Where chips go
The machines a chip was made for.
Data centres and servers
Data centres and cloud servers use the largest and most power-hungry chips: server CPUs with up to a few hundred cores, AI accelerators drawing 700 W to over 1,000 W each, network chips and DPUs, and huge amounts of memory. Intel's Xeon and AMD's Epyc dominate general-purpose servers, while Arm designs such as Amazon's Graviton have grown fast
PCs and laptops
Personal computers and laptops have been built around x86 processors since IBM chose the Intel 8088 for its PC in 1981. Intel and AMD have competed for decades, from the 486 and Pentium to Athlon, Core and Ryzen
Phones and tablets
Phones and tablets are built around a system on a chip (SoC) that combines processor cores, graphics, an AI engine, camera and video processing, and often the modem. Around it sit memory, storage, power management, radio and sensor chips, many dozens in all
Embedded and industrial
Embedded chips run machines rather than people's programs: the controller inside a washing machine, a lift, a printer, a medical pump, a factory robot or a power meter. Most are microcontrollers, small chips with processor, memory and input/output on one die
Cars
A modern car contains hundreds and often more than a thousand chips: microcontrollers for the engine, brakes, airbags and doors, power chips for the motors and battery of an electric car, sensors, radar and camera chips, and processors for the dashboard and driver assistance. Car chips must survive heat, cold and vibration for 15 years or more, and meet strict safety standards.
Games consoles
Games consoles have driven chip design since the 1970s. Each console uses a fixed set of chips for years, made in tens of millions, so its makers can afford custom designs
Supercomputers
Supercomputers join thousands of chips to solve science and engineering problems: weather and climate, nuclear physics, drug discovery and, more and more, AI. The TOP500 list, published twice a year since 1993, ranks them by speed on a standard maths benchmark
AI training and inference
AI chips train neural networks and run them, a task called inference. Training a large language model takes thousands of accelerators running for weeks; inference runs every time someone uses the model
Internet of things
Internet of things (IoT) chips connect small devices to networks: smart plugs, door sensors, wearables, trackers, meters and industrial monitors. They combine a low-power microcontroller with a radio for Wi-Fi, Bluetooth Low Energy, Zigbee, Thread or cellular networks such as NB-IoT
Networks and telecoms
Networks are built from chips: switch chips that forward packets in data centres and offices, router processors, network interface cards in servers, optical transceivers, and the processors and radio chips of mobile base stations. Broadcom, Marvell, Nvidia, Cisco, Intel and Qualcomm are among the main makers
Consumer electronics
Consumer electronics were the first mass market for chips. Pocket calculators drove early chip design: the Intel 4004 was built for a Busicom calculator, and Texas Instruments' TMS1000 went into calculators, toys and microwave ovens
Space and defence
Chips for space and defence must survive radiation, extreme temperatures and vibration, and keep working for years where no one can repair them. Radiation can flip bits, a single-event upset, or slowly damage transistors over time