Big Tech's AI Spending Spree: Is the Bond Market the Real Bottleneck?

The stock market is showing signs of apprehension towards Big Tech, with major players like Google experiencing their worst trading day ever. Intel saw a significant decline, its largest since the Nixon administration, while Amazon and Meta also faced considerable drops this week. The Nasdaq composite index has tumbled, nearing a correction territory with a decline just over a percentage point from its peak. The underlying issue appears to be the immense cost of building the infrastructure to support Artificial Intelligence (AI), a massive expense that Big Tech's current profits are not yet justifying. AI companies are increasingly allocating more capital to data centers than their businesses are generating in revenue.

Despite this market sentiment, Big Tech does not seem to be altering its strategy. Spending on AI has already reached $1.6 trillion since 2023 and is projected by JPMorgan to hit $1 trillion annually for the next decade. This continued investment is being financed by the bond market, which, unlike the stock market, has largely remained unfazed. Lenders continue to provide debt to Big Tech to fund the development of AI technologies that are reshaping the global economy.

This situation suggests that the questions the stock market is posing – whether Big Tech will continue its spending trajectory (which it is expected to) and if AI will soon become profitable (which appears unlikely) – are the wrong ones. The critical question is how much longer the bond market can sustain this extensive spending, as this will ultimately determine the success or failure of AI initiatives.

The economics of AI involve substantial upfront investment in infrastructure such as data centers, servers, chips, power, labor, and building materials. According to Goldman Sachs, AI companies can expect to monetize this investment between three months and two years after their data centers become operational. This extended timeline necessitates significant borrowing by hyperscalers like Google, Amazon, Microsoft, and Meta to fund operations and repay investors.

These four hyperscalers alone have accounted for over a third of the S&P 500's capital expenditures in the past 12 months, according to John Higgins, chief economic adviser at Capital Economics. Goldman Sachs estimates that approximately $250 billion in bonds from hyperscalers alone will be issued this year, financing about a third of their capital expenditures, a figure expected to peak at around 35 percent next year.

However, bond market investors have finite resources, and competition for lending is intense, particularly from U.S. Treasuries which are funding government obligations like Social Security, Medicare, and military operations. The question then becomes the maximum amount of debt the market can absorb for AI companies. While an exact figure is unknowable, Goldman Sachs estimates that the market could potentially absorb $510 billion in investment-grade debt for the largest hyperscalers, more than double their projected borrowing by year-end, indicating substantial borrowing capacity remains.

The focus, therefore, shifts from the quantity of capital to the duration of its availability. Amanda Lynam, chief credit strategist at Goldman Sachs, noted that while access to capital is not a concern, a more nuanced discussion about issuance timing and pricing is expected. Private investment funds, venture capital, and private equity firms possess trillions in