According to a new industry research report from a Swiss banking research institution, the global artificial intelligence (AI) industry value chain is undergoing a structural shift. The AI infrastructure segment, represented by high-performance memory chips and semiconductor equipment, is experiencing accelerating growth, with its value creation capability and return on capital now significantly outpacing those of traditional internet hyperscale cloud service providers (Hyperscalers).

Data shows the economic profit of the global AI infrastructure segment has surged from approximately $200 billion in 2023 to a projected $1.4 trillion by 2027, representing an overall increase of 600%. In contrast, the economic profit of hyperscale cloud service providers is expected to grow only from $200 billion to $400 billion over the same period. In the global value creation ranking within the technology, media, and telecommunications (TMT) sector, the dominance of the traditional five tech giants has been broken. It is projected that by 2027, NVIDIA, Samsung Electronics, SK Hynix, Micron Technology, and Alphabet will reconstitute the top five economic profit generators in the industry, with hardware and semiconductor firms occupying four of those spots.

John Talbott, Head of the UBS Holt research team, emphasized that the current sector rotation within the technology industry is proceeding at a pace surpassing any previous tech cycle. A hardware subsector, once considered highly cyclical and commoditized, has shifted from value destruction to leading global industry value creation within just a few years. The magnitude of this shift is extremely rare in modern industrial history, indicating that the core profit dividends within the AI ecosystem are rapidly concentrating towards the hardware and infrastructure layers.

The report's analysis identifies the core reason for this structural divergence as a shift in capital returns. As global tech giants continue to ramp up investment in fixed assets and computing power for AI, the cash flow return on investment (CFROI) for traditional hyperscale cloud service providers has narrowed by approximately 200 basis points over the past two years. Concerns over whether massive capital expenditures can be translated into substantive commercial returns in a timely manner have intensified, leading to a cumulative decline of over 1% year-to-date for a major US-listed ETF that includes several traditional tech giants.

In stark contrast, the total market capitalization of the AI infrastructure segment, encompassing semiconductors, hardware equipment, and the global supply chain, has already surpassed $1 trillion. Within this, the memory chip subsector alone accounts for half of the expected profit growth for the entire infrastructure industry. Industry experts believe that following a period of weakness in 2023, memory chip companies, underpinned by rigid demand for core hardware like advanced process nodes and high-bandwidth memory (HBM), have undergone a fundamental reversal in their profit cycle, becoming the most rapidly expanding beneficiaries in terms of balance sheet growth during this AI technology cycle.