Tech investors are facing a new form of disruption.

Over the next few years, the huge spending proposed by established tech companies on artificial intelligence infrastructure is astonishing. In 2025 alone, large technology companies, including Amazon (AMZN.US), Microsoft (MSFT.US), Alphabet (GOOGL.US), Meta (META.US), and Apple (AAPL.US), are expected to spend more than 200 billion dollars, almost double what they spent in 2021 (the year before the generative AI chatbot ChatGPT debuted). The increase in capital expenditure is almost entirely due to efforts to build production capacity for generative artificial intelligence.

This highlights the key difference between the surge in AI investment and the tech boom of the previous 20 years: today's investments are focused more on hardware than on software—which is clearly more capital intensive.

If the economy slows down and the business prospects of these tech companies deteriorate, their executives may think twice about these ambitious, completely discretionary spending plans, making it difficult for investors to calculate the possible returns of this emerging technology.

The rise of hardware investment

In the first two decades of this century, software engineers disrupted industry after industry with flexible, scalable, and low fixed cost business models. A small group of savvy entrepreneurs started from scratch, quickly became successful with early prototype products, then developed a series of strategic pivots — Amazon, Netflix (NFLX.US), and many social media companies.

Fast forward to the age of generative artificial intelligence, and the storyline changed dramatically. New business models usually revolve around very smart, complex, and expensive machines that require significant amounts of energy to run and often take a long time to manufacture. For example, TSM.US's foundry in Arizona costs 40 billion dollars, and commercial production will not begin until 2025, four years after construction began.

Importantly, investments in artificial intelligence usually take years to pay off. At the same time, many factors may adversely affect the value of AI infrastructure, including concerns relating to business confidence and cost inflation, as well as regulatory barriers and geopolitical tensions affecting where companies do business. This means that tech investors can no longer easily ignore top-down macroeconomic concerns.

Hardware investment is capital-intensive, and capital costs are sensitive to the macroeconomy

Unlike startups in the software sector, AI startups are often capital-intensive, which makes them highly sensitive to market conditions and financing channels. Most of these young companies rely on private capital, and in recent years, many venture capitalists have been eager to provide private capital. In the first half of 2024, investments in fields related to artificial intelligence and machine learning accounted for nearly half of all US venture capital.

These investments are often huge. In October of this year, OpenAI raised $6.6 billion in equity capital from eight investors and raised $4 billion in debt financing from nine lenders. The average amount of these transactions is over 0.5 billion dollars.

At a time when the S&P 500 index has reached a new high, the growth rate of the US economy is higher than the trend level, and the inflation rate is falling, checks of this scale can be issued. But what happens when the economy inevitably weakens and the price of listed stocks falls? Or what if the cost of capital in the US continues to be high?

At that time, AI startups may find it more challenging to fund their ambitious visions, which in turn may hinder the growth and pace of innovation in the broad AI ecosystem. Of course, this could reduce the need for artificial intelligence infrastructure that big tech companies have already invested hundreds of billions of dollars in.

The cyclical nature of hardware investment is more obvious

The hardware business also showed more cyclical characteristics than the software business. This is because they cannot rely on continuous adjustments to meet changes in customer needs, because creating new products requires significant investment and manpower. This means that these companies are subject to traditional inventory cycles: when demand exceeds current supply, inventory decreases, prices rise, and vice versa. Therefore, unlike flexible software companies, hardware companies will try to expand or reduce production capacity in a short period of time. As a result, the quantity and price of hardware usually fluctuate according to economic conditions.

Notably, semiconductor sales have been positively correlated with the manufacturing PMI for decades. That relationship began to break down in 2022, as the euphoria of artificial intelligence really took off. If the historical pattern remains the same, this may mean that the global semiconductor sales boom is long overdue. This is just one example of why tech investors might need to be as macro-aware as other investors.