US stocks slumped as fears over Big Tech shook Wall Street, continuing a period of volatility that has left investors reassessing the valuations underpinning the AI investment supercycle. The selloff came as questions mounted about whether there is an AI stock market bubble about to burst, and as President Trump prepared to meet AI company leaders to discuss US investment in their firms — a meeting that itself signals how intertwined the AI industry’s financial interests have become with the highest levels of government patronage. Against this backdrop, Anthropic co-founder Daniela Amodei warned that AI development without humans is a risk that requires governance intervention.
The received wisdom
The dominant techno-optimist view, which has driven AI-related equity prices to extraordinary levels over the past three years, holds that large language models and generative AI represent a genuine productivity revolution — the kind of general-purpose technology that, like electricity or the internet, will eventually transform virtually every sector of the economy and justify today’s valuations through future earnings that we simply cannot yet fully see. On this view, the current selloff is a normal correction within a secular bull market, and investors who panic are repeating the mistake of those who sold internet stocks in the late 1990s correction only to miss the decade-long appreciation that followed.
The parallel to the late-1990s internet boom is one that sophisticated bulls use deliberately: yes, many dot-com companies were worthless, but Amazon and Google were not, and the underlying technology genuinely did transform everything. The argument is that AI’s value will similarly prove out over time, and that the enormous capital expenditure currently flowing into data centres, chips, and model training represents a rational bet on a transformational technology.
A different read
The problem with the internet-era parallel is that it is being selectively invoked. Yes, the internet ultimately created enormous value — but it also destroyed enormous investor wealth in the interim. The Nasdaq fell roughly 78% from peak to trough between 2000 and 2002. Many companies that were genuinely working on valuable technologies saw their valuations crater by 90% or more before eventually recovering. The parallel, if apt, cuts both ways.
More substantively, there are real questions about the current AI investment cycle that are not simply the anxieties of technological sceptics. The BBC’s analysis of the bubble question tracks the central puzzle: enormous capital is flowing into AI infrastructure, but the revenue models that would justify that capital remain substantially unproven at scale. Microsoft, Google, and Amazon have all made enormous bets on AI integration; some of those bets are generating real incremental revenue, but the ratio of investment to demonstrated return remains unfavourable by historical standards.
The US stocks slump is also worth reading in the context of rising interest rates and the broader fiscal environment. AI valuations were built in a zero-rate world. In a world where the risk-free rate has returned to historically normal levels, the discount rate applied to uncertain future cash flows matters enormously. Companies whose investment thesis depends on revenues that may not materialise for five to ten years are systematically disadvantaged in this environment, regardless of the underlying technology’s genuine promise.
The political dimension is equally worth attending to. Trump’s meeting with AI executives is not primarily about technology governance — it is about managing a politically influential industry whose CEOs have been cultivating relationships with the administration. This is not new in American history: tech companies have become as politically integrated as defence contractors, banks, and energy companies before them. The risk is that public policy toward AI — including the governance frameworks that Anthropic’s Daniela Amodei says are urgently needed — gets shaped more by the political interests of incumbent large-cap AI companies than by any coherent public interest analysis.
Amodei’s warning about AI development proceeding without adequate human oversight is substantively serious. But it is worth noting that major AI labs calling for regulation consistently tend to favour regulatory frameworks that are expensive to comply with — and therefore serve as barriers to entry that protect established players. This is the same political economy that pharmaceutical companies pioneered with FDA approval requirements, that banks pioneered with Basel capital rules, and that telecoms companies pioneered with spectrum licensing. The form is “safety and governance”; the function is often “protecting the incumbents from competition.” That does not mean the governance concern is wrong — it means it should be evaluated with clear eyes rather than deference to the companies making the argument.
What to watch
- The upcoming earnings cycles for the major AI-exposed companies: Microsoft, Google, Nvidia, and Amazon will all need to demonstrate that their AI capital expenditure is translating into revenue at the pace that current valuations require
- Whether the Trump administration uses its AI executive meetings to announce any policy framework — tax treatment of AI investment, export controls on chips, domestic manufacturing incentives — that changes the investment calculus
- Nvidia’s stock price specifically: it has become something of an index for AI sentiment, and its movements tend to lead broader Big Tech trends
- The governance question: whether any meaningful AI oversight framework emerges from either Congress or the executive branch, and whether it reflects serious public interest analysis or captures by industry lobbying
— J