many components is called an LSI (large-scale integration). One with even more is called a VLSI (very-large-scale integration).
There is no fixed definition of where the lines fall. For example, one university's lecture notes count 1,000 or more components as LSI. Today, being large-scale is normal, so IC and LSI are used to mean almost the same thing.
SoC: a device's main functions on one chip
The chip at the center of computing in a phone holds a CPU. It also holds a circuit that draws images on the screen (a GPU) and circuits suited to AI calculations. An IC that brings together a device's main functions like this is called an SoC (system on chip). In Japan, it is also called a "system LSI." "What Is a CPU?" explains what a CPU does.
| Term | What it means | Where you often see it |
|---|---|---|
| IC (integrated circuit) | A circuit with parts and wiring built into one piece of silicon. When sold, it is a packaged part | The general term |
| Chip | The piece of silicon at the heart of an IC. In everyday speech, also the whole part | News, everyday speech |
| Die | The same piece of silicon as a chip | Manufacturing and design |
| LSI, VLSI | A large-scale IC. Today almost the same as IC | Industry in Japan, textbooks |
| SoC | One chip that combines a device's main functions | Phones, PCs |
| Package | The enclosure that holds the die and connects it to the outside (or the whole part including it) | Assembly, mounting on boards |
Types of ICs, by job
ICs can be roughly grouped by the job they do. Here we look at four groups, with familiar examples.
- Computing (logic): ICs that combine signals of 0 and 1 to calculate and make decisions. CPUs, GPUs and phone SoCs belong here.
- Remembering (memory): ICs that store data as 0s and 1s. DRAM holds the data being worked on. It is fast, but loses its contents when the power is turned off. NAND flash stores photos and the like, and keeps its data without power.
- Handling smooth signals (analog): ICs that handle signals that change continuously, such as sound, radio waves and temperature. They make small signals bigger, or turn them into data made of 0s and 1s.
- Managing power (power management): ICs that make the right voltage for each part from a battery or other source, and deliver it. They are called PMICs (power management ICs), and a phone contains several.
This grouping is simplified. WSTS, a group that compiles worldwide semiconductor shipments, divides ICs into "analog," "micro," "logic" and "memory" in its statistics. CPUs and microcontrollers (a small computer on one IC) go under "micro." Power management ICs mostly go under "analog." Sensors, such as camera image sensors, and parts that handle light are counted separately from ICs.
Worldwide semiconductor sales in 2025 were $791.7 billion. The Semiconductor Industry Association (SIA), a US industry group, announced this figure in February 2026, based on WSTS data. By category, logic was $301.9 billion and memory was $223.1 billion. Together, they make up about two-thirds of the total.
ICs can also be divided another way. Some are general-purpose ICs that many companies buy and use. Others are designed only for one specific product. These are called ASICs (application-specific integrated circuits). The better a product sells, the more it pays to make a dedicated chip.
How an IC is made: design comes first
Before an IC is made, its blueprint is created on a computer. No one can draw billions of transistors by hand, one by one. So designers rely on special software called EDA (electronic design automation). The rough flow is as follows.
- Set the specification: decide what the IC will do, and set targets for speed and power.
- Describe the behavior: write how the circuit should behave, in a special language similar to a programming language (a hardware description language).
- Turn it into a circuit: software converts the described behavior into a network of small circuits that process 0s and 1s (logic gates). This step is called logic synthesis.
- Place and connect: arrange the parts on the chip and decide the routes of the wires. This step is called place and route.
- Check: at each stage, run the circuit on a computer (simulation) to check that it works correctly and at the target speed.
- Hand off to manufacturing: send the finished design data to the manufacturer. This step is called tape-out.
From the design data, the manufacturer makes master plates (masks) that carry the pattern of each layer. Then, in the fab, the circuits are built, tested, cut apart and packaged. This flow inside the fab is explained in "How Are Chips Made?", mentioned earlier.
The company that designs a chip and the one that makes it are often different. "IDM vs. fabless vs. foundry" compares these roles.
Connecting the IC to the outside world
On the die's surface are rows of small electrodes for exchanging signals and power with the outside. The die connects to the package's base in one of two ways. One uses fine metal wires. The other flips the die over and joins it directly with tiny metal bumps. "Wire bonding vs. flip chip" compares the two methods.
The package connects to the device's board through its outer terminals. The traditional form has metal legs sticking out from the sides of the package. The QFP, with legs on all four sides, is still widely used. For example, it is used for microcontrollers that control home appliances.
ICs with many terminals have small balls of solder laid out in a grid across the whole underside of the package. This form is called a BGA. It uses the whole surface, not just the edges, so it can fit many terminals. When it is placed on a board and heated, the balls melt and connect to the board.
The number of terminals varies widely from IC to IC. Intel's 4004 of 1971 had 16 pins. Intel's desktop CPUs (12th to 14th generation) have 1,700 flat contacts on the underside (a form called LGA1700). "BGA vs. QFP vs. QFN" compares package types.
Common misconceptions
Myth: An IC means a CPU
In fact, a CPU is just one kind of IC. A phone also contains several ICs with other jobs, such as memory, power management, and handling sound and radio signals. In sales statistics too, "micro," which includes CPUs, is just one of four IC categories. The myth arises because CPUs and AI chips get the most attention in the news.
Myth: The black square on a board is the chip itself
In fact, what looks black is usually the resin of the package around the die. The die is inside, and it is normally smaller than the package. For example, the 4004's die was about 3 mm × 4 mm, housed in a long, narrow package with 16 pins. The myth arises because "chip" is also used for the whole part.
Myth: More transistors always make a better IC
In fact, performance depends on what the transistors are used for and how they are combined. Adding more with the same process makes the die bigger, so fewer dies fit on one wafer. The share of good dies (the yield) also tends to drop, and price and power tend to rise. For some uses, such as the microcontroller in a rice cooker, being small and cheap matters most. The myth arises because new chips are often introduced by their transistor counts.
By the numbers
| Number | What it is |
|---|---|
| About 5 million | Soldered joints in ENIAC (1946), according to the University of Michigan |
| 1,000 or more | One example of the component count for an IC to be called an LSI. There is no fixed definition |
| About 3 mm × 4 mm | Die size of the Intel 4004 (1971). It had 16 pins |
| 1,700 | Contacts on the package of Intel's desktop CPUs (12th to 14th generation) |
| About two-thirds | Share of logic and memory combined in worldwide semiconductor sales in 2025 (SIA, February 2026) |
Learn more
- Compare in tables: ASIC vs. FPGA vs. SoC, DRAM vs. NAND flash, Front-end vs. back-end process
- Try the numbers: with the die per wafer and yield calculator, see how a bigger die changes the number of dies per wafer and the share of good ones.
- Try circuits of 0s and 1s: with the logic circuit simulator, combine the logic gates found inside ICs.
- Read next: "What Does 2nm or 3nm Mean?" explains the generation numbers used as transistors shrink.