The AI Bubble Is Different — Corporations, Not Retail, Inflate It
The Atlantic calculates a $27tn surge in AI‑linked valuations and argues corporate capex — not retail mania — is inflating the market. That shift changes who wins and how regulators should respond.
What happens when the money behind a market mania is not household investors but the world’s largest corporations? That is the question animating debate after The Atlantic calculated a roughly $27 trillion rise in the market value of AI-linked firms over the past three years — roughly 36% of the U.S. stock market today The Atlantic. The piece argues this is “no ordinary bubble,” driven by corporate capex and balance-sheet bets rather than retail speculation.
Bold claim, unusual mechanics: The Atlantic’s framing matters because the source of capital changes the risk. Unlike the dot‑com era, where margin‑financed retail positions and IPO froth mattered, the current cycle centers on hyperscalers and sovereign-scale allocators spending to build AI stacks. That spending shows up as large infrastructure commitments — estimates put combined AI infrastructure capex from Amazon, Microsoft, Google and Meta in the hundreds of billions Corewire — and it transforms a technology boom into a macroeconomic force.
Real growth or overhang? Proponents argue the scale is justified. SoftBank’s Masayoshi Son told Reuters on July 14 that AI will need roughly $5 trillion a year by 2040 and dismissed bubble talk as ignorant of what AI requires: “Asking if AI is a bubble is absurd. I don’t think people who ask that question know what AI is about,” he said Reuters. That’s a capital‑intensive vision: buy chips, build data centres, sign long‑term power and real‑estate leases.
Yet regulators and allocators are uneasy. Taiwan central‑bank governor Yang Chin‑long warned on July 9 of the “possibility of an AI bubble,” saying the worry is not the technology itself but reckless leverage around it: “AI is driven by real growth potential, but it’s the possibility of over‑expansion via over‑leveraging that concerns us,” he told Reuters Reuters. Hedge funds in China have likewise used phrases like “super bubble” and suggested a collapse point may be near, according to Bloomberg reporting Bloomberg.
The macro effects are visible. The Wall Street Journal has reported AI‑driven data‑centre demand is lifting costs across supply chains — from electricity to smartphones — contributing to what it calls a third wave of inflationary pressure WSJ. BlackRock’s Investment Institute frames the question more conditionally: “Are we in an AI bubble? We think the answer depends on whether AI can turn today’s scarcity into tomorrow’s abundance,” the institute told Swissinfo, underscoring that much of the market’s optimism hinges on unproven productivity gains Swissinfo.
That conditionality is the crack in the bullish case. Corporate capex can sustain valuations for a long period, but it is not immune to mispricing. Capital‑intensive projects have long lead times and high fixed costs; if the anticipated productivity lifts fall short, hyperscalers could face asset write‑downs and slower revenue growth while carrying heavy balance‑sheet obligations. The Atlantic’s point — that this bubble is inflated by balance sheets, not margin debt on Main Street — implies different policy and market responses when cracks appear.
Critics of the “bubble” framing also have a point: spending on AI creates tangible capacity and, in some sectors, immediate operational improvements. Son’s $5‑trillion forecast is one end of a spectrum; central banks’ warnings are the other. Both can be true at once: the industry may deliver long‑term gains while producing short‑term financial instability if expectations run ahead of measurable returns.
Who wins and who loses depends on timing. If AI delivers productivity gains at scale, the hyperscalers and their investors will justify huge up‑front costs and likely dominate the stack. If growth disappoints or financing costs remain elevated, the immediate losers will be the firms that overcommitted to fixed infrastructure and the lenders that underpriced that risk — and public policymakers, who must manage spillovers into energy and labour markets.
The question now is empirical: will the coming quarters deliver revenue growth that matches the $27‑trillion underwriting of expectations, or will rising costs and slower returns force a reprice? Watch hyperscaler capex announcements, quarterly revenue conversion rates for AI products, and central‑bank commentary — each will show whether this is a different kind of bubble or simply a larger, familiar one.
