FORWARD THINKING | ARTICLE — 4 min read
As capital expenditure projections for artificial intelligence (AI) infrastructure dwarf even the dot-com era's excesses, investors are right to ask whether history is repeating itself. In Artificial Intelligence: Not a Bubble… Yet, Pamela Hegarty, Senior Portfolio Manager, Environmental, Social and Governance (ESG) Champion, and Derek Glynn, Portfolio Manager, assess the evidence and arrive at a carefully measured verdict: not yet — but active vigilance is essential.
The infrastructure arms race: echoes of the 1990s
The scale of capital now flowing into AI data centre infrastructure is extraordinary. McKinsey forecasts $6.7 trillion of cumulative capital expenditure (capex) for data centre infrastructure between 2025 and 2030 [1] — a figure that towers over the estimated $500 billion invested in fibre optic and wireless networks in the five years following the US Telecommunications Act of 1996 [2]. By 2002, only 2% of that long-distance telecoms capacity was actually in use [2].
Spending projections for AI vary considerably. Bank of America placed cumulative AI capex at $1.2 trillion by the end of 2030 as of early October 2025 [3], while the leading provider of graphics processing units (GPUs) estimated $3–4 trillion of total infrastructure investment in the year 2030 alone — not a cumulative figure [4].
This arms-race dynamic — where the imperative to lead in AI model development compels vast pre-emptive capital deployment — echoes the behaviour of the late 1990s. Not all participants will succeed, and some of the capacity being built today may, like the telecoms networks of that era, remain significantly underutilised for years. The desire to be first to market is creating risks that investors cannot afford to ignore.
Why this time is structurally different
The comparison with the dot-com era, while instructive, has important structural limits. During the Internet and telecoms bubble, the companies bearing the brunt of infrastructure investment were primarily debt-funded and lacked stable, cash-generative business segments to sustain them through the cycle. Today's leading cloud service providers (CSPs) are large, financially robust organisations that have, to date, self-funded their AI capital expenditure from operating cash flows [4].
Adoption dynamics also differ meaningfully. The infrastructure to deliver AI-based applications — high-speed internet access, smartphones, connected devices — already exists at scale. ChatGPT reached 800 million weekly active users in under three years from its public launch in November 2022 [4]; comparable internet penetration took 13 years [4]. Yet enterprise deployment remains decidedly early-stage: while 78% of enterprises have adopted AI in at least one business function, only 16% have deployed it across five or more [6]. This implies a substantial monetisation runway ahead — a pattern consistent with earlier technology cycles. Cloud computing and software-as-a-service (SaaS), for instance, began their enterprise roll-out in 2006 and expanded steadily without a dramatic digestion period [4].
Valuations and the speculative temperature check
Public-market valuations are elevated relative to the decade before the AI boom, but they remain far below the extremes reached in 1999 and 2000. During the dot-com bubble, the median price/sales multiple of technology sector initial public offerings (IPOs) spiked to 43x in 1999 and 49.5x in 2000 [7]. Today's higher price-to-earnings (P/E) multiples are underpinned by genuinely stronger profit margins and returns on equity (ROE) — not the same untethering from fundamental value that characterised that peak.
The IPO pipeline itself is telling. Between 1991 and 2000, an average of 183 technology sector IPOs were launched each year, reaching 370 in 1999 alone [7]; the past decade has averaged just 37 per year, with only 14 completed in 2024 [7]. Most AI companies are choosing to remain private for longer, concentrating speculative valuations in private equity and venture capital rather than public markets. As of July 2025, there were 793 "unicorns" — private companies valued at more than $1 billion — in the United States [8], with 36 new technology unicorns created in the first half of 2025 alone [9]. Public equity investors, for their part, retain a healthy degree of scepticism about AI economics that was conspicuously absent during the late-1990s hype cycle.
Risks that warrant active monitoring
The absence of a bubble today does not mean the risks can be dismissed. Several factors merit careful, ongoing attention from investors:
- Infrastructure overbuilding: the arms race in AI model development may generate excess capacity, producing a digestion period similar to the telecoms collapse of the early 2000s.
- Debt financing: AI-related issuers accounted for $141 billion in corporate credit issuance in 2025 year-to-date [5], and the asset-backed securities market has seen $20 billion of data centre deals since the beginning of 2024 [5]. The reported use of GPUs — a fast-depreciating asset — as collateral in some structures introduces additional fragility.
- Circular revenue relationships: arrangements where a component supplier invests in its customer and then recognises revenue from sales to that same customer present both accounting and systemic risks.
- Regulatory gaps: the late-1990s bubble was partly fuelled by deregulation; the current absence of comprehensive AI regulation could compound risks in a structurally similar way.
Navigating this environment calls for a disciplined, active investment approach — one grounded in discounted cash flow (DCF) and return on invested capital (ROIC) analyses, diversified across multiple participants in the AI value chain, and continuously alert to the leading indicators of bubble formation [4].
Looking ahead
The AI theme is not yet in bubble territory, but the investment environment demands ongoing vigilance. Valuations for the leading AI companies remain reasonably grounded, yet spending cycles of this magnitude carry inherent risks that can crystallise quickly. Industry consolidation and disruption are likely over time as winners emerge from the current arms race. In this context, bottom-up fundamental research and a disciplined focus on valuation remain the most reliable tools for navigating what is, by any measure, a defining technology transition.
Download the full paper: Artificial Intelligence: Not a Bubble… Yet
Sources
- McKinsey, 2025: "The future of US hyperscale data centers" (Goldman Sachs; S&P Capital IQ; McKinsey Data Center CAPEX TAM & Demand model)
- Robert Litan, Brookings Institute, 2002: "The Telecommunications Crash: What To Do Now?"
- Bank of America, October 2025: AI capital expenditure forecast, cited in Hegarty, P. and Glynn, D., BNP Paribas Asset Management, October 2025
- Hegarty, P. and Glynn, D., BNP Paribas Asset Management, October 2025: "Artificial Intelligence: Not a Bubble… Yet" (document code P2508029)
- Goldman Sachs, October 2025: "AI Capex Turns to the Debt Side"
- McKinsey & Company, March 2025: "The state of AI: How organizations are rewiring to capture value"
- Ritter, J.R., University of Florida, September 2025: "Initial Public Offerings: Updated Statistics", table 4a
- Visual Capitalist, August 2025: "Visualizing Unicorns by Country in 2025"
- TechCrunch, 2025: "At least 36 new tech unicorns were minted in 2025 so far"
Disclaimer
Please note that articles may contain technical language. For this reason, they may not be suitable for readers without professional investment experience. Any views expressed here are those of the author as of the date of publication, are based on available information, and are subject to change without notice. Individual portfolio management teams may hold different views and may take different investment decisions for different clients. This document does not constitute investment advice. The value of investments and the income they generate may go down as well as up and it is possible that investors will not recover their initial outlay. Past performance is no guarantee for future returns. Investing in emerging markets, or specialised or restricted sectors is likely to be subject to a higher-than-average volatility due to a high degree of concentration, greater uncertainty because less information is available, there is less liquidity or due to greater sensitivity to changes in market conditions (social, political and economic conditions). Some emerging markets offer less security than the majority of international developed markets. For this reason, services for portfolio transactions, liquidation and conservation on behalf of funds invested in emerging markets may carry greater risk.
The sustainable investor for a changing world
Topic tags: Equity | Thematic investing | Global | Macroeconomics