Adam Smith Meets Silicon Valley: The Invisible Hand in a Billion-Transistor World

By: Ibrahim Karabay

Imagine Adam Smith grappling with the world of microelectronics. When he passed in 1790, James Watt’s steam engine represented the apex of industrial marvels. If Smith had heard the term “nano”—with his classical training, he might have recognized its Greek root meaning “dwarf,” though he could scarcely have imagined it would describe units as infinitesimal as billionths of a meter, the scale of today’s semiconductors. Numbers in the billions would have sounded surreal, and the notion of 200 billion transistors on a sliver of silicon, like in Nvidia’s latest Blackwell AI chip, might have seemed more sorcery than science to even the most enlightened Scottish thinker.

Smith’s famous assertion that “the division of labour is limited by the extent of the market” would, however, feel at home in the global semiconductor industry. Distilled to its essence, the supply chain reflects his principle on a grand scale. Market demand for chips has expanded exponentially from $45 billion in 1979 to an anticipated $600 billion in 2024, with projections nearing $1 trillion by the early 2030s and the labor behind each chip is more finely divided than ever.

Gone are the days when giants like Intel managed every stage of chip production. In the age of hyperspecialization, designers like Nvidia create the blueprints, while specialized foundries, such as TSMC in Taiwan, manufacture the chips. These foundries rely on silicon providers and equipment assemblers, who, in turn, depend on components from niche suppliers.

Despite this precision, the spoils of the AI surge are unevenly split. Since mid-2022, the top 100 semiconductor companies by market cap have added $5.4 trillion in value, but foundries and equipment makers have only netted about $1 trillion of this; fabless firms, those that design but don’t produce chips themselves, have taken the lion’s share, with Nvidia leading, having captured three-quarters of this sum. Nvidia’s astronomical performance alone, with nearly quadrupled revenues since early 2022, is driving market excitement.

The dichotomy is all the starker given how fragmented and specialized the industry has become. The semiconductor supply chain is now a string of micro-monopolies. TSMC and Samsung lead in foundries; ASML of the Netherlands dominates advanced lithography; DISCO of Japan supplies 85% of wafer-cutting tools. Barriers to entry remain nearly insurmountable. The capital needed for R&D and equipment dwarfs the budgets of most competitors, and geopolitical considerations now restrict Chinese firms from accessing Western markets. The incumbents also enjoy billions in subsidies from Western and Japanese governments eager to safeguard domestic production.

One might assume this oligopoly could yield easy profits, but even within micro-monopolies, dynamics are nuanced. For example, some toolmakers find their clout limited, dealing with similarly powerful clients who hold sway over prices. Others refrain from wielding their full pricing power, particularly with giants like TSMC, to maintain long-term relationships, as Chris Miller, a semiconductor expert, points out.

Yet, the scene is changing. Demand from tech titans like Microsoft and Google, both flush with resources, enables Nvidia to set steep prices for its top-tier models, with margins now reaching 75%. TSMC, sensing a more favorable climate, is planning to hike prices by up to 15% for its cutting-edge services. How this plays out will hinge on whether these players can align pricing without hampering demand.