AI & Capital: A Pragmatic Assessment

The current obsession with Artificial Intelligence is, predictably, attracting capital. It is rarely wise to follow the herd, but to ignore a significant shift in investment is equally foolish. The promise, as always, is of exponential returns. The reality is likely to be more…selective. A dispassionate assessment suggests two primary avenues for potential profit: the manufacturers of the necessary hardware, and those who provide the computing power itself. To suggest these are the only beneficiaries would be naive, but they represent the most direct exposure to this unfolding phenomenon.

Six companies appear reasonably positioned to capitalize on this trend, though “capitalize” is a word often divorced from genuine value. A cautious investor should not mistake enthusiasm for insight.

The Hardware Manufacturers

The fabrication of the machinery that powers this “intelligence” is, at present, the most tangible area of investment. These companies are receiving payment now for components essential to AI workloads. Nvidia is the most obvious participant, and its valuation reflects that. However, to focus solely on Nvidia is to overlook the supporting players. Broadcom and Taiwan Semiconductor Manufacturing, while less glamorous, are equally crucial.

Nvidia’s graphics processing units (GPUs) are, undeniably, the current standard for AI training and inference. They are also, it must be noted, expensive. One pays a premium for performance, and whether that premium is justified depends on the individual investor’s risk tolerance.

Broadcom offers a more pragmatic alternative. By partnering directly with clients to design application-specific integrated circuits (ASICs), they circumvent the need for generalized computing power. This specialization, while less flexible, can result in a more cost-effective solution. The expectation of a doubling in AI semiconductor revenue by early 2026 is not a guarantee, but it suggests a viable challenge to Nvidia’s dominance.

Taiwan Semiconductor Manufacturing (TSM) occupies a particularly interesting position. It is, in effect, the factory floor for companies like Broadcom and Nvidia. They design the components; TSM manufactures them. This intermediary role offers a degree of insulation from the inherent risks of AI development itself. As long as demand for these chips remains high – Nvidia projects a rise to $3 to $4 trillion annually by 2030, a figure that should be treated with healthy skepticism – TSM is likely to prosper.

The ultimate consumers of these chips are, predictably, the cloud computing providers.

The Cloud and the Rental of Intelligence

Cloud computing represents a shift in capital expenditure. Rather than investing in physical infrastructure, developers now rent computing power from those who possess it. This model is efficient, but it also creates a dependency. Amazon, Alphabet, and Microsoft are the dominant players in this space, and their continued growth is, to a degree, assured. The question is not if they will grow, but at what rate, and at what cost.

Amazon Web Services (AWS) was the pioneer, but its size now hinders rapid expansion. Alphabet’s Google Cloud is growing faster, but lags in profitability. The pursuit of growth, without a corresponding increase in returns, is a common failing. Microsoft Azure, however, appears to be striking a more favorable balance. A reported growth rate of 39% in Q2 FY 2026 is impressive, though it is worth noting that Microsoft has also been utilizing some of this capacity internally.

All three companies are investing heavily in data centers, and rightly so. The cloud is, at its core, a rental agreement. A continuous stream of revenue, secured by monthly payments. The replacement of computing units will be an ongoing expense, but the infrastructure – the land, the utilities, the buildings themselves – represents a fixed cost. Once sufficient capacity is established, profitability should increase significantly. This, at least, is the theory. Whether it will materialize as predicted remains to be seen. The allure of long-term investment should not blind one to the inherent uncertainties.

2026-02-09 08:02

The Hardware Manufacturers

The Cloud and the Rental of Intelligence

Read More