Where Chips Go

“A chip” isn’t one thing — it’s a category covering wildly different jobs, from crunching video game graphics to sensing whether your car’s airbag should fire. The device that makes your phone smart contains a dozen or more very different types of chip working together, each doing a job the others can’t do as well. This page connects the chip types you’ve read about elsewhere in this site to the real products they end up inside.

Matching chip types to jobs

CPUs (Central Processing Units) are general-purpose computing chips, built to handle a huge variety of different instructions quickly, one (or a modest number of parallel) at a time. They’re the “do anything, reasonably fast” chip — the main processor in your laptop, phone, or game console runs a CPU core to handle the operating system, apps, and general logic.

GPUs (Graphics Processing Units) are built the opposite way: instead of a few powerful cores that handle any instruction, a GPU packs thousands of simpler cores that all do the same simple operation on different pieces of data at once. That massive parallelism was originally built for rendering millions of pixels for video game graphics, but the same structure turned out to be extremely good at the matrix math behind modern AI — which is why GPUs power both the graphics in your games and the training of large AI models.

Memory chips — DRAM and NAND flash — store data rather than compute with it. DRAM is fast but forgets everything when power is removed, so it holds whatever your device is actively working on right now (open apps, in-progress calculations). NAND flash is slower but keeps data without power, so it holds your photos, apps, and files even when the device is switched off.

Microcontrollers are small, self-contained, low-power chips that combine a simple processor with memory and input/output circuitry on one chip. They’re built for control tasks rather than raw computing power — running the same simple program forever, reliably and cheaply. They’re everywhere you don’t expect a “computer” to be: inside a washing machine’s control panel, a car’s window motor, a thermostat, or a toy.

Sensors convert something physical — light, pressure, acceleration, temperature, sound — into an electrical signal a chip can process. Your phone’s camera sensor turns light into pixel data; its accelerometer senses motion to know when to rotate the screen; a car’s tire-pressure sensor watches for a slow leak.

RF (radio frequency) and analog chips handle signals that are continuously varying rather than clean digital 0s and 1s — radio waves, audio, power regulation. Your phone’s cellular modem, Wi-Fi radio, and Bluetooth chip are all RF chips translating between the analog world of radio waves and the digital world of bits your CPU can process.

🧩 Think of it like… a kitchen full of specialized staff instead of one person doing everything. The head chef (CPU) makes broad decisions and handles varied tasks. A huge line of prep cooks (GPU) all chop vegetables at once — slow individually, but incredibly fast together on repetitive work. The pantry and fridge (memory) store ingredients, some for quick reach (DRAM) and some for long-term storage (NAND flash). A dishwasher running the same cycle forever (microcontroller) doesn't need a head chef's flexibility. Someone tasting the soup (sensor) reports back what they perceive. And a waiter relaying orders between the dining room and kitchen (RF/analog chip) translates between two different "languages." A restaurant needs all of these roles — and so does your phone.

Tracing chips through a real device

    flowchart LR
    subgraph Phone["Inside a smartphone"]
    CPU["CPU<br/>runs the OS & apps"]
    GPU["GPU<br/>renders the screen"]
    DRAM["DRAM<br/>active app memory"]
    NAND["NAND flash<br/>photos & files"]
    Sensor["Sensors<br/>camera, accelerometer"]
    RF["RF chip<br/>Wi-Fi, cellular, Bluetooth"]
    MCU["Microcontroller<br/>battery/power management"]
    end
    Sensor --> CPU
    CPU <--> DRAM
    CPU --> NAND
    CPU <--> GPU
    CPU <--> RF
    MCU -.controls power.-> CPU

A car tells a similar story at a larger scale: dozens to over a hundred microcontrollers manage individual systems (windows, mirrors, brakes, infotainment), RF chips handle key fobs and wireless connectivity, sensors watch everything from wheel speed to airbag deployment conditions, and — increasingly, in vehicles with advanced driver-assist or self-driving features — GPUs or dedicated AI accelerator chips process camera and radar data in something close to real time.

“ASIC” (Application-Specific Integrated Circuit) is the general term for a chip custom-designed for one narrow job rather than general-purpose use. A CPU and a GPU are both technically general-purpose in different ways; a chip built to do exactly one company’s exact AI workload and nothing else, or a chip built only to manage a car’s battery, is more precisely called an ASIC.

Why specialization wins

You could, in principle, use a powerful general-purpose CPU to do everything — process camera data, manage battery charging, render graphics. But a chip optimized for one specific job is almost always faster, cheaper, and far more power-efficient at that job than a general-purpose chip pressed into service for it. A microcontroller running a washing machine’s cycle doesn’t need billions of transistors; it needs to be cheap, reliable, and sip almost no power. A GPU doesn’t need to be good at running an operating system; it needs to do the same math operation across enormous amounts of data as fast as possible. This is the same economic logic that shapes the foundry-fabless-idm split — specialization, done well, beats generalism at almost every layer of the industry.

Key takeaways

CPUs handle general-purpose computing; GPUs handle massively parallel workloads like graphics and AI.
DRAM stores active, in-use data; NAND flash stores data permanently without power.
Microcontrollers are small, cheap, low-power chips built for dedicated control tasks in appliances, cars, and toys.
Sensors convert physical signals (light, motion, pressure) into data a chip can process; RF/analog chips bridge the digital and analog worlds for radios and audio.
A single device — a phone, a car — typically contains many different chip types, each specialized for a job the others do less efficiently.
See DRAM and NAND Flash for how memory chips physically work, and Major Players Map for which companies design these chip types.

Major Players Map The Global Supply Chain