NederlandsChip
A chip is a small piece of silicon with billions of tiny switches, the transistors, on which a computer calculates and remembers. AI chips are built to do the huge amount of computation behind AI models quickly.
A chip, or integrated circuit, is a small slice of silicon carrying billions of minuscule switches: transistors. By switching them on and off, a computer calculates and remembers data. Every phone, car and server contains dozens of chips. You can think of a chip as a city on a postage stamp, with streets, junctions and storerooms that all work together.
How does it work?
Chips are made by printing patterns with light onto a wafer of silicon, layer by layer. Sizes are expressed in nanometres, billionths of a metre. The smaller the components, the more fit in the same space. That usually makes a chip faster and more economical, although it also depends on the design.
Running an AI model means a great deal of computation that can be done at the same time. That is why data centres use chips that are good at this, such as Nvidia's graphics chips. In addition, companies design their own AI chips. A chip also needs memory, and the closer that sits to the computation, the faster things go.
Why it matters
Chips determine how fast and how cheap AI can become. At a major data centre trade show, the American company d-Matrix showed that an AI model answers faster when ordinary graphics chips and its own chips share the work, each doing the part it is good at (link).
It is also about who can make the chips. Japan gave TSMC the go-ahead to make 3-nanometre AI chips in Kumamoto from the end of 2028, the first time in that country (link). Meanwhile China is showing how far it has got with its own chips: there, the company Sugon linked 100,000 domestically made computing cards into a computing cluster (link). Reports like these show that chips are not just technology, but also decide where computing power for AI becomes available.
Articles about Chip
-
Laser flash turns plastic into trillions of tiny diamonds
German physicists fire a powerful laser at a thin sheet of the plastic that soft-drink bottles are made of. Each flash produces about 10 trillion tiny diamonds averaging 3 millionths of a millimetre, almost all the same size. The next step is making enough of them to work with.
-
A different order makes AI servers start answering more than three times faster
Engineers from Amazon and Mistral have made the open software vLLM, which many companies use to run AI models, work together more cleverly. With long questions, the first word of the answer appeared more than three times faster in their test. The new method is free to use.
-
Siemens factory in Erlangen teaches AI cameras to inspect with simulated photos, before a prototype exists
In a German electronics factory, Siemens and Nvidia let AI prepare the production of new circuit boards with a digital copy of the factory. According to Siemens, this saves up to 85 per cent of the preparation work. It is a first step towards a factory that adjusts itself.
-
Large open AI model from DeepSeek now fits on a computer with 128 gigabytes of memory
DeepSeek's new open AI model is too big for an ordinary computer. The company Unsloth made a heavily reduced version that runs on a powerful laptop or mini computer with 128 gigabytes of memory. This costs some accuracy.
-
Japan gives the green light: TSMC will make the most advanced AI chips in Kumamoto
The Japanese government has approved the new plan for TSMC's second chip factory in Kumamoto. From the end of 2028, 3-nanometre chips for AI will come off the line there, for the first time in Japan. The factory will, however, start smaller than first planned.
-
Nvidia simulates a 104-qubit quantum circuit on a European supercomputer, real errors included
Nvidia and the German research centre Jülich used the JUPITER supercomputer to simulate how a 104-qubit quantum computer makes an important building block, including the errors that real hardware makes. This lets designs be tested before they are built.
-
AI model Maxwell makes existing software faster while it runs
The Israeli company doubleAI presents Maxwell, an AI that rewrites programs while they run so that they go faster, without touching the source code. According to the company, one customer's computing costs fell by 67 per cent.
-
AI model Mercury 3 writes more than a thousand tokens per second by creating the whole text at once
The American company Inception has released Mercury 3, a language model that does not write word by word but brings a whole text into focus in a few steps. According to the company, this makes it four to eight times as fast as comparable models.
-
d-Matrix splits the work of an AI model across two kinds of chips for faster answers
At a major data centre conference, the American chip company d-Matrix showed how an AI model answers faster when ordinary graphics chips and its own chips each do the work they are good at. The company now also supplies complete computing clusters.
-
Chinese company Sugon shows computing cluster with 100,000 home-grown AI chips
China's Sugon is showing its Sugon 8000: a computing cluster that links 100,000 computing cards made in China. The system combines AI with classic supercomputing and, according to Sugon, has already run ten research projects.
-
Working AI server rack lets chips talk through light, with the optics right next to the chip
At a major data centre trade show in the US, three tech companies showed a working server rack in which chips talk through light, with the light components right next to the computing chip. That could make AI data centres much more energy-efficient. Large-scale production is next.
-
Microsoft finds a more economical recipe for training large AI models
Microsoft AI has completed the first training phase of its new base model MAI-Base-2. According to the company, its new approach reaches the same level as a strong Chinese model with a quarter to a twelfth of the computing. The model is now being trained further.
-
Brain-inspired chip keeps up with a million measurements per second from a fusion reactor
An American team tested a superconducting chip that works like a network of brain cells. It processed measurements from a fusion reactor as fast as they came in. Next year, a chip ten times larger will monitor the reactor itself.
-
Nvidia releases open AI models for digital assistants on servers and laptops
Nvidia has released three new open AI models for assistants that carry out tasks on their own. The largest runs in companies' data centres; the smaller ones simply work on a laptop. According to Nvidia, the large model beats top models on specialist tasks after extra training.
-
Printed brain-inspired chip dissolves in six minutes after use
Researchers in China have printed a flexible chip that, like our brain, calculates and remembers in the same place. It recognises handwritten digits almost as well as ordinary software and dissolves in six minutes after use.
