The scale of current investments in artificial intelligence has reached a fever pitch, with hyperscalers ramping up capital expenditures to historic levels. Recent data from Goldman Sachs highlights the intensity of this boom, noting that AI related issuers now make up a quarter of all new US investment grade corporate debt. For many of the world’s most creditworthy companies, spending has surged by at least thirty five percent annually for ten consecutive quarters. However, beneath these staggering numbers lies a growing tension between genuine structural demand and speculative fervor.
Lu Zhang, the founder and managing partner of Fusion Fund, suggests that the signal to noise ratio in the current market is dangerously low. While she acknowledges that big tech firms are determined to secure their computing power to avoid being left behind by competitors, she warns that much of the apparent demand may be illusory. Financial red flags have already begun to emerge, with giants like Microsoft, Alphabet, Amazon and Meta seeing capital spending grow significantly faster than their revenues. In some cases, free cash flow has plummeted or turned negative, leading critics to wonder if some of this growth is merely circular vendor financing rather than organic market expansion.
To cut through the hype, Zhang employs a rigorous three part test when evaluating AI ventures. Rather than focusing on model quality or massive spend totals, she looks for companies that possess curated industry specific data, optimized architectures that lower operational costs beyond mere training, and established partnerships with key industry players who control essential workflows. Without these pillars, Zhang argues that any perceived competitive advantage is likely temporary. She believes true durability is found not in headline revenue figures—which can be inflated in a seller’s market—but in actual budget allocations from non tech sectors like healthcare and insurance.
Beyond the balance sheets, Zhang identifies governance failures as one of the primary risks that could derail the AI thesis entirely. High profile security breaches and pauses in model development due to safety concerns illustrate how quickly regulatory or ethical lapses can freeze progress. As global powers discuss international safety standards and emergency hotlines to manage AI incidents, the industry faces a reckoning where technical capability must finally align with responsible oversight to ensure long term stability.
