Technology stocks extended their decline on October 15 as a sell-off deepened amid compounding concerns over persistent US inflation and the extraordinary capital expenditure demands of artificial intelligence infrastructure, according to The Guardian’s financial reporting. The market weakness coincided with news that AI company Anthropic is wooing potential investors ahead of a stock market debut, a timing that raises questions about whether the company is moving to capture valuation while investor enthusiasm for the sector remains elevated despite turbulence. US inflation rose more than expected in the latest reading, reinforcing Federal Reserve caution about rate cuts and extending the period of elevated borrowing costs that make the debt-funded capital expenditure model that underlies large-scale AI infrastructure substantially more expensive than it was two years ago.
The received wisdom
The consensus view in technology investment circles holds that the current AI spending cycle, while expensive, represents a foundational infrastructure buildout analogous to the internet’s early expansion: capital-intensive, temporarily loss-making, but ultimately generative of productivity gains that will justify the investment many times over. The comparison to the 1990s fibre-optic buildout is frequently invoked — yes, individual companies went bankrupt, but the infrastructure they laid enabled the digital economy of the following two decades. On this reading, the current sell-off is healthy volatility, a market correctly pricing higher interest rates and adjusting expectations for near-term AI monetisation without abandoning the long-run thesis. Anthropic’s IPO preparations are framed as validation: a serious safety-focused AI company accessing public markets is a sign of sector maturation, not desperation.
A different read
The fibre-optic analogy is instructive, but not in the way its proponents intend. The 1990s infrastructure buildout did indeed lay the groundwork for the next decade’s digital expansion — but it also destroyed an estimated $2 trillion in shareholder value, rendered major telecoms companies insolvent, and left the financial system with write-downs that contributed to the early 2000s recession. The lesson of the internet’s infrastructure phase is not that the technology was overhyped — it was not, in the long run — but that the financial structure through which it was built was catastrophically mispriced, and the correction was sharp, painful, and took years to work through.
The current AI infrastructure buildout has structural characteristics that should give pause to anyone drawing easy parallels. The capital expenditure involved is of a different order of magnitude. Microsoft, Google, Amazon, and Meta collectively announced AI-related capital expenditures exceeding $200 billion for 2025 alone. These are real-economy investments in data centres, electricity infrastructure, and specialised chips — not paper valuations. They are being made at a moment of elevated US inflation that keeps borrowing costs high, against a backdrop of deepening market uncertainty specifically about AI’s ability to generate sufficient returns to service the debt those investments require.
The Anthropic IPO preparation is a particularly revealing data point. Anthropic has positioned itself as the responsible alternative to OpenAI, with a Constitutional AI approach, a safety-first culture, and a corporate structure (public benefit corporation) intended to insulate it from pure profit maximisation. All of that is genuine and worth taking seriously. But approaching investors ahead of a stock market debut means Anthropic is now entering the logic of public equity markets, which require revenue growth, margin expansion, and capital returns. The tension between a safety-first mission and the quarterly earnings cadence of a publicly listed company is not theoretical — it is the same tension that has deformed OpenAI’s governance as it has moved toward commercialisation. The resolution of that tension, in every previous case, has not favoured the mission.
There is also a geopolitical dimension that the pure financial analysis tends to understate. The revelation last week that a Saudi national had allegedly been planting recording devices inside OpenAI’s San Francisco office is not an isolated incident of corporate espionage — it is a window onto the systematic great-power competition for AI capability that is reshaping the technology sector’s risk profile. AI infrastructure is increasingly dual-use, geopolitically contested, and potentially subject to export controls, state intervention, and national security review in ways that conventional technology investment has not had to price. The NPR piece on the Musk-Bezos space race offers a parallel: private capital competition in domains with enormous strategic implications tends to generate a public subsidy dynamic where profits are private and risks are socialised. The AI sector may be approaching a similar inflection.
The market’s current discomfort is therefore not irrational caution about interest rates and quarterly earnings. It is a belated recognition that AI infrastructure investment involves risks — technological, financial, geopolitical, and regulatory — that were systematically excluded from the models that generated the valuations of the past three years. Markets often see these things late and price them brutally.
What to watch
- Federal Reserve language on rate cuts: the higher-than-expected inflation print makes a December cut less likely, and every month of elevated rates increases the financial cost of the AI infrastructure buildout — watch for Fed commentary at the next FOMC meeting.
- Anthropic’s IPO valuation range and investor reception: if it prices significantly below its last private valuation round, that is a signal about where institutional money assesses the sector.
- Big Tech Q3 earnings calls, specifically the language around AI capital expenditure guidance for 2027: any reduction or hedging of forward commitments will be treated as a meaningful signal about internal confidence.
- Regulatory developments in the EU and the US on AI model disclosure requirements: increased compliance costs would further strain the economics of the buildout at a moment of already elevated interest rates.
— J