By Jean-Claude Muller and John-Guy Park

Summary

The article " Oil and semiconductors: a story that repeats itself but is not the same" explores the parallels between oil and semiconductors, two crucial resources that have shaped not only technology and industry, but also global geopolitics. Oil, long considered the lifeblood of the global economy, now faces a new competitor in semiconductors, the cornerstones of the digital age. This article highlights the implications, geopolitical context and historical parallels of these two vital resources.

Oil has been synonymous with economic and military power since the dawn of the 20th century. The nations that controlled the vast reserves of oil, or possessed the capacity to exploit them, often dictated the rules of world order. The transition to new forms of energy and associated power is inevitable, as summed up by the quote from former Saudi Oil Minister Zaki Yamani: "The Stone Age didn't end because the world ran out of stones; the Oil Age won't end because we run out of oil."

The article also discusses the evolution of semiconductors as the "new energy" of the future, powering almost every advanced technology, from smartphones to electric cars and artificial intelligence. Semiconductors are becoming the vital industrial resource that will power the vast majority of the economy of the future. This reality affirms that the transition to an economy where information processing and storage are just as critical as physical energy was in the 20th century.

The article also discusses the importance of GPUs (Graphic Processing Units) in the semiconductor field, their role in artificial intelligence and the geopolitical competition for access to the most advanced GPUs. Nations that dominate semiconductor production and innovation, and GPUs in particular, gain a strategic advantage in the global AI race.

In the complex fabric of the recent global economy, two resources have played and continue to play crucial roles, shaping not only technology and industry but also global geopolitics: oil and semiconductors. Oil, long considered the lifeblood of the global economy, now finds a new competitor in semiconductors, the cornerstones of the digital age. This analysis explores the parallels between these two vital resources, highlighting implications, geopolitical context and historical parallels.

Since the dawn of the 20th century, oil has been synonymous with economic and military power. Nations that controlled vast oil reserves, or possessed the capacity to exploit them, often dictated the rules of world order. As former Saudi Oil Minister Zaki Yamani put it, "The Stone Age didn't end because the world ran out of stones; the Oil Age won't end because we run out of oil." This quote sums up the inevitable transition to new forms of energy and associated power.

Petroleum, this age-old natural resource, has long been used in relatively simple applications such as lighting via kerosene lamps. It was only at the beginning of the 20th century, with the advent of the automobile and aeronautics industries, that oil became a strategic resource. These booming sectors had no viable alternative to oil as an energy source, catapulting its demand and value to unprecedented levels. This transformation not only reshaped the global economy, but also played a part in contemporary geopolitical dynamics revolving around energy resources.

The recent replacement of combustion engines by powerful, reliable electric motors illustrates a major turning point in the history of technology and innovation. It has taken almost a century to develop and refine electric motor technology to the point where they can now compete with, and often surpass, their internal combustion counterparts in terms of performance and energy efficiency. This transition to electrification reflects the ongoing move towards more sustainable solutions that are less dependent on fossil fuels.

Today, semiconductors are at the heart of almost every advanced technology, from smartphones to electric cars and artificial intelligence. Semiconductors are becoming the vital industrial resource that will power the vast majority of the economy of the future. This reality asserts that an information processing and storage economy will become just as critical as access to oil energy was in the 20th century.

A computer is made up of several key components: the processor (or CPU), which is the brain of the computer executing program instructions; memory (RAM), which temporarily stores data for rapid access by the CPU; and semiconductors, which are the materials used in the manufacture of electronic components such as transistors, essential to the operation of processors and memory. The GPU (Graphic Processing Unit) is the electronic component specialized in graphic data processing. It is designed to rapidly perform the complex calculations required to render images, videos and animations with great precision and detail.

Unlike oil, none of these components is a natural resource; they have to be designed, developed and constantly improved. This innovation takes place mainly in developed countries, while manufacturing is often relocated to countries in the South, reflecting an international division of labor influenced by production costs and the availability of specialized labor.

Semiconductors are materials whose ability to conduct electricity can be precisely controlled, making them indispensable in the manufacture of electronic circuits. Moore's Law, which predicted a doubling in the number of transistors on a microprocessor every two years or so, has long guided the industry, driving constant increases in power and efficiency. Today, it's not just a question of quantity, but of power and energy efficiency, making semiconductors a strategic issue comparable to oil in terms of importance to the global economy.

Access to semiconductors has become a major issue due to their crucial importance in a wide range of industries, from automotive to information technology, and the concentration of their production in a limited number of geographical regions. The recent global shortage of semiconductors has highlighted their critical role and the vulnerabilities of global supply chains, exacerbating geopolitical and economic tensions.

Explosive demand for artificial intelligence (AI) capabilities has turned GPUs into strategic commodities. Technology companies and research institutions are scrambling for the most powerful GPUs to fuel their AI projects, causing shortages and driving up costs. This situation is reminiscent of the race for oil, where access to a resource determines a nation's or company's ability to innovate and maintain competitiveness.

Initially, central processing units (CPUs) were seen as essential components of computing, while graphics processing units (GPUs) were seen as additional accessories, with a higher tolerance for error in calculations, due to their use mainly in video games. However, as graphics rendering technologies advanced, notably towards photorealism and high-resolution 3D, computing requirements began to diverge. Whereas CPUs were adapted to handle high volumes of information rapidly, GPUs distinguished themselves by their ability to perform parallel calculations, thus reducing latency, i.e. the delay between an event and its graphical representation.

This development has prompted engineers to explore the use of GPUs for parallel mathematical calculations, including finite element calculations, where they have demonstrated remarkable efficiency. The CUDA architecture, for example, has been adopted in professional applications such as Dassault Systèmes' CATIA, paving the way for new uses of GPUs beyond video games.

Furthermore, the rise of blockchain and, by extension, cryptocurrencies, has revealed another facet of GPU use. Cryptocurrency mining, requiring intensive computing capacity, has led to the emergence of GPU-equipped "mining farms", often at the expense of environmental considerations. This increased demand for GPUs has also influenced the gaming computer market, where a mid-range model can today fetch a price of 2,000 euros, with half of this sum attributed to the GPU alone.

The rapid evolution of artificial intelligence technologies has propelled GPUs to the status of indispensable resources. Initially designed to improve graphics performance in video games, GPUs have proved extremely efficient for executing complex AI algorithms, particularly LLMs.

A Large Language Model is a type of machine learning model designed to understand and generate text in natural language. It is trained on a large volume of text data, enabling it to learn the structures, patterns and nuances of human language. It is a computer tool capable of reading and understanding text in a way similar to that of humans. It can predict which word or phrase is likely to follow a given text, answer questions, summarize texts, translate languages and even generate original text.

These models require colossal computing power to process the vast datasets on which they are trained, making GPUs essential for the development and deployment of advanced AI. According to Jenssen Huang, one of Nvidia's three co-founders, "The revolution is just beginning". The company's sales jumped 265% to $22.1 billion in the last quarter of 2023. "The first wave concerns data centers, "veritable factories" for manufacturing AI. Once AI models have been trained, their use will require more and more computing power. The part of Nvidia's business related to inference has grown enormously because of the use of AI models to generate text, images or videos" explains Jensen Huang.

The competition for access to the most advanced GPUs also has a geopolitical dimension. Nations that dominate semiconductor innovation and production, and GPUs in particular, gain a strategic advantage in the global AI race. The main semiconductor-producing countries include Taiwan, South Korea, the USA and China. Taiwan, thanks to TSMC (Taiwan Semiconductor Manufacturing Company), is the world leader in semiconductor manufacturing. Samsung Electronics in South Korea and Intel in the USA are also major players in semiconductor production.When it comes to innovation, particularly in GPU development, the USA stands out with companies like Nvidia and AMD. Nvidia, in particular, is recognized for its significant contributions to the advancement of GPUs, both for video games and artificial intelligence applications.This has led some countries to invest heavily in their own semiconductor production capacity, in the hope of reducing their dependence on foreign suppliers. Faced with the concentration of semiconductor production in Asia, the USA has launched initiatives to revitalize its domestic semiconductor industry. The CHIPS Act is an example of these efforts, aimed at stimulating domestic semiconductor production through subsidies and financial incentives.European Union: The EU has also announced plans to increase its share of global semiconductor production, with ambitious targets for achieving self-sufficiency in this critical sector. The European plan aims to double its share of global semiconductor production by 2030.China: China is investing massively in its semiconductor production capacity as part of its "Made in China 2025" strategy, with the aim of reducing its dependence on foreign suppliers and becoming a world leader in this field.South Korea and Taiwan also continue to invest in expanding their production capacity, with massive investment plans announced by companies such as Samsung, SK Hynix, and TSMC to maintain and strengthen their dominant market positions.These efforts reflect worldwide recognition of the strategic importance of semiconductors and GPUs in the global digital economy, underlining the race for technological sovereignty in the context of geopolitical tensions and economic competition.

The growing importance of GPUs raises a number of questions, particularly in terms of accessibility and sustainability. The growing demand for these components puts pressure on the natural resources required for their production, and their energy consumption poses significant environmental challenges:

Silicon: Silicon remains the basic material for the manufacture of semiconductors, including GPUs. Although silicon is indeed abundant, the type of silicon used in semiconductors must be of extremely high purity. Transforming natural silicon into a semiconductor-grade material is a costly and energy-intensive process. What's more, the production of high-purity silicon can be concentrated in certain regions, which can create supply bottlenecks.

The transition from oil to semiconductors and GPUs as strategic resources marks a fundamental shift in global economic and geopolitical dynamics, reflecting the gradual transition from the computer age to the age of artificial intelligence. This mutation underlines the importance of innovation, sustainable resource management and adaptation to new technological realities for nations,  companies and individuals aiming to prosper in a rapidly changing landscape. To date, it is clear that the USA and China have taken the lead in this field, and that Europe, Japan and Korea are not (yet) competitive.

History never repeats itself identically, but political and economic leaders should remember the tensions and conflicts generated by difficulties in accessing or delivering oil. The same could be said of access to high-performance semiconductors, the new strategic resource that is indispensable to artificial intelligence, and which all future human activities will increasingly need in the near future.

Written in Paris and Singapore, March 6, 2024

The authors

Jean-Claude Muller is Executive Editor of btobioinnovation, a blog that publishes reports and articles on innovative topics in the life sciences. He was Senior Vice President in charge of Prospective and Strategic Initiatives at Sanofi.

John-Guy Park is an engineer and serial entrepreneur. After 15 years in listed companies in aeronautics, telecoms and pharmaceuticals, he founded 3 startups specializing in data and AI under the mentorship of Jean-Claude Muller. He works as an independent advisor to venture capital firms and VCs.