AI and semiconductors

Semiconductors and AI are closely intertwined. Semiconductors are the backbone of modern computing and are present in a vast array of electronic devices, from servers and data centres to smartphones and laptops. At the same time, AI is a quickly expanding technology that critically depends on computing power and data processing abilities.

AI for improving semiconductors

The synergy between semiconductors and AI is essential for advancing AI capabilities and propelling innovation in the semiconductor industry. Semiconductors serve as the foundation for AI technologies by providing the necessary computing power and data processing capabilities. AI algorithms rely on the processing capabilities of semiconductors, such as central processing units (CPUs), graphics processing units (GPUs), and specialised AI hardware, like tensor processing units (TPUs) and neural processing units (NPUs). These semiconductor components accelerate AI computations and enable the execution of complex AI algorithms.

AI also enables improved design, more effective manufacturing processes, and enhanced chip performance. AI can be used, for instance, to anticipate equipment breakdowns, optimise process variables, and enhance quality control, all of which can result in more effective and economical manufacturing processes. Overall, AI’s integration into the semiconductor industry can help enhance the entire lifecycle of semiconductor chips, making semiconductors more efficient, cost-effective, and reliable. AI has also become a valuable tool in semiconductor design, assisting chip engineers in navigating the complexities of their work.

Challenges in using AI in the semiconductor industry

The use of AI in the semiconductor industry presents several challenges, including increased complexity and costs. Specialised hardware and algorithms must be developed to effectively implement AI, necessitating significant investments in research, development, and infrastructure. The proprietary nature of AI algorithms, data, and models used for optimisation and design can potentially lead to infringements on intellectual property rights, thus raising ownership disputes. With AI systems being connected to networks, they are more susceptible to cyberattacks and unauthorised access.

Semiconductors, often referred to as microchips, or simply chips, are an essential component of electronic devices that have become an important part of our everyday life. We can find them in our smartphones, computers, TVs, vehicles, advanced medical equipment, military systems, and countless other applications. In 2021, the sales of semiconductors reached a record $555.9 billion, according to the Semiconductor Industry Association. It is estimated that we use 120 chips per person on the planet on average. For example, a typical car uses between 50 and 150. However, a modern electric vehicle can use up to 3,000.

Semiconductor chips power our world. They are a key component of nearly every electronic device we use and they also power factories in which these devices are produced. Think for a minute of all the encounters you have with electronic devices. How many have you seen or used in the last week? In the last 24 hours? Each has important components that have been manufactured with electronic materials.

pic 1
Image by axonite from Pixabay

To understand the important role of semiconductor chips we have to explain what they are and how they are designed and produced. A substance that does not conduct electricity is called an insulator. A substance that conducts electricity is called a conductor. Semiconductors are substances with the properties of both an insulator and a conductor. They control and manage the flow of electric current in electronic equipment and devices. 

The most used semiconductor is silicon. Using semiconductors, we can create electronic discrete components, such as diodes and transistors and integrated circuits (ICs). An IC is a small device implementing several electronic functions. It is made up of two major parts: a tiny and very fragile silicon chip and a package, which is intended to protect the internal silicon chip and to provide users with a practical way of handling the component. Semiconductor devices installed inside many electronics appliances are important electronic components that support functioning of the world.

Types of chips

We can categorise types of chips according to the ICs used or to their functionality. 

Sorted by types of IC used, there are three types of chips:

  • Digital 
  • Analog 
  • Mixed

Most computer processors currently use digital circuits. These circuits usually combine transistors and logic gates. Digital circuits use digital, discrete signals that are generally based on a binary scheme. Two different voltages are assigned, each representing a different logical value. On the other hand, in analog circuits, voltage and current vary continuously at specified points in the circuit. Power supply chips are usually analog chips. Another application using analog circuits is communication systems. Mixed integrated circuits are typically digital chips with added technology for working with both analog and digital circuits. An analog-to-digital converter (ADC) and a digital-to-analog converter (DAC) are essential parts of these types of circuit. 

pic 2
Image by Recklessstudios from Pixabay

Sorted according to functionality, categories of semiconductors include:

  • Memory chips 
  • Microprocessors 
  • Graphic processing units (GPUs)
  • Application-specific integrated circuits (ASICs) 
  • Systems-on-a-chip (SoCs)

Memory chips

The main function of semiconductor memory chips is to store data and programs on computers and data storage devices. Electronic semiconductor memory technology can be split into two main categories, based on the way in which the memory operates:

  1. Read-only memory (ROM
  2. Random-access memory (RAM)

There are many types of ROM and RAM available. They stem from a variety of applications and also the number of technologies available. This section contains a brief overview of the functionality of the main memory chip types.

table 1
Overview of main memory chip types
pic 3 Memory Sharing infographic July 2022
Semiconductor Memory Technologies by Diplo Creative Lab


Microprocessors are made of one or more central processing units (CPUs). Multiple CPUs can be found in computer servers, personal computers (PCs), tablets, and smartphones.

The 32- and 64-bit microprocessors in PCs and servers today are mostly based on x86 chip architectures, first developed decades ago. Mobile devices like smartphones typically use an ARM chip architecture. Less powerful 8-, 16-, and 24-bit microprocessors (called microcontrollers) are found in products such as toys and vehicles. We will address these architectures later in the Technology and production section as the first step of chip production.

Graphic processing units 

A graphics processing unit (GPU), which is a type of microprocessor, renders graphics for the smoother display that is expected in modern videos and games by most consumers of electrical devices. GPU rendering is the use of a GPU in the automatic generation of two-dimensional or three-dimensional images from a model, done by computer programs. 

A GPU can be used in combination with a CPU, where it can increase computer performance by taking some more complex computations, such as rendering, from the CPU. This is a big improvement, since it accelerates how quickly applications can process data; the GPU can perform many calculations simultaneously. It also allows development of more advanced software in fields such as machine learning and cryptocurrency mining.

Application-specific integrated circuits

Application-specific integrated circuits (ASICs) are made for a specific purpose. They enable significant amounts of circuitry to be incorporated onto a single chip, decreasing the number of external components. They can be used in a wide range of applications, such as bitcoin mining, personal digital assistants, and environmental monitoring.


The system-on-a-chip (SoC) is one of the newest types of IC chips, a single chip that contains all of the electronic components needed for an entire electronic or computer system. The capabilities of an SoC are more comprehensive than those of a microcontroller chip, because they almost always include a CPU with RAM, ROM, and input/output (I/O). The SoC may also integrate camera, graphics, and audio and video processing in a smartphone.

Technology and production

pic 4
Image by Ranjat M from Pixabay

Production phases

Most semiconductor companies choose to work on two main stages of production: manufacturing and/or design. Those focused solely on manufacture/fabrication are called foundries (also known as fabs or semiconductor fabrication plants). Those focused on design are called fabless companies. Fabless companies such as Broadcom, Qualcomm, and HiSilicon (the in-house design firm of China’s Huawei) specialise in chip design and outsource fabrication, assembly, and packaging. They contract the Taiwan Semiconductor Manufacturing Company (TSMC) and others to fabricate for them. A third type semiconductor company focuses on both manufacturing and design, and are called Integrated Device Manufacturers, or IDMs. Intel and Samsung are among the world’s biggest IDMs. Other semiconductor companies work on assembly and packaging, and the manufacture of semiconductor equipment. 

Process nodes and wafers

One term you might often notice when reading about chips is process node. This represents the standardised process used across a whole range of products. The semiconductor process is based on a set of steps to make an IC with transistors that have to meet certain levels of performance and size characteristics. Standardising the process allows faster production and improvement of these chips. Separate teams are not needed for each smaller group of products; the same solutions can be used for many products at the same time. This makes production more efficient. Creating a smaller process node means coming up with a new manufacturing process with smaller features and better tolerances by integrating new manufacturing technologies.

Process nodes are usually named with a number followed by the abbreviation for nanometer: 7nm, 10nm, 14nm, etc. Nowadays, there is no correlation between the name of the node and any feature of the CPU. TSMC’s vice president of corporate research, Dr Philip Wong said concerning the node names: “It used to be the technology node, the node number, means something, some features on the wafer. Today, these numbers are just numbers. They’re like models in a car – it’s like BMW 5-series or Mazda 6. It doesn’t matter what the number is, it’s just a destination of the next technology, the name for it. So, let’s not confuse ourselves with the name of the node with what the technology actually offers.” 

Another term you might run into is wafer. It is a thin slice of semiconductor, such as a crystalline silicon, used for the fabrication of ICs. The larger the wafer, the more chips that can be placed on it.

Picture of a wafer by Samuel Faber from Pixabay

Investment in producing semiconductor chips

The main goal of producing semiconductor chips is to try and make them as small as possible. If we can create a smaller process node, we can have smaller chips and we can fit more of them on a wafer, which results in higher profit. However, transistors are physical objects; there is a physical limit to how small they can be.

The history of semiconductor chips

In 1965, Gordon Moore, the co-founder of Fairchild Semiconductor International Inc., and Intel (and former CEO of the latter), predicted that manufacturers would go from 65 to 65k transistors per processor in the next 10 years. Moore’s predictions of the exponential growth trajectory that the industry was on were captured in Moore’s Law, which states that the number of transistors in a dense IC doubles about every two years. The Law not only predicted the increasing computer power, it was also a self-fulfilling prophecy. The improvement in semiconductors over the years attracted more investment in production, materials, and manpower, which in turn brought a lot of profit.

Steps in the chip production

chart 1
Chart of chip production split

Instruction Set Architecture 

As a first step, how the processor will perform its most basic instructions is defined, for example do calculations or access memory. The Instruction Set Architecture (ISA) acts as an interface between the hardware and the software, specifying both what the processor is capable of doing as well as how it gets done. The main goal is to turn this model of how the CPU is controlled by the software into an industry standard. This allows processors and operating systems of multiple companies to follow the same standard and become interoperable. 

For example, Windows, Mac iOS, and Linux can run on a variety of Intel and AMD chips through the power of x86. The x86 ISA family was developed by Intel; it is the world’s predominant hardware platform for laptops, desktops, and servers. For mobile phones, Arm chips are used in most cases. Arm is a reduced instruction set computing (RISC) architecture developed by the British company Arm Limited. Some companies, such as Samsung and Huawei, create their own chips. Intel and AMD own most of the x86 and only licence their ISA to a single active competitor: VIA Technologies Inc., in Taiwan. Moving a complex operating system to a new ISA would take a lot of time.

Chip design

Today, circuit diagrams are created by companies. Some of them, like Intel and Samsung, manufacture what they design. However, most are fabless companies; they outsource the manufacturing to foundries. This allows them to focus only on the design part, while other parts of the process are left to other players. In addition, a lot of companies that use specialised chips now design their own chips, so that they don’t have to rely on Intel, for example, to create a chip that suits their needs. Examples include Apple, Samsung, and Huawei designing chips for their phones; Google for its AI service Tensorflow; Microsoft and Amazon for their data centres. 


In this step, the goal is getting that design onto silicon wafers. This is a complex process that also requires a lot of capital. It is extremely expensive to produce chips, as manufacturers have to spend around 30%–50% on capital expenditures, compared to the 3%–5% designers spend. In most cases it is done by foundries, such as TSMC. Leaders in the field, foundries can use their machines in multiple production parts for different kinds of chips. Most competitors gave up on trying to compete with TSMC since it didn’t make sense economically

Some of the world leaders are deciding to split their design and fabrication business, such as Samsung and Samsung Foundry, and AMD and GlobalFoundries. Even Intel might start outsourcing their manufacturing to an external foundry. 

Equipment and software

Custom equipment and software are required for each chip. For example, extreme UV lithography machines (EUVs) are required in lithography, in which the design is transferred to the silicon wafer using EUVs. The Dutch company ASML is the sole producer of high-end EUV machines. ASML CEO Peter Wennink said they have sold a total of about 140 EUV systems in the past decade, each one now costing up to $200 million. TSMC buys around half of the machines they produce. This is just one example of a monopoly in the production of equipment.

Packaging and testing

Silicon wafers are cut up into individual chips. Wires and connectors are attached. And the chips are put into a protective housing. They’re tested for quality before being distributed and sold. 

The future of semiconductor chips

As semiconductors get increasingly complex, it will be more and more expensive to compete in this space, creating a further concentration of power, which in turn creates economic and political tensions. Other factors, such as experiments with new materials for semiconductors, changes in the prices of metal materials, and the increase in development of new technologies in artificial intelligence (AI), internet of things (IoT), and similar fields will affect future sales and add new challenges and opportunities.

Supply chain disruption

The supply chain issue focuses on the ongoing global chip shortage, which started in 2020. The issue is simple: demand for ICs is greater than supply. Many companies and governments are searching for a solution to accelerate chip production. As a consequence of the supply chain disruption, prices for electrical devices have increased; production times are longer; and devices such as graphics cards, computers, video game consoles and gear, and automobiles are in short supply. 

Trend to fabless companies

The move from foundries to fabless companies helped complicate the chip shortage. More and more major semiconductor factories are adopting the fabless model and outsourcing to major manufacturers like TSMC and Samsung. For example, Intel talked with Samsung and TSMC to outsource some chip production to them.

In 2020, the USA had captured 47% of the global market share of semiconductor sales, but only 12% of manufacturing, according to the Semiconductor Industry Association. The country has put semiconductors at the top of its diplomatic agenda as it tries to work out export policies with its partners. 

How COVID-19 created a global computer chip shortage

The global pandemic had a major influence on the chip shortage. COVID-19 forced people to work and do everything from home, which for most of us meant we needed to upgrade our computers, get better speakers and cameras, make home theatres, and play a lot of video games.

Most businesses struggled to set up remote work systems, and there was an increased need for cloud infrastructure. All this together, along with the pause in production during the lockdowns, caused a massive supply chain disruption for electronics companies. Some governments are now increasing their investments in this industry, so they can hopefully lessen the impact of the disruption.

The semiconductor production process is very complex. Typically, the lead time is over four months for products that are already established. Trying to switch to a new manufacturer can take over a year, specifically since chip designs need to match the manufacturer’s ability to produce those designs and make them function on a high level.

pic 6
Semiconductor development and production timelines by McKinsey & Company

How the car industry contributed to the chip shortage

Cars are getting more advanced each year and they need more semiconductors, such as advanced semiconductors to run increasingly more complex in-vehicle computer systems; and  older, less advanced semiconductors for things like power steering.

During the pandemic, the auto supply chain was disrupted. Cars require custom chips, which are commissioned by automakers. Chips for phones are not in short supply on that level, because they are designed around standardised chips. Car manufacturers use custom components to prevent aftermarket profits for third parties. The lead time to build standard semiconductors is about six months. The lead time for custom chips is two-to-three years. The pause in production created a massive delay of a lot of vehicles.

For instance, car manufacturers cut chip orders in early 2020 as sales of vehicles decreased. After sales recovered, the demand for chips increased even more than expected in the second part of 2020, which meant manufacturers had to move the production lines even later.

How the China–US trade war contributed to the chip shortage

At the base of this conflict stand two competing economic systems. The USA imports more from China than from any other country, and China is one of the largest export markets for US goods and services. Ever since September 2020, when the USA imposed restrictions on China’s Semiconductor Manufacturing International Corporation (SMIC), China’s largest chip manufacturer, it has made it harder for China to sell to companies that cooperate with the USA. Consequently, TSMC and Samsung chips were used more, creating an issue for those companies as they were already working and producing at maximum capacity.


  • Only IDMs
  • Only fabrication
  • Fabrication and IDMs
  • Design and fabrication
  • All three (design, fabrication and IDMs)
  • None of three options

In the past couple of years, semiconductors have become a geopolitical issue. The strategic technology of semiconductors is not only the foundation of modern electronics, but also the foundation of the international economic balance of power. The transnational supply chain is a big part of this technology that is now distributed in its production and supply across the world, with multiple countries specialised in particular parts of the production chain.

Since the supply chain is so internationally distributed, there has been an increase in patent infringement lawsuits in this field, lawsuits on the grounds of misappropriation of intellectual property, and ones such as GlobalFoundries seeking orders that will prevent semiconductors produced with the allegedly infringing technology by Taiwan-based TSMC from being imported into the USA and Germany.

Policy measures for cooperation have been proposed, such as the EU Commission’s proposed CHIPS Act to confront semiconductor shortages and strengthen Europe’s technological leadership, and the World Semiconductor Council (WSC) series of policy proposals to strengthen the industry through greater international cooperation. However, creating global policies and regulations that respect the national legal frameworks of each global actor in the semiconductor industry is not easy to achieve, but there is a trend towards international cooperation with policies set in place.

pic 7
Image by Gerd Altmann from Pixabay


China’s role in this supply chain is that of a vast consumer of semiconductors, importing a sizable percentage. China still cannot meet its semiconductor needs domestically. However, it is working on building a chain of production and wants to move to higher-value production. 

The manufacturing industry is built on semiconductors. China uses them in a variety of electronic manufacturing sectors. Thus, if anyone takes action against China in the semiconductor industry, they would disrupt the production chain in many other sectors. For example, US export controls directed at Huawei have had a significant effect on the global smartphone market. This has undermined Huawei’s capacity to deliver cutting-edge consumer devices, which consequently cut their market shares compared to their competitors, but acted as a stimulus to the industry. Beijing has made chips a top priority in the next 5-Year Plan. It will invest $1.4 trillion to develop the industry by 2025. In 2020, the country invested 407% more than the previous year. Its main goals are semiconductor independence with 77% of chips used in China, coming from China.


The USA is home to most chip design companies, such as Qualcomm, Broadcom, Nvidia, and AMD. However, these companies increasingly have to rely on foreign companies for manufacturing. 

“We definitely believe there should be fabs of TSMC [and] Samsung being built in America, but we also believe the CHIPS Act should be preferential for U.S. IP [intellectual property] and U.S. companies like Intel,” said Patrick P. Gelsinger, CEO of Intel in an online interview hosted by the Washington-based Atlantic Council think tank on 10 January 2022. 

In 2022, the USA announced an investment of more than $20 billion to build two new chip plants in the state of Ohio. Construction is set to begin in late 2022, with production predicted to go onstream in 2025.

During the pandemic, the Biden administration presented its plans to end the supply chain crisis by changing the supply chain, starting with changing the production of certain elements in the USA. The USA will work more closely with trusted friends and partners, nations that share US values so that their supply chain cannot be used against the country as leverage.

The US administration needs to look at both national and economical security. A thriving US semiconductor industry means a strong American economy, high-paying jobs, and a national ripple-effect, such as the impact on transportation with new vehicles increasingly relying on chips for safety and fuel efficiency.


Taiwan-based TSMC has a huge role in the global semiconductor supply chain. As the number one chip manufacturer, it has built its market dominance for years. TSMC has set such a high standard for chip production it will take a long time for a competitor to reach its level. 

Taiwan doesn’t have the same trade issues that China has, since it cooperates with many countries. For example TSMC committed to building a $12 billion fabrication plant in Arizona, USA, to start producing 5nm chips by 2024 (not 3nm, which will be the cutting edge then produced in Taiwan).

Building has begun; TSMC is hiring US engineers and sending them to Taiwan for training, although, according to Taiwan’s Minister of the National Development Council Ming-Hsin Kung, the pace of construction depends on Congress approving federal subsidies.

Taiwan is preparing to introduce tougher laws to protect the semiconductor industry from Chinese industrial espionage. “High-tech industry is the lifeline of Taiwan. However, the infiltration of the Chinese supply chain into Taiwan has become serious in recent years. They are luring away high-tech talent, stealing national critical technologies, circumventing Taiwan’s regulations, operating in Taiwan without approval and unlawfully investing in Taiwan, which is causing harm to Taiwan’s information technology security as well as the industry’s competitiveness,” said Lo Ping-cheng, Minister without Portfolio and Spokesperson for the Executive Yuan.

The overreliance on a single Taiwanese chip fabrication company carries supply chain risks for the broader semiconductor industry.

In February of 2022, Taiwan’s Economy Minister Wang Mei-hua emphasised: “Taiwan will continue to be a trusted partner of the global semiconductor industry and help stabilise supply chain resilience.” The statement said that Taiwan has “tried its best” to help the EU and other partners resolve a global shortage of chips. TSMC has said it was still in the very early stages of assessing a potential manufacturing plant in Europe, as they are currently focusing on building chip factories in the USA.

South Korea 

Samsung is a major manufacturer of electronic components, semiconductors being one of them. Although the company is catching up with TMSC, this still means only two companies are able to provide that type of service at the cutting edge of technology and it will be difficult to change this situation anytime soon. In 2021, Samsung’s market share of the global semiconductor industry was 12%. In addition to exporting semiconductors to countries such as the USA and China, Samsung uses its own semiconductors in its other products as well as selling them to technology companies in South Korea that use them, too.

South Korea has its own version of the CHIPS Act offering state support to the domestic chip industry currently led by Samsung and SK Hynix. Unlike other global actors, South Korea‘s new chip law does not specify quantitative targets for how much it would cost for their government to carry out its plans and what the consequence would be on economic growth or job creation. As a result, South Korea’s large corporations will be subject to a 6%–10% tax break for facility investment and a 30%–40% tax credit for research and development, while smaller companies will have a larger degree of tax relief.

The EU 

As a consequence of the crisis in the shortage of the semiconductor chips that led to problems such as a lack of components and European companies closing, the EU has started taking steps towards the goal of doubling chip manufacturing output to 20% of the global market by 2030. Security, energy efficiency, and green transition are additional goals it is focusing on. The new Digital Compass Plan will fund various high/tech initiatives to boost digital sovereignty.

Emerging market opportunities, such as AI, edge computing, and digital transformation bring a lot of demand for chip production. New needs for AI will bring new production models and collaboration. The EU’s strengths are R&D, manufacturing equipment, and raw materials. Its weaknesses lie in semiconductor IP and digital design, design tools, manufacturing, and packaging.

As a result, the EU needs the CHIPS Act. The goals will be to strengthen its research and technology leadership; build and reinforce its own capacity to innovate in the design, manufacturing, and packaging of the advanced chips; put in place an adequate framework to substantially increase its production capacity by 2030; address the acute skills shortage; and develop an in-depth understanding of the global semiconductor supply chains. 

Three pillars of the CHIPS Act (by European Semiconductor Board) :

  1. Chips for Europe Initiative – Initiative on infrastructure building in synergy with the EU’s research programmes; support to start-ups and small and medium enterprises (SMEs).
  2. Security of Supply – First-of-a-kind semiconductor production facilities. 
  3. Monitoring and Crisis Response – Monitoring and alerting, a crisis coordination mechanism with member states, and strong commission powers in times of crisis. 

Member states are enthusiastic about this topic. They all understand the importance and feel the effects of the shortage. They have already started working on these problems in the expert group, trying to find possible solutions to face these kinds of challenges. 

The next steps

The prospect of European or American firms that could do similar service in the next five years is unrealistic. The EU is promising to invest €43 billion over the period of 2030. TMSC will invest over $40 billion in 2022 alone on its capital expenditure, and Samsung will try to match it, which shows the difference in the investments and also further proves how it will be hard to catch up with the leaders of chip production.