Is AI a bubble, or is the best yet to come?
Key Takeaways
- Bubble dynamics: A defining feature of market bubbles is the feedback loop between management ambition and investor capital. Today’s AI boom shows similar traits—surging spending, sky-high valuations, and a heavy reliance on investor funding rather than customers.
- Stacking the odds: Forecasting winners and losers is notoriously difficult; even extreme valuations can precede extraordinary returns. Rather than trying to call the top, investors can stack the odds by owning well-run, cash-generative businesses with robust economics across multiple outcomes.
- Selective exposure: The promise of AI is real, but so is the risk of overpaying. Selective positions in firms like Taiwan Semiconductor Manufacturing Company (TSMC), Nebius, and Samsung enable us to capture the structural benefits of AI adoption without the speculative multiples that often accompany bubbles.
Author: Ben Preston
Benjamin Preston, Master of Arts (Honours) in Mathematical Sciences (University of Oxford), Chartered Financial Analyst. Ben joined Orbis in 2000. He currently directs client capital in the Orbis Global Equity Strategy and has oversight of our Responsible Investing initiatives. During his time at Orbis Ben has served as the leader of the Global Sector Team and as a director of both Orbis Holdings Ltd and Orbis Investment Advisory Ltd.
Key Takeaways
- Bubble dynamics: A defining feature of market bubbles is the feedback loop between management ambition and investor capital. Today’s AI boom shows similar traits—surging spending, sky-high valuations, and a heavy reliance on investor funding rather than customers.
- Stacking the odds: Forecasting winners and losers is notoriously difficult; even extreme valuations can precede extraordinary returns. Rather than trying to call the top, investors can stack the odds by owning well-run, cash-generative businesses with robust economics across multiple outcomes.
- Selective exposure: The promise of AI is real, but so is the risk of overpaying. Selective positions in firms like Taiwan Semiconductor Manufacturing Company (TSMC), Nebius, and Samsung enable us to capture the structural benefits of AI adoption without the speculative multiples that often accompany bubbles.
A hallmark of any self-respecting stock market bubble is not just the dramatic rise in share prices but also the surge in corporate spending. It’s that coalition between managements eager to grow and investors cheering them on with ever-larger infusions of capital that lays the groundwork for future disappointment. When managements go too far—giddy on easy money and optimistic forecasts—they sow the seeds for a bust created by their own overexpansion. That risk looms large today, given the enthusiasm around AI. OpenAI, for instance, is being valued at $500 billion in private markets, making it one of the world’s most highly valued companies despite generating revenue comparable to a mid-sized US regional bank. It has made eye-wateringly large spending commitments totalling over a trillion dollars while expecting to remain cash-flow negative for at least four years. So far, vastly more of its money has come from investors than from customers. Add in the circular money flows—the “infinite money glitch”—between the likes of OpenAI, Nvidia, and Oracle, and alarm bells should be ringing. Yet the technology behind AI looks genuinely transformative. Proponents describe the transition from “generalised computing” to “accelerated computing” as being as significant as the commercialisation of electricity or the shift from horses to cars. Adoption remains at an early stage relative to its full potential.
“The headache for investors is that the same ingredients that often characterise bubbles, can also describe an emerging investment trend that's too important to miss.”
“Successful investing has little to do with predicting the future. It's about discipline-owning quality businesses at attractive valuations across a range of possible outcomes.”
The headache for investors is that the same ingredients that often characterise bubbles—high share prices, rapid expansion before profitability, and extravagant forecasts for unproven demand—can also describe an emerging investment trend that’s too important to miss. How, in advance, can one tell the difference between a dotcom bubble, featuring disasters like Global Crossing and pets.com, and a digital revolution that produced Apple, Netflix, and Amazon? Hindsight always makes it look easy. After the dust settles, there will always be those taking victory laps for having called it correctly. But if it were truly easy, everyone would see it, and bubbles would never form. Apart from at “extreme extremes”, making those big market calls is surprisingly hard. There is no single metric that provides certainty. In 2002, after the dotcom bust, the CEO of Sun Microsystems famously chastised investors for having paid “ridiculous” multiples of 10× revenues just a few years earlier. Any investors who heeded that warning would have struggled to buy Nvidia at the same valuation 20 years later—and would have missed returns of 2,700%. The future is inherently unknowable. History is a useful guide, but it’s far from perfect because markets are dynamic and learn just as quickly as investors, if not more so. While not perfectly efficient, they’re very good at pricing shares at a level that keeps decisions difficult. If that uncertainty feels uncomfortable, take heart: nobody knows what the future of AI holds—not even its creators. The playing field is more level than it appears: we’re all operating in an environment of unknowns. More encouraging still, successful investing isn’t about predicting the future. While a crystal ball would no doubt be useful, investors can do exceptionally well by focusing bottom-up on companies with sound economics, outstanding management, and valuations attractive across a wide range of possible outcomes. Investors who assemble a portfolio of such businesses stack the odds in their favour. Fortunately, we’ve been able to buy shares in companies that have turned out to be clear beneficiaries of AI—such as TSMC, Nebius and SK Square—without having to pay the bubble-level multiples that come with significant downside risk.
The headache for investors is that the same ingredients that often characterise bubbles—high share prices, rapid expansion before profitability, and extravagant forecasts for unproven demand—can also describe an emerging investment trend that’s too important to miss. How, in advance, can one tell the difference between a dotcom bubble, featuring disasters like Global Crossing and pets.com, and a digital revolution that produced Apple, Netflix, and Amazon? Hindsight always makes it look easy. After the dust settles, there will always be those taking victory laps for having called it correctly. But if it were truly easy, everyone would see it, and bubbles would never form. Apart from at “extreme extremes”, making those big market calls is surprisingly hard. There is no single metric that provides certainty. In 2002, after the dotcom bust, the CEO of Sun Microsystems famously chastised investors for having paid “ridiculous” multiples of 10× revenues just a few years earlier. Any investors who heeded that warning would have struggled to buy Nvidia at the same valuation 20 years later—and would have missed returns of 2,700%. The future is inherently unknowable. History is a useful guide, but it’s far from perfect because markets are dynamic and learn just as quickly as investors, if not more so. While not perfectly efficient, they’re very good at pricing shares at a level that keeps decisions difficult. If that uncertainty feels uncomfortable, take heart: nobody knows what the future of AI holds—not even its creators. The playing field is more level than it appears: we’re all operating in an environment of unknowns. More encouraging still, successful investing isn’t about predicting the future. While a crystal ball would no doubt be useful, investors can do exceptionally well by focusing bottom-up on companies with sound economics, outstanding management, and valuations attractive across a wide range of possible outcomes. Investors who assemble a portfolio of such businesses stack the odds in their favour. Fortunately, we’ve been able to buy shares in companies that have turned out to be clear beneficiaries of AI—such as TSMC, Nebius and SK Square—without having to pay the bubble-level multiples that come with significant downside risk.
30 Sep 2025 | Source: CapIQ, Orbis. *Market capitalisation as at 30 September 2025, or in the case of OpenAI, based on its latest private equity raise. ^Consensus estimated revenue for 2025, calculated as the look through revenue of SK Square's 20.1% holding in SK Hynix, or in the case of OpenAI its latest reported annual recurring revenue.
Looking ahead, we expect the same disciplined approach will allow investors to continue to find opportunities to gain exposure to AI’s potential, without taking a directional bet on whether it will prove the most transformative innovation or the biggest bubble in history. Artificial Intelligence is real, and it will change the world—perhaps in ways we can’t yet imagine. Whether today’s early leaders will dominate or collapse dotcom-style, paving the way for new winners to emerge from the ashes, remains to be seen. But by staying close to the ground, remaining adaptable, and assessing each company on its merits, we believe the hunting ground will remain fertile for bottom-up investors seeking to buy companies for less than they’re truly worth.
Orbis Investment Management Limited © 2025 The trademarks ‘Orbis’, ‘Orbis Invest Differently’ and ‘Orbis Invest Differently & Design’ are owned by Orbis Holdings Limited and used with permission by Orbis Investment Management Limited. This is a marketing communication. Past performance does not predict future results. The value of investments in the Orbis Funds may fall as well as rise and you may get back less than you originally invested. It is therefore important that you understand the risks involved before investing. This report represents Orbis' view at a point in time and provides reasoning or rationale on why we bought or sold a particular security for the Orbis Funds. We may take the opposite view/position from that stated in this report. This is because our view may change as facts or circumstances change. This report constitutes general advice only and not personal financial product, tax, legal, or investment advice, and does not take into account the specific investment objectives, financial situation or individual needs of any particular person. This report does not prohibit the Orbis Funds from dealing in the securities before or after the report is published. This document contains information relating to one or more Funds managed by Orbis Investment Management Limited (together ‘Orbis Funds’). This paper is directed only at professional advisers and institutional/professional investors and is not suitable for a direct retail audience. Please do not distribute or copy any part of this document without our consent. You should consider the relevant offering documents including the Fund Prospectus and Key Information document (for a SICAV Fund) before making any final investment decisions. These offering documents, as well as a summary of investor rights, are available in English in the Fund Documents section of our website. Additional notes for Australian clients: Equity Trustees Ltd AFSL No. 240975 (EQT) is the issuer of units in the Orbis Funds domiciled in Australia. You should consider such funds’ Product Disclosure Statement (PDS) or Information Memorandum (IM), as applicable, before acquiring or disposing units in such funds’. The PDS or IM can be obtained from www.orbis.com.au. Target Market Determinations (TMDs) for the Orbis Funds can be found on our 'Forms' page under 'How to Invest'. Each TMD sets out who an inestment in the relevant Fund might be appropriate for and the circumstances that trigger a review of the TMD.