AI investing: boom, bubble, or crash in the making?
- TEBI
- Aug 18
- 7 min read
Updated: Aug 20
The AI investing frenzy has reached that peculiar stage where even the person leading the revolution is urging caution. OpenAI's Sam Altman warned last week that smart people are getting "overexcited" about AI, and the market is now a bubble. At the same time he still maintains that AI is "the most important thing to happen in a very long time".
That contradiction captures the essential paradox of AI investing: transformative technology and speculative excess often go hand in hand.
History offers a sobering lesson: revolutionary technologies that transform civilisation don't necessarily transform investor portfolios. Steam railways, automobiles, and the internet all reshaped society while delivering mixed results for those who funded them. The benefits typically flow to consumers, not shareholders.
We only see bubbles clearly after they've burst
Robert Shiller noted in Irrational Exuberance that bubbles are "often conclusively recognised only after they have 'popped'." This retrospective clarity creates an awkward dilemma for investors trying to navigate the present moment.
Current AI investment exhibits classic bubble indicators. Palantir trades at nearly 600 times earnings — stratospheric by any standard. CB Insights reports that AI captured 50% of all venture capital dollars in Q2 2025, totaling $47.3 billion across 1,400 transactions.
Yet AI differs from pure speculation. Unlike many dot-com companies that lacked viable business models, today's AI leaders demonstrate robust profitability. Microsoft's Azure grew 39% year-over-year, and Meta generated $71 billion in net profit while investing heavily in AI infrastructure.
Current AI company valuations (18 August, 2025)
Company | Market cap | P/E ratio (TTM) | Revenue growth (YoY) | Recent performance |
Nvidia (NVDA) | $4.44T | 58.73 | +69% | Beat revenue expectations |
Microsoft (MSFT) | $3.87T | 38.13 | +18% | Azure +39% growth |
Amazon (AMZN) | $2.61T | 35.25 | +13% | AWS guidance concerns |
Meta (META) | $2.0T | — | +22% | Record highs post-earnings |
Tesla (TSLA) | $1.07T | 191.59 | -12% | Missed profit expectations |
Palantir (PLTR) | $420.1B | 588.76 | +48% | Beat adjusted EPS |
The table above shows how current AI companies stack up financially—note the extreme valuations alongside genuine revenue growth.
Railway mania: productive over-investment
So what can history teach us? The Railway Mania in 1840s Britain offers perhaps the closest parallel to today's AI investing frenzy. Railways represented genuinely transformative technology, with The Times declaring them "the wonder of the world". Share prices doubled before collapsing, and more than 200 railway companies went bankrupt.
Yet the enduring legacy was Britain’s railway network, which transformed trade and communication for decades. Gareth Campbell’s analysis shows how “productive over-investment” can wipe out individual wealth while still creating lasting economic value. In other words, investors lost heavily, but the infrastructure funded during the frenzy delivered benefits that endured.
A similar thing happened with the car industry, which is dominated today by a small number of very large global players. Another example is the crash that followed the speculative frenzy surrounding dot-com stocks in the late 1990s.
Historical technology bubbles versus long-term impact
Technology era | Peak bubble period | Market decline | Consumer benefit | Investor outcomes |
Railway mania | 1840s | Prices doubled then collapsed | Extensive rail network | 200+ companies bankrupt |
Dot-com bubble | 1995-2000 | NASDAQ fell 78% | Internet infrastructure | $5 trillion in losses |
Auto industry | 1900-1930 | Hundreds failed | Transportation revolution | Mixed; most failed |
Current AI | 2020s-? | TBD | Productivity gains | TBD |
This comparison reveals the pattern: massive consumer benefits alongside mixed investor outcomes.
Why investors often lose
To put it another way, transformative technologies create enormous value, but that value typically flows to consumers rather than investors. Railways provided faster, cheaper transport. Automobiles enabled personal mobility. The internet delivered access to information and commerce.
These benefits resist capture through pricing because competitive forces drive costs down. Companies maintaining high margins face competition from new entrants offering similar services at lower prices.
Several factors work against investor returns in transformative sectors:
Intense competition: Revolutionary technologies attract massive investment, driving down profit margins over time.
Innovation cycles: Rapid technological change quickly obsoletes existing investments, requiring continuous capital expenditure.
Regulatory response: Successful technologies attract oversight that can limit profitability.
The AI sector already exhibits these characteristics. The EU AI Act, which came into force on 2 August, imposes new compliance obligations. In the US, the SEC's AI Task Force signals increased scrutiny of corporate AI claims.
AI investment flow (Q2 2025)
Metric | Value | Source |
Global VC funding to AI | $47.3B | CB Insights |
AI share of total VC dollars | ~50% | Multiple trackers |
AI-focused ETF performance (YTD) | +34.2% | Roundhill CHAT ETF |
The concentration of capital in AI ventures mirrors historical patterns before bubble corrections.
Is AI investing different?
Unlike dot-com companies, today's AI leaders already generate substantial profits. Barry Schwartz of Baskin Wealth Management, who lived through the dot-com crash, told CBC News: "Unlike the dot-com pre-revenue companies, these companies are profitable. They have global distribution, captive customers."
Google, Apple, Meta and Amazon have billions of customers. Those businesses will continue whether AI becomes a game-changer or not. If it does, these giants are positioned to benefit.
The venture capital landscape has also matured. Buyouts now outpace IPOs as the dominant exit route, providing more stable value realisation than the IPO-heavy dot-com era.
The enduring risks of concentration
Despite these compelling arguments for why AI investing might be different, academic research consistently demonstrates that concentrated investing, whilst potentially rewarding, carries disproportionate risks that most investors underestimate. The behavioural finance literature reveals a systematic bias called "first stock bias" where investors become overly attached to their initial investment themes and fail to diversify adequately. This psychological trap leads to portfolio concentration that undermines long-term wealth creation.
Consider the internet revolution of the late 1990s. The technology clearly represented the future — and indeed it was. Yet investors who concentrated their portfolios in dot-com stocks experienced devastating losses when the bubble burst in 2000. Amazon, which has delivered extraordinary returns over two decades, fell 94% from its 2000 peak before recovering. Even transformative technologies experience severe volatility that can destroy wealth for insufficiently diversified investors.
The current AI investment landscape exhibits similar warning signs. Price-to-earnings ratios for many AI-focused companies reflect extraordinarily optimistic growth assumptions. When expectations become this stretched, any disappointment — whether from slower-than-expected adoption, regulatory challenges, or competitive pressures — can trigger significant corrections.
Recent research on algorithm aversion and appreciation reveals that investors hold contradictory beliefs about AI-driven investment strategies. They simultaneously fear algorithmic decision-making whilst over-allocating to AI investments, creating misaligned risk-return expectations. This behavioural paradox suggests that many current AI investment decisions are driven more by emotion than evidence.
How to invest rationally in AI
Concentrating on a single theme is tempting, but history shows it’s dangerous. Even Amazon, one of the biggest long-term winners from the internet revolution, fell 94% in the dot-com crash before recovering. Investors who failed to diversify paid a heavy price for betting too narrowly on a transformative trend.
Decades of research confirm that diversification is the closest thing to a free lunch in investing. By spreading capital across sectors, geographies, and asset classes, you can capture the upside of technologies like AI while limiting the damage if expectations prove too optimistic. A global approach also balances different strengths — Europe in industrial AI, Asia in consumer AI, and the US in platform companies — while avoiding dependence on any single regulatory regime.
Practical implementation matters too. Thematic AI funds often tilt heavily toward growth and large-caps, and they tend to charge higher fees than broad index funds. Understanding those factor exposures and keeping costs low are critical to long-term outcomes.
The lesson isn’t to avoid AI entirely, but to treat it like any other exciting innovation: own it as part of a broadly diversified, low-cost portfolio rather than a concentrated side bet.
Revolution without guaranteed returns
We cannot definitively determine whether AI investing constitutes a bubble. The nature of financial bubbles — their retrospective clarity combined with real-time uncertainty—means definitive answers only come after the fact.
What we can say with confidence: history favours consumers over investors when it comes to revolutionary technologies. The pattern is remarkably consistent: massive societal benefits, intense competition, eventual commoditisation, and mixed investor outcomes.
The railway investors of the 1840s funded infrastructure that transformed Britain, but many lost fortunes in the process. Today's AI investors may be funding infrastructure that transforms the 21st century — but they should prepare for the possibility that future consumers, not current shareholders, will be the primary beneficiaries.
This doesn't mean avoiding AI investments entirely. Rather, it means approaching them with appropriate humility about our predictive abilities and caution about concentration risk. The most successful long-term investors share a common characteristic: they maintain broad diversification even when compelling investment themes emerge. They understand that individual brilliance in identifying trends matters less than disciplined implementation of proven portfolio principles.
Your colleague at work who’s gone all-in on AI might see spectacular gains — or devastating losses. Either way, it’s the wrong lesson.
The future undoubtedly belongs to artificial intelligence — but also to healthcare, energy, finance, and other sectors. A broad portfolio captures these opportunities while weathering the volatility that comes with transformative change.
That's not exciting dinner-party conversation, but it's the path to long-term financial success.
As Barry Schwartz told CBC News: "It just comes down to one simple question. Do you think we're gonna be using more AI and data in the future or less?" Most investors are betting the answer is more. The question is whether they'll profit from being right.
PREVIOUSLY ON TEBI
HOW TO FUND THE LIFE YOU WANT
If you want to learn more about saving for retirement, we recommend you read the award-winning book How to Fund the Life You Want by TEBI founder Robin Powell and Jonathan Hollow.
The book is published by Bloomsbury and is primarily aimed at a UK audience.
Buy the book here on Amazon, or, if you prefer, here on bookshop.org.