Is the AI Money Machine Running Out of Steam?

The rapid expansion of the artificial intelligence sector has captivated global markets, but recent volatility in South Korea provides a sobering reminder of the risks inherent in this boom. John Stepek, senior reporter at Bloomberg and author of the Money Distilled newsletter, recently sat down with Neil Callanan, Bloomberg’s private companies managing editor, to pull apart the complex web of capital driving the AI industry.

The Anatomy of an AI Value Chain

Stepek and Callanan break down the AI industry into a functional hierarchy, moving from the foundational hardware to the end-user applications. At the top of the chain sit the equipment manufacturers like ASML, followed by the essential chip foundries like TSMC and the design leaders such as Nvidia (NASDAQ:NVDA). These chips eventually populate the massive data centers managed by hyperscalers, the modern equivalent of internet hotels, which provide the infrastructure for AI models developed by firms like OpenAI and Anthropic.

This structure is increasingly energy-intensive. "They need energy, and so we are seeing a massive boom in energy, and people are talking about needing up to 300GW of additional energy power by 2030," noted Callanan. This demand is rippling through the economy, fueling growth in construction and energy infrastructure, but it also creates a dependency on end-users being willing to pay the high costs required to sustain these massive capital expenditures.

The full discussion can be found on Bloomberg Podcast's YouTube channel.

The Risk of Circular Financing

A central theme of the discussion is the rise of circular deal-making, which Callanan compares to the speculative excesses of the 1990s fiber optic boom. When chip designers and model developers engage in complex, cross-linked funding arrangements, it can create misaligned incentives and a false impression of market demand. "These circular deals can create a false impression of demand. You may think that all these companies are generating massive revenues, but if it is all just moving around and sloshing around and if something falls out of bed, then there could be wider implications," warned Callanan.

StartupHub.ai data provides a benchmark for this competitive landscape, with OpenAI holding a score of 84/100, while competitors like Anthropic (76/100), Arena (71/100), Prometheus (64/100), Subquadratic (60/100), and One AI (50/100) continue to vie for market share. As these companies seek to scale, they are increasingly turning to the credit markets, with some firms like CoreWeave seeing their credit default swap spreads rise toward record levels as investors become more discerning.

China and the Cost of Tokenization

The conversation also addresses the growing competitive threat from Chinese AI developments. Callanan suggests that China’s advantage in cheaper electricity makes the tokenization process more cost-effective, which could disrupt the business models of Western incumbents. "If you have these cheaper models coming out of China and people start switching to that, then what happens with the names we are all familiar with like OpenAI and Anthropic?" Callanan asked. This potential shift is forcing investors to reevaluate the valuations of AI-adjacent companies, which have seen massive gains this year.

Ultimately, while the AI sector remains a significant driver of economic growth, the era of unquestioning capital allocation appears to be fading. As the credit markets push back and regulators monitor for potential systemic risks, the AI industry must prove that its underlying business model can generate sustainable value beyond the initial wave of infrastructure investment.