But as contrarians note, surely this argument is incomplete. After all, the companies leading the current AI investment cycle — Nvidia Corp, Amazon.com and Google, among others — did not build their digital empires by throwing money at mirages.
Time and again, their track records have demonstrated prudent capital management and the willingness to scale only what works, and to exit what does not. In Google’s case, this disciplined approach has even been immortalised with comedic precision at www.killedbygoogle.com, an unofficial catalogue of products and services that met an early and quiet demise.
So, with prices surging and anxieties rising, what gives? At heart, the bubble question comes down to timing and purpose.
Timing matters
See also: We said it ... but Singapore did it
In the short term, have markets shot ahead of reality? Probably.
Stretched over any meaningful horizon, however, the current AI wave is almost certain to be transformative on a scale and scope that eclipses past paradigm shifts.
Indeed, the potential payoffs are staggering. While recent news of Softbank Group Corp’s Masayoshi Son shedding tears at the prospect of selling Nvidia shares to raise capital verges on hyperbolic, he raises a valid point by asking: [What] if AI is able to earn 10% of global GDP over the long term?
See also: Outlook for the ringgit: Between structural drag and cyclical relief
Current International Monetary Fund estimates place global GDP in 2025 at US$117 trillion ($151.4 trillion), whereas — to give a sense of the scale of current AI investments — McKinsey projects AI-related data centre spending worldwide at US$6.7 trillion through 2030, cumulatively. Should AI capture even just 5% of global GDP every year, the potential economic payoff would far outstrip current AI spending and yield returns relatively quickly — making the ongoing spending spree more than justified.
In part, the problem is that our perceptions struggle to capture this longer tail of reality. Amara’s Law captures this neatly: We tend to overestimate the impact of new technologies in the short run while underestimating their long-term consequences. More on this shortly.
For now, the mismatch leaves us in an awkward middle ground, caught between the dawn of a fledgling technology and its eventual disruptive reality. Expectations swing wildly, pulled between the promise of big payoffs and the echo of past financial follies.
Not all bubbles are created equal
Speculative manias take many forms, and history offers countless examples — some rational, most not. The distinction matters because the underlying purpose of the speculation fundamentally shapes whether a bubble ultimately drives value or dissipates with nothing to show for it.
Take the Tulip Mania of the Dutch Golden Age, long regarded as the poster child of irrational exuberance. The flashy but functionally useless flower was bid to stratospheric extremes. At its peak, a single bulb reportedly sold for more than six times the average annual salary. Prices soared for no underlying economic reason, and when the mania collapsed, fortunes evaporated overnight.
By contrast, today’s AI spending has a purpose, whereby the aggressive investments are underpinned by genuine, world-changing potential — exemplifying what Nobel laureate Michael Spence calls a “rational bubble”. At present, governments, companies and investors are pouring capital into AI, knowing full well that only a handful — the few with the potential to redefine entire industries — will ultimately succeed. The speculation is speculative in nature, but not senseless.
For more stories about where money flows, click here for Capital Section
Large swathes of capital will inevitably be incinerated along the way. But such is the nature of any technological transition. The key point is that these cumulative investments drive real progress and expand the technological frontier, making the spending rational even if individual bets fail.
Racing to stay relevant
For individual companies, the logic is brutally simple. With everyone else pouring in capital aggressively, no company can afford to sit on the sidelines for fear of being left behind.
Meta Platforms CEO Mark Zuckerberg put it plainly earlier this year: The risk of over-investing is far preferable to the alternative of being late to an era-defining technological transformation. For context, the company’s commitment to the AI buildout is staggering at US$600 billion in AI-related capex through 2028. Even so, it is far from the most profligate spender among Big Tech.
Indeed, for the incumbent tech giants, their underlying fears are real — existential, even. Technology has increasingly tended towards “winner takes all” dynamics since the turn of the century. Since the modern internet era, connectivity and scale have defined digital dominance. Later, the rise of Software-as-a-Service (SaaS) in the 2010s accelerated platformisation and the shift towards ecosystem-based business models, reinforcing network effects that rewarded early and differentiated movers.
Within these ecosystems, incumbents have established substantial moats and pricing powers, enjoying sustained growth that has ballooned their share of the overall market. Big Tech has only grown bigger (see Chart 2), operating now as leviathans of sorts, with invisible networks that span continents.
AI now represents the next frontier in enabling these giants to maintain dominance and scale. The current rivalry therefore carries enormous implications for who will control this sprawling digital landscape. In a winner-takes-all environment, no tech giant can afford to fall behind.
Long tail of AI transformation
So, we understand now that the stakes are high. At the same time, however, it is necessary to acknowledge AI’s frustratingly uneven progress so far. A recent study by the Massachusetts Institute of Technology, for instance, found that 95% of Gen AI pilot programmes failed to deliver any return whatsoever. But as we noted earlier, this gap between expectations and current reality is a textbook case of Amara’s Law.
AI is still in its infancy, and we are only beginning to scratch the surface of its potential. Current performance should not distract us from its long-term promise. To unpack Amara’s Law, there are actually many factors behind our habitual underestimation of new technologies. One is the “horseless carriage syndrome”: our tendency to interpret the new through the lens of the familiar, the way early cars were initially conceived as carriages without horses.
Beyond this innate short-sightedness, technology itself often evolves in unexpected ways once critical mass has been reached. Novel uses proliferate, driving further adoption, and, over time, the effects compound. Joseph Schumpeter’s concept of “creative destruction” captures this pattern: New technologies tend to disrupt old paradigms in uneven, non-linear ways; yet, over the long run, they become a primary engine of productivity growth.
This time, it’s different
Crucially, the transformative potential of AI is even more profound than previous technological waves. Be it that investors have good reason to be sceptical of the phrase “this time, it’s different”, it is undeniable that the current AI revolution carries meaningful differences from past technological transitions, ones that merit attention.
For one, AI belongs to the rare category of innovations known as “general-purpose technologies”, or GPTs (not to be confused with the shorthand for ChatGPT, which stands for “Generative Pre-trained Transformer”). Unlike railways or the printing press, which were undeniably revolutionary, yet highly specific in function, AI lends itself to a multitude of use cases across countless industries. Its pervasiveness magnifies the scale and scope of its impact.
At the same time, however, successful AI deployment requires unmatched levels of coordination across multiple layers of interlocking technologies, ones whose R&D demands are immense yet highly uncertain, and where practical challenges hinder rapid deployment even as the push for adoption mounts. In other words, the other half of the AI puzzle is execution, given that it is not a single product but an interdependent stack.
The challenge of bringing together all the moving parts needed to bring AI to life is far from trivial. Take Microsoft Corp, for instance. CEO Satya Nadella admitted in an interview in early November that the company currently sits on a sizeable stockpile of Nvidia graphics processing units (GPUs) for the simple reason that there are not enough “warm shells”, or powered data centres, to plug them into.
Beyond that, equipment refresh cycles add another layer of complexity. As Moore’s Law observes, semiconductor performance doubles roughly every two years, meaning that cutting-edge AI hardware quickly becomes obsolete. This shortens the payback period and raises the threshold for justified spending. Yet, as mentioned above, the intense competition between peers racing in lockstep forces companies to invest even more heavily to stay in the race.
Patience and perspective
In truth, the term “bubble” has been used so expansively that it hardly explains much. So, returning to the topic of current valuations, the question is not whether AI is in a “bubble”. What matters more is understanding the root of present-day elevated valuations.
Far from being purely irrational or driven by greed, though these factors certainly exist at the margins, much of the apparent excess we see in the markets today reflects the “growing pains” of an economy recalibrating to a new technological paradigm.
Valuations swing wildly because investors are tasked with picking future winners before the terrain has fully taken shape. Companies face a similar imperative: Invest aggressively despite the uncertainty, or risk falling irreversibly behind. The steady stream of breakthroughs dominating headlines every day highlights just how fluid the AI frontier remains.
Our understanding of this paradigm-shifting technology is constantly being challenged and recalibrated, and we simply do not — and cannot — know what tomorrow brings. And compounding this uncertainty is the equally daunting challenge of scaling AI systems, whose complex, interlocking parts make the task anything but straightforward.
So, it is a glorious mess for now, but one that is only to be expected. As the “destruction” in Schumpeter’s “creative destruction” makes clear, technological progress inevitably destabilises the status quo. From here on, some companies will emerge as winners, whereas many others will fail.
Ultimately, what is required is patience. AI, as a generation-defining technological innovation, will reshape the economy and our lives. It just takes time.
Our belief? Stay invested in AI. What to invest in? Stay tuned to this column.
The Malaysian Portfolio came in flat for the week ended Dec 10, performing better than the broader market. The FBM KLCI was down 0.7% over the same period. Hong Leong Industries (+0.8%), LPI Capital (+0.4%) and United Plantations (+0.2%) were the gaining stocks while Kim Loong Resources (-0.4%) and Malayan Banking (-1.4%) ended in the red. Total portfolio returns now stand at 197.7% since inception. This portfolio is outperforming the benchmark FBM KLCI, which is down 12% over the same period, by a long, long way.
The Absolute Returns Portfolio, on the other hand, gained 1.4%, lifting total returns to 42.3% since inception. The top three gainers were Ping An – H (+9%), Ping An – A (+7.5%) and Trip.com (+2.8%). Berkshire Hathaway (-2.6%) and ChinaAMC Hang Seng Biotech ETF (-0.2%) were notable losers last week.
The AI Portfolio also ended higher, up 2.6%. Total returns now stand at 7.3% since inception. The biggest gainers were Horizon Robotics (+16.3%), Robosense Technology (+10.8%) and Naura Technology (+6.0%); and the biggest losers were Marvell Technology (-7.7%), Datadog (-3.0%) and Amazon.com (-0.3%).
Disclaimer: This is a personal portfolio for information purposes only and does not constitute a recommendation or solicitation or expression of views to influence readers to buy/sell stocks, including the particular stocks mentioned herein. It does not take into account an individual investor’s particular financial situation, investment objectives, investment horizon, risk profile and/or risk preference. Our shareholders, directors and employees may have positions in or may be materially interested in any of the stocks. We may also have or have had dealings with or may provide or have provided content services to the companies mentioned in the reports.
