The buildout of the US railroads between 1860 and 1900 drew the largest capital expenditures in US economic history, creating not only an enduring transport infrastructure but also the institutions of modern Wall Street including the stock market. Only today’s buildout of AI infrastructure – data centers, networks, software and related facilities – is expected to surpass that historic investment.[1] Alongside the tech giants, newly emerged start-ups led by Open AI are pouring hundreds of billions of dollars into capital expenditures on AI. Google, Amazon, Microsoft, and Meta alone have committed to spending $725 billion in 2026, a drastic increase from $410 billion in 2025, racing against each other to secure dominance.[2] OpenAI, which expects to be in the red until 2029,[3] is planning to spend $600 billion for computing infrastructure by 2030.[4] This massive investment has been driving the U.S. stock market and the growth of U.S. GDP.
To fund their competing AI infrastructures, tech giants, no longer able to self-finance, are going into the domestic and overseas debt market and depleting their enormous cash reserves. As of March 2026, Alphabet has an outstanding debt a $77.501 billion, up from a mere $10.88 billion in 2024;[5] Amazon is carrying $65.6 billion in debt;[6] and Oracle holds debt of over $100 billion, driven by AI infrastructure spending. Tech companies’ layoffs have been due not to AI automation but to funding their AI infrastructure expenses.
If demand actually turns out to justify this massive capital expenditure, the tech companies’ share prices will continue to balloon – the combined value of the seven tech companies is $23.14 trillion, which is roughly 45% of the S&P 500 [7] and accounted for 92% of U.S. GDP growth in the first half of 2025.[8] However, despite the exuberant stock prices and the promise and hope surrounding AI, the immediate benefits for corporations’ bottom lines from the use of AI are remain murky.
U.S. companies spent $37 billion in 2025 on generative AI alone, but they are struggling to see immediate returns (ROI) on their AI investment.[9] Nvidia vice president Bryan Catanzaro said, “AI compute now costs more than the employees using it, making AI more expensive than human labor.”[10] Google, fully aware that corporations are trying to justify the skyrocketing cost of AI, has recently released its new, more “cost effective” Gemini AI model.[11] According to a survey done by the Harris Poll commissioned by Dataiku, “71% of global chief information officers said their AI budgets would be frozen or cut if value from AI couldn’t be demonstrated within two years.”[12] According to a recent Financial Times article, the capital investments by hyperscalers are expected to rise by 20 percent between 2025-2030, but revenues are generously calculated to grow by only 15 percent.[13] In light of all this, tech giants desperately need demand that generates revenue. Otherwise, they may face the same reality that confronted railroad investors during the 1890s: overcapacity, and all that comes with it.
For this moment, Anthropic’s release of Mythos did a spectacular job of generating demand in the cybersecurity realm. Consequently, the company is hurriedly searching for computing power, expanding its alliances with competitors Google, Amazon, and xAI and leasing data center capacity wherever it can. While, at least for the moment, Anthropic’s Claude could be the best AI model, it is forced to rely on its competitors for computing infrastructure that it does not have yet. In turn, paradoxically, almost half of Google and Amazon’s profit this quarter actually came from their investments in Anthropic.[14]
Google, Microsoft, Amazon and Nvidia all invested in OpenAI and Anthropic and are now betting on their upcoming trillion-dollar IPOs to expand their ever-growing data centers.[15] This circular financing holds up the debt-ridden AI industry and has momentarily staved off the problem of overcapacity. But how long can this last?
No one can predict, but what is clear is that there is a widening gap between demand and the tech giants’ hugely inflating infrastructure expenditures and infrastructural capacity. The result is likely to be price wars and, ultimately, something harsher.
If tech companies are not able to generate greater demand – soon – then the circulation of capital to the AI sector via the stock market, private equity, and investment banks, could slow or come to a sudden stop. This in turn would propel a gigantic shockwave through the financial system and, indeed, through global capitalism. Time is ticking.
Shinjoung Yeo & Dan Schiller
[1] See “The AI Capex Boom: Bubble or Infrastructure Supercycle?” LongYield, March 5, 2026, .
[2] “Alphabet, Amazon tap overseas debt markets to fund AI infrastructure push,” Reuters, May 11, 2026.
[3] Jeremy Laird, “OpenAI’s own forecast predicts $14 billion loss in 2026 but Nvidia-style $100 billion revenues by 2029 according to new report,” Yahoo!Finance, January 21, 2026,
[4] Ashley Capoot and Kate Rooney, “OpenAI resets spending expectations, tells investors compute target is around $600 billion by 2030,” CNBC, February 20, 2026.
[5] “Alphabet Long Term Debt 2012-2026,” Macrotrends.
[6] Rich Duprey, “AI’s Coming Trillion-Dollar Hangover: Amazon Leads Hyperscalers Back to the Debt Well,’ Yahoo!Finance, March 10, 2026.
[7] Sinchita Mitra, “AI Stocks near 45% of S&P 500 weight, Goldman Sachs says,” Seeking Alpha, April 22, 2026.
[8] Nick Lichtenberg, “Without data centers, GDP growth was 0.1% in the first half of 2025, Harvard economist says” Fortune, October 7, 2025.
[9] Tim Keary, “AI Pilots Still Don’t See Returns. Here’s Why,” Forbes, April 30, 2026; Thomas H. Davenport and Laks Srinivasan, “7 Factors That Drive Returns on AI Investments, According to a New Survey,” Harvard Business Review, March 17, 2026.
[10] Sherin Shibu, “Nvidia VP Says AI Costs ‘Far’ More Than Human Employees,’ Yahoo!Finance, April 29, 2026.
[11] Yifan Yu, “Google says new AI model could save companies billions in token costs,” Nikkei Asia, March 20, 2026.
[12] Makenzie Holland, “Most CIOs regret AI vendor, platform decisions: report,” Yahoo!finance,February 12, 2026,
[13] Joachim Klement, “The impossible maths of the AI boom,” Financial Times, May 20, 2026.
[14] Eva Royburg, “Google and Amazon’s biggest profit driver last quarter was their Anthropic stakes—which they haven’t sold,” Fortune, April 30, 2026.
[15] Joachim Klement, “The impossible maths of the AI boom,” Financial Times, May 20, 2026.
