The first phase of the artificial intelligence revolution created a clear winner: Nvidia. As demand for AI computing exploded, the company’s advanced processors became the essential engines powering everything from chatbots and large language models to hyperscale data centres. Yet as the AI industry enters a new stage of development, investors are increasingly asking a different question: where will the next wave of profits emerge?
The answer may lie far beyond the companies designing AI chips. As technology groups and governments commit hundreds of billions of dollars to AI infrastructure, attention is shifting toward the manufacturing ecosystem that makes advanced computing possible. Increasingly, the beneficiaries of the AI boom may include suppliers of silicon wafers, advanced packaging technologies, semiconductor equipment, specialised materials and memory systems—the industrial foundations upon which the AI economy is being built.
Understanding this shift requires distinguishing between chips and wafers. AI chips, such as Nvidia’s graphics processing units (GPUs), perform the calculations that power artificial intelligence systems. Silicon wafers, by contrast, are the ultra-pure circular discs from which those chips are manufactured. Every advanced processor begins life as a wafer before passing through multiple stages of fabrication, packaging and testing. Chips generate computing power, but wafers provide the physical platform upon which that power is created.
The growing importance of these upstream technologies helps explain why Nvidia chief executive Jensen Huang recently said the company expects to spend around $150 billion annually within Taiwan’s technology ecosystem. More than a spending commitment, the figure represents a signal about how the AI industry expects demand to evolve. Nvidia’s reliance on Taiwan’s manufacturing network suggests the company believes AI infrastructure investment will remain elevated for years, creating opportunities far beyond chip design itself.
At the centre of this opportunity stands Taiwan. While Taiwan Semiconductor Manufacturing Company (TSMC) remains the world’s leading producer of advanced semiconductors, the island has evolved into a broader AI infrastructure hub. Companies including Foxconn, Quanta and Wistron play critical roles in assembling servers, integrating memory systems and deploying the hardware that supports AI applications worldwide.
For investors, Taiwan demonstrates an increasingly important reality: AI is no longer simply a software or chip-design story. It is becoming an industrial ecosystem story.
One of the clearest examples is advanced packaging. Once viewed as a routine manufacturing process, packaging has become one of the industry’s most valuable capabilities. Modern AI processors combine multiple computing chips with high-bandwidth memory inside a single package, enabling them to process vast amounts of information at extraordinary speed. The sophisticated technologies required to integrate these components have become one of the industry’s most significant bottlenecks.
This matters because bottlenecks often create pricing power. Companies controlling scarce manufacturing capacity frequently enjoy stronger margins and more resilient demand than participants in less constrained segments of the value chain. As AI infrastructure spending accelerates, advanced packaging providers may become some of the most strategically important businesses in the semiconductor industry.
A similar trend is emerging in silicon wafers. Industry estimates suggest Chinese manufacturers could account for approximately 32% of global 12-inch silicon wafer manufacturing capacity by 2026, up from only 3% in 2020. Although China remains behind global leaders in advanced semiconductor fabrication, the expansion highlights the growing importance of upstream manufacturing assets in the AI era.
Japan’s leading wafer producers, including Shin-Etsu Chemical and SUMCO, provide another illustration. Despite attracting far less attention than Nvidia, these companies occupy critical positions within the semiconductor supply chain. Their products form the foundation of advanced chip manufacturing, making them indispensable participants in the AI economy.
The same dynamic extends to semiconductor equipment makers, specialised chemical suppliers, substrate manufacturers, memory producers and other industrial companies that rarely feature in headlines. As AI deployment expands, many of these businesses could become increasingly valuable because they control technologies and production capacity that cannot easily be replicated.
The implications extend beyond investors. Governments increasingly view semiconductor manufacturing, packaging technologies and critical materials as strategic assets. The result is a shift away from supply chains built solely for efficiency and toward systems designed for resilience, security and technological independence.
Unlike many previous digital revolutions, artificial intelligence is proving remarkably capital-intensive. Building AI systems requires not only software and algorithms but also factories, packaging facilities, energy infrastructure, specialised materials and highly sophisticated manufacturing equipment. The industry is beginning to resemble sectors such as energy, aerospace and advanced manufacturing, where ownership of physical infrastructure can be as important as technological innovation.
For much of the past three years, investors focused on identifying the companies designing the most powerful AI chips. That strategy proved highly profitable. The next stage of the AI revolution may require looking deeper into the supply chain.
Nvidia may remain one of the dominant beneficiaries of artificial intelligence. But the next generation of AI winners could emerge from the less visible businesses producing wafers, packaging technologies, materials and manufacturing equipment. In the AI economy, the most valuable asset may not be the chip itself, but the industrial ecosystem that makes the chip possible.
