The emergence of generative AI presents a significant global challenge, catalysing considerable investment by major public and private players. Though foundation models like OpenAI and Mistral are arousing great enthusiasm among the public and corporates alike, Sundar Pichai, CEO of Google, notes, ‘Chips and hardware represent the invisible infrastructure of the AI era, on which the whole edifice of modern technology rests.’ His words highlight the critical importance of the underlying infrastructure, often eclipsed by the more visible applications of AI.
Let’s investigate the key components of this value chain to untangle the knot of players, the economic dependencies and the geopolitical issues shaping the future of generative AI.
CHIPS, AT THE HEART OF GENERATIVE AI AND A CRITICAL PIECE ON THE GEOPOLITICAL CHESSBOARD
The electronic chip is the fundamental component of generative artificial intelligence. Initially the fruit of technical wonders, it now represents an issue of global power and shapes global economic strategies.
- Raw materials, a powerful lever for China
China holds a dominant role in the supply of silicon and rare-earth elements, which are essential to the production of semiconductors. Controlling 85% of the rare earths processing market, the country uses its monopoly strategically to influence the global market, particularly in response to protectionist policies like those of the United States, which have seen restrictions placed on germanium and gallium exports.
- ASML and European domination in high-precision machines
The Dutch company ASML reigns over the market for extreme ultraviolet (EUV) lithography machines, essential to etching circuits of less than 10 nm on microchips. Just one such machine is a veritable colossus, weighing some 150 tonnes, sent in 13 containers and then assembled by 250 engineers for at least six months! This technological monopoly places ASML at the centre of geopolitical debates, particularly as part of the import restrictions imposed by the United States on China, highlighting the crucial importance of these technologies in international relations.
- Nvidia and American supremacy in chip design
Nvidia, the US giant, is at the forefront of the design of specialised chips (graphics processing units – GPUs), with innovations pushing artificial intelligence to new heights. In 2023, the company generated record revenues of $49 billion, testament to the fundamental impact of these technologies in the semiconductor sector and strengthening America’s position in global technological competition. The company’s latest innovation, the Blackwell platform, represents a leap forward in the calculating capacity for generative AI, capable of managing large language models with trillions of parameters with improved energy efficiency and considerably reduced costs.
- TSMC, the Taiwanese titan in chip manufacturing
The CEO of Nvidia summarised the situation well: ‘There is water, there is air and there is TSMC.’ TSMC’s leadership in the manufacture of advanced microchips is undeniable. Controlling 90% of the market for 3 to 7 nm chips, TSMC symbolises Taiwan’s supremacy in this field. TSMC’s 3 nm manufacturing technology is particularly noteworthy and should generate more than 20% of the company’s revenues in 2024. Against this domination, the American Intel is making considerable investments to rival with TSMC, specifically more than $100 billion in the next five years to extend its manufacturing and research capacities in the US.
THE VERTICAL INTEGRATION STRATEGY OF TECH GIANTS
Vertically integrated cloud companies like AWS, Google and Microsoft are Nvidia’s largest customers. The fact is that cloud infrastructure is at the heart of generative AI, providing the calculation and storage resources necessary to train, deploy and develop these intelligent systems. At the same time as shaping the development of the generative AI application ecosystem, these players are looking to move up the value chain and integrate the critical elements of their technological infrastructure.
- Development of specific chips
To reduce their dependence on others, these tech giants have started creating their own semiconductors. Google has launched its own chips dedicated to generative AI, known as TPUs, whilst Amazon has deployed its own Graviton chips to optimise the energy efficiency of its infrastructure. Microsoft is intensifying its efforts to develop alternatives to Nvidia by making considerable investments. Finally, Apple has taken significant steps towards vertical integration by developing its own material components, such as the Apple Silicon chips, thus reducing its dependence on external suppliers like Intel and Samsung. These developments make it possible for hyperscalers to strengthen their positions on the market by offering technological solutions that are closely integrated and optimised for their services.
- Securing energy resources
To answer the extreme growth in energy requirements for data centres, tech giants are adopting strategies centred on nuclear power. These investments are earmarked for the purchase of land around nuclear power stations, the construction of new infrastructure, and partnerships with existing energy companies. Amazon has acquired a data centre powered entirely by a nuclear power station in Pennsylvania, as part of a $650 million deal with Talen Energy. Microsoft also has a clear strategy in the development of small modular nuclear reactors (SMRs) to improve its own energy efficiency but also to potentially offer these solutions to other energy suppliers around the world.
- Checking foundation models
The crucial importance of generative AI for GAFAM products is driving them to invest heavily in start-ups offering the most effective foundation models (a foundation model uses its acquired knowledge to generate original new content). This strategy has led to the concentration of the industry around a small number of powerful players, which, endowed with significant capital, are shaping the development of generative AI. As a result, the AI industry is characterised by significant transactions that illustrate the centralisation of power and the concentration of the market. By controlling the foundation models, these influential players ensure not only a technological lead but also a decisive influence on the future direction of generative AI. Microsoft has invested more than $13 billion in OpenAI to acquire a 49% stake. This strategic investment will enable Microsoft to integrate OpenAI’s technologies into its Office suite. Amazon, meanwhile, has committed up to $4 billion to Anthropic, a start-up specialising in the development of chatbot models like Claude. Apple acquired up to 32 AI start-ups in 2023, more than any other tech giant.
NEW DEPENDENCIES, NEW CHALLENGES
The concentration of the generative AI market in the hands of a few tech giants presents major challenges for global businesses and organisations. By controlling the key resources needed to develop AI, these companies can potentially limit access to the technologies needed by other players, thereby holding back innovation in other sectors or geographies. This situation could also lead to anti-competitive practices, where hyperscalers exploit their dominant positions to establish unfavourable market conditions for challengers and reinforce their monopolies.
This reality poses major strategic challenges for regulators and policymakers, who must find ways to encourage healthy competition whilst promoting innovation and fair access to cutting-edge technologies. For Europe and other regions seeking to strengthen their technological autonomy, it is becoming imperative to develop viable alternatives and support infrastructures that enable greater diversity in the development of generative AI. This can include diversifying technology suppliers, investing in open or interoperable technologies, and participating in industry alliances that can offer some leverage against the dominant powers of the tech giants.
Jean Barrère – Partner, Accuracy / Helena Javitter – Manager, Accuracy
Accuracy Talks Straight #10 – Industry insight