It is not an overstatement to say that the advent of generative artificial intelligence (GenAI) is a pivotal moment in history. The breadth, depth, and scope of this genuinely revolutionary technology mean that things will never be the same again.
We are in the first phase of the GenAI era, which has been going on for about 18 months since ChatGPT broke onto the scene and became one of the best-known names in the field. Yet it was not its foremost pioneer nor even its most prominent force. For a quarter of a century, the world has turned to the big search engines in their hunt for answers and knowledge, Google being the biggest.
Increasingly, GenAI is where people now turn to for research and content generation. This has opened the door to profound change and revised priorities, being no longer tethered to static data.
GenAI has disrupted the foundations on which countries and corporations operate, such as fitting a new and very different engine to the same car. As with all the big technological innovations of the past millennium, those who do not keep pace with the change risk being left behind.
Championing chips
The development of AI hinges on the evolution of semiconductors (the basic materials used to make microchips and integrated circuits) that power devices like phones, computers, cars, satellites, and medical devices like MRI machines. They store, process, analyse, transmit, and multiply data, which is needed to operate everything from data centres to drones and fighter jets. In short, they have become as indispensable as oxygen.
When a new technology is shown to have such profound potential, it can have a similarly profound effect on global geopolitics. Such is the case with GenAI.
Tension between the United States and China has intensified over Taiwan. Home to the world’s leading chipmaker, Taiwan Semiconductor Manufacturing Corporation (TSMC) produces 90% of the world’s most advanced chips used in AI and quantum computing and is opening plants in Japan and the US because both Washington and Tokyo are eager to expand domestic chip production. Yet its main operations are in Taiwan, which Washington regards as an independent country but Beijing considers a breakaway province.
Regardless, TSMC is of such strategic global importance that both the US and China will want to keep access to it. Several major US companies, including Apple, Qualcomm, NVIDIA, and Advanced Micro Devices (AMD), depend on TSMC's tiny chips, which are measured in nanometres.
To understand scale, there are one million nanometres in a millimetre. Bacteria are around 1000 nanometres (nm) wide, viruses 100nm, proteins 10nm, and DNA molecules 2.5nm. TSMC produces 3nm chips, but these will soon shrink to 2nm—about the length of four atoms.
Industry ambitions
Over in Silicon Valley, ChatGPT's developer OpenAI is largely credited with starting the AI boom. Its far-sighted founder and chief executive, Sam Altman, has emerged as a global GenAI luminary. Backed by the likes of the United Arab Emirates and Singapore's Investment Corporation, Altman now wants raise a staggering $7tn to build chip factories and data centres to establish dominance in these sectors.
It reflects the US aspiration to reclaim global tech industry leadership. Washington is also encouraging allies and partners to bolster their own capabilities, calling for joint partnership working to replace competitive secrecy. This approach aims to prevent global fragmentation in the face of monopolisation. The US believes semiconductors will shape the economy – and geopolitics – in the forthcoming decades in the same way oil reserves defined previous eras.
The industry is already huge. In 2023, global sales of microchips topped $520bn and were expected to increase by 13% to $588bn in 2024. Investors are all too aware of their value. In December 2023, the combined market value of the top 10 global chip companies was a mind-blowing $3.4tn, up 74% from $1.9tn just over a year earlier.
A report by consultancy Deloitte projected AI-related sales to pass $50bn this year, representing around 8.5% of total industry sales. Lisa Su, the chief executive of AMD, has said AI chip sales could pass $400bn by 2027. Asia dominates semiconductor production, accounting for between 75% and 90% of global output, with a significant presence in China and Taiwan.
While semiconductors contribute a small proportion of global economic output (as measured by gross domestic product, or GDP), they are crucial in powering economically valuable products and industries worth trillions of dollars. A quarter of America's 226 manufacturing industries rely on semiconductors as a primary input, constituting 39% of industrial GDP.
As chips become more integral in more sectors, this dependence will only grow, so any disruption in chip production will reverberate throughout the economy, slowing growth and hitting supply chains.
Giants of the sector
Intel, a US-based semiconductor company, powers the chips in personal computers, corporate servers, self-driving cars, and Internet of Things (IoT) connected devices in sectors such as retail, industry, and healthcare. The company, which designs and fabricates the chips, is expanding nicely. It has invested $20bn in Ohio, opened a new plant in Ireland, plans to open another in Poland, and invested $33bn in Germany. As of January, it was worth $180bn.
In terms of value, however, this looks sluggish next to NVIDIA, a competitor that is also based in Santa Clara, California, whose growth has been nothing short of stratospheric. That is partly because it decided several years ago to focus on selling high-performance semiconductors for gaming and AI. Investors also love that it only designs the chips, outsourcing their fabrication to others.
Read more: NVIDIA's ascent shows the wild digital capitalism of the AI age
Those chips power the data centres that provide cloud access to Amazon, Google, Microsoft, China's Alibaba, and most of the world's big IT firms and science teams that model everything from new drugs to weather patterns. In the words of The Economist, "It has created a broad, deep moat that protects its competitive advantage", which seems to have worked. On 30 May 2023, its value reached $1tn. Just nine months later, in February 2024, it hit $2.3tn.
NVIDIA is now the third most valuable company in the world, behind Microsoft and Apple. Giants such as Saudi Aramco and Google-owner Alphabet are worth less. The reason is the world's soaring demand for GenAI chips and cloud services. Its clients, such as Meta and Tesla, helped NVIDIA's fourth-quarter revenue in 2023 reach $22bn, up 265% from last year. The company's latest chip, the Blackwell B200, boosts its control of the chip design market to 80%.
The Blackwell B200 is 25 times more energy efficient and has significantly enhanced "human interaction capabilities". It recognises voices far more accurately and even understands users' psychological states, giving tailored responses. It overshadowed Intel's recent AI chip launch, the Gaudi 3, while AMD launched its new chip, the M1300, back in January. It was compared to NVIDIA's top chip, then the H100, but this has since been overtaken by the B200.
As the industry evolves, NVIDIA faces competition from its own huge tech customers venturing into AI-based chip development for cloud computing services. The top three providers—Amazon, Microsoft, and Alphabet—collectively comprise 66% of the world's cloud computing market share. Despite the economic and political turbulence, spending on cloud infrastructure services continues to grow.
It is not just American firms that are planning to build. South Korea's Samsung is investing $17bn in a new chip factory in Texas and significantly expanding its semiconductor manufacturing at home. Based in Idaho, Micron Technology is building a new $3bn chip facility in India and expanding its manufacturing capacity in Japan.
ARM Holdings, a UK chip designer, generates revenue from licensing fees for its semiconductor designs and collects a royalty for each chip sold that uses its technology. Its designs power nearly every smartphone in the world. Bought for £24bn by Japanese conglomerate SoftBank in 2016, it listed on Nasdaq last year, where it was valued at $54bn. Since then, its value has doubled, having briefly been worth $149bn in February.
Concentration and scarcity
SoftBank has allocated $100bn for AI-related projects, including semiconductor manufacturing, yet despite these investments, there are still concerns over global chip supply, which relies on intricate and fragile supply chains, with around 16,000 suppliers across different countries.
In response, the European Union and the US are trying to boost local chip production using legislation and taxation, among other things. Both are concerned about the concentration of production and the scarcity of materials.
Alongside the dominance of companies like TSMC and NVIDIA, analysts worry about a scarcity of metals used in chip manufacturing and the dependence on Advanced Semiconductor Materials Lithography (ASML), a Dutch company, for the lithography machines that make the chips.
ASML's chief executive, Peter Wennink, expects a surge in demand, with more than 20 new semiconductor plants being built. These are big projects, too. According to Intel, building a semiconductor factory takes around three years. As such, the resilience of chip supply chains is still some way off.
Given its importance, microchip production is a concern for big companies and governments, who are worried about national strategies and priorities. While dominance in this field confers influence and control, it also leaves the industry susceptible to fragmentation and geopolitical tensions, particularly regarding overseas supply lines.
Big players need to cooperate in this highly interconnected supply chain. For instance, while US companies lead in intellectual property and chip design, equipment and materials used in the manufacturing come from Europe and Japan.
While South Korea designs, manufactures, assembles, and tests memory chips, China provides the silicon—the main raw material for chip production—while Taiwan holds the largest share of the chip industry. Each element is critical, yet most are highly concentrated in limited locations, heightening the risk of disruption. Technological relations between the countries are, therefore, integral to national security considerations.
Repatriating the tech
Washington has legislated in response to recently disrupted commodity supply chains, especially in the chip sector, due to pandemic-era restrictions and the economic fallout from Russia's invasion of Ukraine. Measures include the US CHIPS and Science Act (CHIPS stands for Creating Helpful Incentives to Produce Semiconductors). It earmarked $280bn over the next decade, including $53bn to entice semiconductor production to the US.
Another $200bn is allocated for research, development, and commercialising cutting-edge technologies, including quantum computing, AI, clean energy, and nanotechnology. New regional high-tech centres are also envisaged.
Stringent export controls prevent companies from selling their products to China, where firms like Huawei and the Semiconductor Manufacturing International Corporation (SMIC) still rely on US tech for their chips. Despite the incentives to "reshore" chip-making, there are challenges and limited opportunities for emerging companies because industry giants dominate the landscape to such an extent.
Another issue is the shortage of the skills needed for the semiconductor industry. There is an estimated shortfall of 70,000 jobs in the US alone. Two million people work in the semiconductor workforce today. Deloitte thinks that to sustain industry growth, another million will be needed by 2030.
Some analysts worry that US incentives may be insufficient, given the vast sums involved. TSMC, for instance, has invested $44bn since the beginning of 2022. Like the US, the European Union has also legislated and created a $46bn fund to double the bloc's share of global chip production from 10% to 20% by 2030.
Japan, which has expertise in producing the ultra-pure chemicals crucial for semiconductor manufacturing, is allocating $13bn to support the industry, while New Delhi's India Semiconductor Mission has been set up to foster the development of its own chip ecosystem.
Margrethe Vestager, the European Commission's vice president for competition and the digital economy, thinks the US and EU will together account for around 50% of the world's semiconductor production but denies that this is driven by protectionism or a quest for self-sufficiency.
The role of China
Not one to miss out, China is raising $27bn to accelerate the development of cutting-edge tech to counter growing US restrictions on chips and AI, according to Bloomberg. Beijing has stepped up its support for the domestic chip industry to reduce reliance on foreign semiconductor suppliers. In late March, it banned the use of US chips (mainly from Intel and AMD) in Chinese government computers.
Last year, sales to China accounted for $54bn in revenue for Intel (or 27% of its total) and $23bn for AM (15% of its total). Beijing has now put $91bn aside for Chinese-made IT alternatives, to be in place by 2027.
The competition among chip makers to acquire advanced lithography machines from ASML has been striking. China's imports of these machines surged by 450% at the end of last year before Dutch export restrictions were introduced after US pressure.
Wennink of ASML says politics continues to drive the chip industry and semiconductor market. China needs sophisticated equipment to further its domestic progress.SMIC produced a 7nm processor for Huawei's Mate 60 Pro in 2023, but both companies lag behind industry leaders like Samsung and TSMC. The latter began producing 3nm chips in 2022 when the US stopped exports of the most advanced chips and manufacturing tools to China (Huawei has been sanctioned by the US since 2019).
China wants to establish a comprehensive chip manufacturing ecosystem, with Huawei having developed design automation software to create blueprints for chips. This could establish China's presence in the semiconductor industry.
Companies such as SMIC, whose market share reached 16% last year, are also expected to grow, perhaps to as much as 30% by next year. Other Chinese companies like Northern Huachuang Technology (NAURA) and Zhejiang Wazam New Materials (a supplier of films used in semiconductor insulation) have also experienced growth.
Chris Miller, a professor of international history at Tufts University and author of Chip War: The Fight for the World's Most Critical Technology, says China is the only G7 nation trying to establish its own independent semiconductor ecosystem. Others are not convinced that self-sufficiency in semiconductor production is feasible or even desirable, he suggests.
Most think dominance in chip design and manufacturing could lead to global economic dominance in the coming years, so China's threats towards Taiwan are treated with the utmost seriousness in Washington. Military action could cripple the global chip industry, which no one can afford.
Mark Liu, chair of TSMC's board of directors, said: "If military force is employed or an invasion occurs, our factories would cease operations. They rely on real-time communication with key partners across Europe, Japan, and the US."
Challenges and worries
One of the risks associated with increased chip production is increased water usage. According to Standard & Poor's, the US ratings agency, the water consumed by foil mills to cool machinery and maintain cleanliness is equal to the water consumed in Hong Kong, home to 7.5 million people. Given the world's water scarcity, the chip industry's significant water consumption, which is rising by 10% annually, has become a growing concern.
TSMC's ultra-pure water consumption is up 35% per unit produced, particularly as it advances to 16nm processing chips. Global demand for AI could lead to the consumption of up to 6.6 billion cubic metres of water by 2027—equivalent to half the water consumed in the UK every year. This has hindered several big data centre projects, such as a Google site in Santiago, Chile, where concerns about water consumption stopped construction.
Data centres already use up to 1.5% of the world's electricity. If current trends continue, AI energy demands will surpass those of the Netherlands by 2027. There are other factors, too. A 7.4-magnitude earthquake in Taiwan in early April meant that operations at TSMC were suspended and its factories evacuated. The quake destroyed entire batches of microchips, highlighting the world's dependence on TSMC for semiconductor manufacturing. It could have been much worse, though, and therefore acts as a warning.
Securing the future
Diversifying the sources and spread of chip production is now a critical geostrategic, commercial, and security matter. The US has found $40bn for TSMC to build two chip factories in Arizona.
The company is also building its first plant in Japan. Just like the oil industry, any disruption in the flow of chips and semiconductors has the potential to cause big economic problems, including potential collapses.
The semiconductor industry's challenges and constraints are not limited to resources and politics. As per Deloitte's report, it is also uniquely susceptible to cyber threats, not only to chip companies but also to suppliers, distributors, and manufacturers. To protect against such vulnerabilities, one solution may be to use the very technology that these chips enable: AI.
In summary, the speed of AI's development and the scope of its impact is like nothing the world has seen before. With global powers racing to repatriate key industries, algorithms alone cannot predict future developments. Predicting such a fast-moving industry—even in the near future—is foolish.
Yet it seems clear that the intersection of AI and chip technology will shape the trajectory of technological progress, with implications for ethics, governance, and society, which is yet to feel the full impact.
The world is in for a wild ride. Working together to secure the sustainability and security of the global chip and semiconductor industry would be one way of grabbing the reins.
