The truth about the AI bubble debate
Forget the old rules of tech investing: the AI bubble debate misses what’s actually different about this cycle, according to Igor Pejic.
Henry Jennings recently sat down with Igor Pejic, author of Tech Money and internationally recognised expert on the intersection of technology and finance, to discuss the new rules of investing in the AI age, which industries face disruption next, and why the metrics that defined 20th-century investing no longer apply.
Why this time really is different
Pejic’s central thesis in Tech Money is that technology has fundamentally restructured stock markets and investment decision-making since around the year 2000. Price-to-earnings ratios and book value analysis were built for a world dominated by traditional businesses. In the tech age, the real returns come from catching exponential growth curves, and doing that requires a different set of tools entirely.
That said, Pejic is careful to distinguish between genuine structural change and the irrational exuberance that destroyed the dot-com generation. He cites John Templeton’s warning that “this time it’s different” are the four most dangerous words in investing, but notes that Templeton himself acknowledged, in the same letter, that 20% of the time things genuinely are different. The dot-com crash happened because burn rates outpaced any credible revenue model and the internet was narrowband. Today’s AI companies have real revenues, real demand, and real monetisation. The product works, and the numbers back it up.
Big tech and the AI arms race
One of the more counterintuitive arguments in Tech Money concerns the dominance of large technology companies. Historically, joining the top ten by market cap was a signal of future underperformance – companies got too big to stay nimble. That pattern has broken down with big tech, which has demonstrated an unusual ability to ride successive technological waves. Kodak had the first digital camera patented and still failed to adapt. Google, Microsoft, and Amazon have shown they can leap onto new technologies because they have the capital, talent, and infrastructure to do so.
That capital is now being deployed at a scale that raises legitimate questions. The hyperscalers are collectively spending somewhere in the range of $750–800 billion US dollars this year on AI infrastructure, and the financing mix is shifting. Free cash flow gave way to debt, and now to equity raises and circular arrangements where chip producers take stakes in model providers that then buy their chips back. Pejic sees this as a warning sign worth watching, though not yet at the level of systemic risk that characterised the dot-com era. The underlying demand is real. The open question is whether the investment is proportionate to it.
Which industries are next
AI is what economists call a general purpose technology. It doesn’t disrupt a single industry; it disrupts every industry that can apply it. Pejic sees biotech as one of the clearest near-term beneficiaries, with clinical trials that previously took years now being compressed using AI. Space is another theme gaining momentum, fuelled in part by the data centre buildout and the economics of satellite infrastructure. Energy will become the biggest bottleneck in the AI race, making nuclear technology increasingly relevant. And cybersecurity faces an arms race of its own, as the capability of frontier AI models to exploit security loopholes is already forcing a defensive response across the industry.
On jobs, Pejic sees two scenarios. The more likely near-term path is a significant workforce shift, with fewer people doing the same tasks more efficiently and new roles emerging around working with AI – much as tractors reduced farm labour without ending agriculture. The more disruptive scenario involves artificial general intelligence that reasons fully autonomously. That remains technically distant, requiring breakthroughs well beyond stacking more GPUs into data centres, and it is far from certain it will arrive in any near-term timeframe.
Valuations and the SpaceX question
On valuations, Pejic is direct: some stocks are priced for outcomes that strain credibility. SpaceX (NASDAQ: SPCX) is his clearest example. With a price-to-sales ratio of around 94 – against Spotify’s ratio of 16 at IPO, which was already considered stretched – the implied growth required to justify the valuation is extraordinary. Pejic notes that Elon Musk’s compensation package requires both a permanent Mars colony of one million inhabitants and a corporate valuation of $7.5 trillion. A four-times increase contingent on interplanetary colonisation is not an ideal risk-reward proposition.
The broader valuation concern is not that the AI trade is over, but that the action is moving toward the application layer – and picking winners among thousands of competing applications is a significantly harder problem than owning five or six hyperscalers. For most investors, the infrastructure names remain the more legible bet.
Quantum computing and what comes after AI
Looking a decade out, Pejic nominates quantum computing as the technology with the greatest transformative potential – with the significant caveat that the timeline is genuinely unknown. Today, quantum computers can solve theoretical puzzles but remain too unstable for any commercially relevant task. If that changes, the implications extend across every field that requires massive computational power, from drug discovery to space exploration to AI that reasons like a human. Cybersecurity is the nearer-term theme – less hyped than AI, but facing its own escalating arms race as frontier models become capable of exploiting virtually any security loophole.
The core investing lesson
For investors trying to apply any of this practically, Pejic’s framework from Tech Money comes down to a few principles. Technology is the biggest driver of long-term returns, and the goal is to catch a growth curve early – before it becomes consensus. That is harder than it sounds, because tech investing is psychologically demanding. Valuations swing violently, sentiment shifts fast, and the temptation to abandon a well-reasoned thesis at the worst moment is constant.
Pejic’s advice, particularly for younger investors with long time horizons, is to set a thesis, stress-test the assumptions, and stick to the game plan. The investors who do that across a full cycle – through the drawdowns as well as the runs – are the ones who capture the compounding that makes tech investing worthwhile in the first place. For those interested in Tech Money, the book is available through major retailers.