While in the investment world, AI is presented as an omnipotent solution that will transcend the limitations of the human mind to create a utopian future, in reality, the tech has several chokepoints, which make its promises lean more towards wishful thinking rather than tangible outcomes. Dr. Laszlo Gömöri, CEO of market-timing and behavioral analytics firm Spectralwaves, takes this view. Gömöri, who has been advising institutional traders for decades, is also the former CMO of Pictet Asset Management in Geneva, CIO Emerging Markets of Bank J. Safra Sarasin, and chief economist at Novartis.

The Electricity Chokepoint

The physical limitations of AI are undeniable, says Gömöri. First, and most importantly: the electricity. «The investment announcements made in this sector over the past months are remarkable», says Gömöri. 

AI firms such as Google have made their energy requirements for the coming years public. «They are currently buying electricity for the next five years». Add up these requirements, he says, and the result is striking: the U.S. would need to increase its electricity generation capacity by 150 GW over the next five years to power new data centers. «Such an increase is simply unfeasible».

To put this into perspective, a typical nuclear reactor has an installed electrical capacity of around 1 GW, according to the U.S. Department of Energy. In other words, meeting the additional demand would require the equivalent of at least 150 reactors, he says.

McKinsey’s projections lend weight to Gömöri’s assessment. «Even in a low data center demand scenario, the US power sector could add more than 150 GW of power capacity to supply the IT load demand by 2030», the consultancy states. It also estimates that data centers are being built at a pace that will require almost 30 GW of power annually.

In practice, more installed capacity would be needed to deliver the same amount of electricity over a full year, as maintenance, refuelling and unplanned outages reduce reactors’ average output.

However, there is the catch: according to the International Atomic Energy Agency, it takes around 10 years for a nuclear power plant project to be completed successfully. Meeting that demand within five years therefore looks all but impossible, says Gömöri.

«As for wind and solar energy, their contribution remains limited due to the intermittent nature of their output, which is incompatible with the continuous demand of data centers».

The Energy Illusion

While tech giants have wanted to address this issue by turning to small modular nuclear reactors (SMRs), which only produce 5 to 300 MW, Gömöri points to the fact that none of these firms have obtained a single permit or licence to construct such initiatives. 

«The total number of nuclear power plants scheduled to come online worldwide over the next five years amounts to about 50 units at best—and that’s assuming no unforeseen events», he says.

Those who would argue in favour of more traditional energy sources will likely also fall short, says Gömöri.

«As for wind and solar energy, their contribution remains limited due to the intermittent nature of their output, which is incompatible with the continuous demand of data centers».

The construction times for this type of infrastructure are well-known, he says. «What is perhaps less well-known is that the lead time for delivering a gas turbine is currently about three years. Power grids are already under heavy strain, and high-capacity transformers have even longer lead times». 

The Dr.Copper Malady

The second blind spot and bottleneck that Gömöri believes investors are missing is copper. «Without copper, there’s no electricity», explains Gömöri.

Not only did the price of copper hit stratospheric highs in January 2026 at USD 14 500 per ton (intraday), but there are also physical limits to copper, says Gömöri. Many of the major copper mining regions sit in politically volatile regions, he says.

«The commodity is found in extremely politically volatile countries riddled with civil wars and guerrilla groups and the like, who often control power. Many major copper deposits are found in Africa, especially in Congo. It’s a bit like Panama, where politically unstable governments can suddenly nationalize the mining companies, so to speak—which has actually happened».

«Secondly, you can’t just find a copper mine overnight. If you find a copper deposit somewhere today, it takes three to five years to actually extract the copper, because you need political approval due to environmental restrictions and so forth», he explains. 

While one could propose gold as an alternative conductor, Gömöri highlights that this material is far too expensive for it to be a realistic option.

Dotcom Bubble Deja Vu

Another Achilles heel that Gömöri foresees is the spending vs. income discrepancy, which he says is eerily starting to resemble a bubble. «OpenAI, for example, is spending several millions every day», he says. «This amounts to about two billion a year. At present, no one knows how they will recoup those billions?»

«These companies remind me a lot of the year 2000», says Gömöri, referencing the Dotcom Bubble. 

«There’s an interesting example: Cisco Systems. Back in 2000, Cisco was one of the high-flyers and one of the world’s largest tech companies. Today, Cisco’s stock price, twenty-five years after the dot-com bubble of 2000—is at the same level as it was 25 years ago. If you had bought Cisco stock twenty-five years ago, you’d be back at break-even now. However, and here’s the interesting part: Cisco Systems earns eight times more money today than it did back then, however the stock price is exactly the same».

But what does that mean? 

«It means the stock was overbought. It was a bubble», he concludes.  «The stock was way too expensive back then, because the investors in the stock market extrapolated it». 

However, unlike CISCO, we currently don’t know how these AI firms will make this money back, he says.  «But we do know that these stocks are generally valued way, way too high and that these valuations are unrealistic».

It’s worth bearing in mind that the subscription models don’t even scratch the surface of the spending of these firms, he says. Recent reports on tech giants appear to support Gömöri’s thesis. In July it was reported that Google recorded a negative free cash flow of $5.9 billion in the second quarter of 2026, driven by the soaring capital expenditures of its artificial intelligence infrastructure.

The Psychology Behind the AI Boom

The AI-investment euphoria or hypnosis can be explained through behavioral finance, explains Gömöri. He says that the collective interest and momentum in AI investments is best understood through the work of economist John Maynard Keynes. More precisely, through Keynes’ theories of Animal Spirits and the Keynesian Beauty Contest.

How AI became the belle of the ball at the Keynesian Beauty Contest

The Keynesian Beauty Contest is a concept developed by Keynes in which investors guess what other investors will find attractive (popular stocks) rather than what they personally find valuable, in order to profit from market trends rather than intrinsic value. This can lead to speculation and bubbles. The economist’s work also shows that when we predict what others find valuable, we often get it wrong. 

In this regard, AI has become the belle of the ball at the Keynesian Beauty Contest. Investors are investing in AI due to the popularity or momentum of the AI stocks, while the current value of AI currently remains somewhat of a mysterious, Gömöri says. Few can truly explain what the actual value of AI is and whether this ‹value› truly has longevity. 

Keynes Animal Spirits on Steroids

Secondly, Gömöri attributes investors’ bullishness on AI to Keynes’ theory of Animal Spirits.

Keynes' Animal Spirits theory refers to the instinctive, emotional, and psychological factors that drive human economic decisions, beyond purely rational calculation. It includes drivers such as herd mentality, instincts, self-confidence, or fear, rather than data and spreadsheets. It posits that with these Animal Spirits, positive mood and high confidence drive risky investments.

Keynes also emphasized that in an uncertain future, people often act on intuition and belief, rather than on empirical evidence. And not only is AI completely uncertain, «AI is the ultimate dream», Gömöri explains. 

This is the crux of the matter

The illusion that AI has the ultimate potential has driven an excessive optimism in investors that exceeds rationality. «It’s completely new and no one knows where it’s headed. That's why we had these extremely subjective expectations. It's purely an extrapolation into the future. They are completely unrealistic valuations, often because we’re in an industry that we simply don’t understand».

Precisely this herd mentality and enthusiasm can blind investors to the fact that they may be running off the side of a cliff. 

«I believe a significant portion of investors betting on AI-related companies will be disappointed.»

Waking up from the AI Hypnosis

While AI may be the ‹ultimate dream›, it may stay exactly that, if solutions aren’t found to the material impossibilities of the tech.  Is the ‹everyone’s darling› of the financial markets really the emperor’s new clothes?

Should it turn out that the AI emperor is in fact naked, what does this mean for investors and institutions who are heavily invested, such as the US pension funds? 

«I believe a significant portion of investors betting on AI-related companies will be disappointed, as it is materially unfeasible for all of their projects to succeed», says Gömöri. It perhaps may be prudent for investors to proceed with caution when it comes to the AI hypnosis. They could possibly soon face a rude awakening should the ‹bubble› go bang.