From Sand to Silicon Wafers
At the beginning of every modern computer chip there is something that looks very unspectacular: sand. More exactly, factories use quartz sand that contains a lot of silicon dioxide. Silicon is a perfect material for chips because it behaves like a semiconductor. That means it sometimes conducts electricity and sometimes blocks it – exactly what we need for building tiny switches.
However, the silicon inside a chip must be almost perfectly pure. Normal sand contains many other elements that would disturb the electronics. In special high-temperature processes the silicon is separated, cleaned and melted. From this melt, a long, round crystal is pulled very slowly. This is called a silicon ingot. It can be more than a meter long, but the structure inside is extremely ordered.
The ingot is then sliced into very thin discs – the wafers. A typical wafer has a diameter of 300 mm and is polished until it is as smooth as a mirror. On this shiny surface, the future circuits of hundreds or thousands of chips will be created.
Photolithography – Drawing Tiny Circuits with Light
To build the circuits, the wafer is coated with a light-sensitive material called photoresist. A pattern of the circuit is projected onto the wafer using ultraviolet light and something like a very precise slide called a mask. Wherever the light hits, the properties of the photoresist change.
Afterwards the wafer is put into a chemical bath. Parts of the photoresist are washed away, other parts stay and protect the silicon underneath. In the next step, the unprotected silicon is etched away or doped with other atoms to change its electrical properties. In this way transistors, wires and insulating layers are formed.
This procedure is repeated many times with different masks and different chemicals. Modern high-end chips can require more than 100 separate processing steps. The structures that are created are unbelievably small: a single transistor on a 3-nanometre chip is much smaller than a virus. A finished chip can contain tens of billions of these switches.
Testing, Cutting and Packaging
When all layers are done, the wafer is tested while it is still in one piece. Tiny probes touch contact points on each chip and check whether it works as expected. Chips that already show errors are marked.
Then the wafer is cut into individual rectangular dies. Each die is placed into a package that protects it from damage and allows it to be soldered onto a circuit board. Very fine metal wires or microscopic bumps connect the chip inside the package to the visible metal contacts on the outside.
Finally, the finished chips are tested once again under different temperatures and loads. Only the ones that pass all tests are shipped to customers and built into devices. A single factory can produce millions of chips per month, but the whole process requires extremely clean rooms, expensive machines and a lot of know-how.
Where Computer Chips Work in Our Lives
If all computer chips stopped working at the same moment, our modern world would almost immediately come to a standstill. That sounds dramatic, but it shows how deeply chips are integrated into our lives. Often we do not even notice them, because they work silently in the background.
Everyday Devices
In smartphones, tablets and laptops, the main processor chip executes apps and the operating system. Additional chips handle mobile networks, Wi-Fi, Bluetooth, the camera, charging and the display. Even a simple action like taking a photo involves many chips working together: one controls the sensor, one processes the image, another stores it and others send it to the cloud.
At home, chips control washing machines, dishwashers, ovens, fridges, robotic vacuum cleaners and smart speakers. They make sure programmes run correctly, temperatures are measured and energy is used efficiently. Without chips, many modern devices would go back to being completely mechanical and much less flexible.
Medicine, Industry and Agriculture
In hospitals, computer chips are built into heart monitors, ventilators, MRI scanners and surgery robots. They help doctors measure important values precisely and react quickly when something changes. In industry, chips steer robots on assembly lines, manage warehouses and monitor machines so that they do not break suddenly.
Even agriculture now relies on chips: tractors use GPS to drive more exactly, drones monitor fields from the air and sensors in the soil measure moisture and nutrients. All these systems need chips to collect data and make smart decisions.
Large Systems and Data Centres
In big computers and data centres, thousands of chips are connected in racks. They store our photos, run online games, stream films and answer our search queries in fractions of a second. More and more often, they also run artificial intelligence that recognises speech, translates languages and detects patterns in large data sets.
Because chips are so universal, they are sometimes called the “building blocks of the digital age”. Understanding them means understanding a big part of how the modern world functions.
Chips Over Time – Progress, AI and Crisis
1. Strong Technological Progress
In the first decades of chip development, engineers saw incredibly fast progress. Every new generation of chips could hold many more transistors than the previous one. This development is known as Moore’s law. It states that the number of transistors on a chip roughly doubles every two years.
Because of this doubling, computing power grew exponentially. What once required a whole room full of hardware now fits into a laptop or even a smartphone. At the same time, each individual calculation became cheaper and used less energy. This made digital technology attractive for many new areas – from music players to navigation and later to smartphones.
2. The Rise of Special AI Chips
When machine learning and deep learning started to grow, classical processors reached their limits. Training a large AI model can require billions of mathematical operations. To handle this, developers began using graphics processing units (GPUs) and later designed special AI accelerators.
These chips are optimised for doing many simple calculations in parallel. This makes them ideal for operations in neural networks. Thanks to them, applications like face recognition, automatic translation and advanced voice assistants became fast enough to be used in real products. Without AI chips, this would either be too slow or far too expensive.
3. The Global Chip Crisis
During the COVID-19 pandemic the world experienced a global chip shortage. Some factories had to stop production because of lockdowns, while demand for laptops, webcams and game consoles increased. At the same time, car manufacturers had cancelled chip orders earlier in the crisis and later suddenly needed them again.
Because building new chip factories takes years, the system could not react quickly. As a result, car plants stood still, delivery times for electronics exploded and prices rose. The crisis showed how fragile the supply chain is and how much our economy depends on a small number of high-tech factories.
Taiwan – A Central Player in the Chip World
Taiwan is one of the most important locations for chip production on the planet. A large part of the world’s most advanced chips is manufactured there, especially by the company TSMC. Its factories produce chips for many famous brands – including smartphone makers, car manufacturers and cloud providers.
What makes Taiwan special is not only the number of factories, but the very high technical level. Producing chips with structures of only a few nanometres requires extremely precise machines, long experience and a highly trained workforce. At the moment, only very few companies in the world can do this at large scale.
Because of this, many countries depend on chips from Taiwan. If the factories there had to stop – for example due to a natural disaster or political conflict – the effects would be felt worldwide. Cars could not be finished, smartphones would become more expensive and cloud services might slow down.
That is why governments in the USA, Europe and Asia are now investing billions of dollars to build more chip factories in their own regions. They want to reduce their dependence on a single island and protect themselves from future crises. Chips have become not only a question of technology, but also of economic security and political power.
AI Chips – The New Gold and an Environmental Challenge
AI chips are sometimes called the new gold. They are scarce, expensive and extremely valuable for companies that want to develop or use artificial intelligence. Big tech firms order whole warehouses full of these chips to train ever larger AI models.
To operate them, companies build enormous data centres. In such a building, thousands or even hundreds of thousands of servers run day and night. Many of them are equipped with powerful GPUs or other AI accelerators that consume a lot of electricity.
Energy Use and CO₂ Emissions
Data centres already account for a noticeable part of global energy consumption. When they use electricity produced from coal, oil or gas, this leads to high CO₂ emissions. Training one very large AI model can use as much energy as several households over many years.
Because of this, there is now a strong discussion: How much AI is useful and worth the energy, and where do we need to be more careful? Some companies promise to power their data centres completely with renewable energy, but this is not always available everywhere and at all times.
Water Use and Local Conflicts
Another issue is water. In many data centres, water is used to cool the servers. In regions that are already dry, this can create tension with local communities and farmers who also need water. Some planned AI data centres in the USA have faced criticism for exactly this reason.
All of this shows that AI is not just a question of clever software. It is also a question of infrastructure and resources. If AI is to be sustainable, we need more efficient chips, better cooling technologies and electricity from renewable sources such as solar, wind and hydro.
In summary, AI chips drive innovation and open new possibilities – from medical research to language tools. At the same time they remind us that every digital service has a physical side: servers, power plants, cables and cooling systems in the real world.
AI Data Centers – Powering AI, Polluting Cities?
AI chips do not work alone. To run modern artificial intelligence, companies build huge AI data centers. These are large buildings filled with racks of servers and thousands of GPUs or other AI accelerators. Together, they train AI models, answer questions, generate images and run many online services.
But this comes with a cost. AI data centers need enormous amounts of electricity to power the chips and to cool them. If this electricity comes mainly from fossil fuels such as natural gas or coal, the result is a lot of CO₂ and other pollution. In some regions, data centers also use huge quantities of water for cooling, which can cause conflicts in areas that already suffer from water stress.
Case Study: xAI Data Center in Memphis, Tennessee
A very current example is the AI supercomputer project of xAI, the AI company founded by Elon Musk, in South Memphis, Tennessee. The supercomputer, called Colossus, is built to power xAI’s chatbot and other AI systems. To feed this machine with energy, xAI installed a large number of methane gas turbines directly at the data center site.
According to environmental groups, xAI set up dozens of these gas turbines near residential areas in South Memphis. Together, they have the capacity to produce hundreds of megawatts of power – enough to supply electricity to tens of thousands of homes. However, the turbines also emit nitrogen oxides (NOx), formaldehyde and other pollutants that are linked to asthma, heart problems and an increased cancer risk.
Local organisations like the Southern Environmental Law Center, NAACP and community groups in South Memphis argue that the data center makes air quality worse in a neighbourhood that is already heavily burdened by industrial pollution and has a high share of low-income and Black residents. They criticise that xAI has operated some turbines without proper permits and that people in the area were not transparently informed about the plans.
This example shows how the AI boom can create environmental justice questions: Who gets the benefits of AI, and who has to live with the pollution and health risks of the supporting infrastructure?
How This Connects Back to Chips
Inside a data center like the one in Memphis, hundreds of thousands of computer chips are working at the same time. Many of them are powerful AI chips from companies like Nvidia. Each chip on its own is small, but together they form a gigantic power-hungry machine. Without the chips, there would be no need for the turbines – and without the turbines, there would not be enough electricity for the chips.
So when we talk about the environmental impact of computer chips, we also have to look at the whole ecosystem: chip factories, data centers, power plants and the communities around them. The goal for the future must be to make this system more sustainable – for example by using renewable energy, improving cooling and designing AI models that need less computing power.
Video: What Are Computer Chips Made Of?
The following video explains what computer chips are made of and how they are produced. It helps to understand what is actually inside all those servers in data centers like the xAI facility in Memphis.
The Future of Chips – Beyond Simple Shrinking
For many years, “better chips” mainly meant “smaller structures”. This strategy is slowly reaching its limits. When transistors become only a few atoms wide, quantum effects disturb their behaviour and manufacturing becomes incredibly difficult and expensive.
Quantum Technologies
One possible next step is quantum computing. Quantum chips use qubits instead of classical bits. Qubits can be in several states at the same time, which can make some calculations much faster. For example, quantum computers could help to design new materials, simulate molecules for medicine or break certain types of encryption.
However, quantum computers are still in a very early stage. They are extremely sensitive to noise and need to be cooled to almost absolute zero. It will probably take many years until they are used widely, and even then they will only be helpful for specific problems.
New Materials
Another research direction are new materials like graphene or other two-dimensional semiconductors. These materials are sometimes only one atom thick and have exciting electrical properties. They could allow faster switching speeds or lower energy consumption.
If such materials can be produced cheaply and reliably, they may complement or partly replace silicon in some components. Flexible electronics, transparent displays or sensors integrated into clothing are possible applications.
Smarter Architectures
Even without completely new physics we can build smarter chips. Neuromorphic chips, for example, imitate the structure of the human brain. Instead of separating memory and processing, they combine both and send small electrical spikes between artificial neurons. This can make certain AI tasks much more energy-efficient.
Other architectures bring memory and processing closer together or specialise chips for concrete tasks such as video decoding, encryption or sensor fusion. The goal is not only maximum speed, but also a good balance between performance, energy and cost.
The future of chips will therefore be a mix of many ideas: some based on silicon, some completely new. What they all have in common is the same mission: to handle more information with less energy and to make the digital world smarter without overloading the planet.
Thank You for Reading
This website is meant as a detailed explanation of our topic. If you have read everything, you now understand much more than just “chips make computers work”. You have seen how they are built from sand, where they are used, why AI needs special chips, how this affects the environment and which ideas researchers have for the future.