The distinction between a transformative technology and a profitable investment is a key point to keep in mind. The challenge with the AI sector is that investors need to separate the technology’s potential from the companies that are most likely to generate sustainable investment returns.

Why AI stocks have run so hard

The AI rally isn't occurring in isolation. Moody's estimates that global investment in the infrastructure powering AI, from data centres and advanced chips to energy supply, will surpass US$3 trillion over the next five years.

That capital must be spent somewhere, and it usually flows first to the companies selling the picks and shovels - chipmakers, cloud providers, data-centre operators and the utilities that power them.

Webull believes earnings for the largest AI suppliers have, so far, largely justified the enthusiasm. When a handful of companies post genuine, cash-backed growth on the back of a structural shift, money usually follows. And in the case of AI infrastructure, it appears it has.

There's a behavioural layer on top of the fundamentals too. Nobody wants to be the investor that sat out on what may potentially be the defining theme of their era, it is the fear-of-missing-out dynamic that could push AI valuations beyond what the numbers alone would support. Both forces are real. The trick is telling one them apart.
 

Why AI valuations are volatile

The same concentration that drove the gains cuts the other way. A large share of recent index returns has come from a very small number of very large companies. When sentiment on those names wobbles, the whole market feels it.

Add AI valuations priced for years of flawless execution, and you have a setup where a single-earnings miss or a rumour about slowing demand has potential to erase months of share-price gains in a single trading session.

Then there's the question nobody can answer yet - when does all this spending actually pay off? Building a US$50 billion data-centre pipeline is a cost today against revenue that's largely a promise in a prospectus. Every time the market suspects that payoff is further away than hoped, it may reprice AI valuations.

Volatility here isn't a malfunction. It's the sound of a market arguing with itself in real time about the future value of entities fueling the AI ecosystem.
 

The AI stack is not one trade

One of the most common mistakes I see is treating 'AI' as a single button you press. It isn't. It's a stack, and the risk and return profile changes dramatically depending on where you sit in it.

• Infrastructure and hardware: chips, servers, cooling and networking. This layer is seeing the most concrete revenue today, but it's also the most capital-hungry and the most exposed if the buildout slows.
• Energy and 'picks and shovels': data centres are voracious consumers of power, which has pulled utilities, grid operators and even property into the story. These are often less glamorous, sometimes more defensive plays on the same theme.
• Platforms and applications: the software companies building AI into products. Here the promise is enormous, but so is the uncertainty about who ultimately captures the profit.
• Enablers and adjacencies: cybersecurity, data management and the specialists that grow regardless of which model wins. As AI expands the attack surface of every business, protecting these systems becomes non-negotiable.

The point isn't to tell you which layer to favour; that depends entirely on your goals, timeframe and risk appetite. The point is that “I'm invested in AI” can mean at least four completely different things and knowing which one you're inclined to commit to is half the battle.
 

Potential upside

If the optimists are even partly right, AI represents a genuine productivity shift on the scale of electrification or the internet - the kind that lifts economic output across entire industries rather than just one.

For investors, that can mean exposure to a multi-decade structural trend rather than a passing fad. The infrastructure being built now, much like that fibre-optic cable in the 1990s, will likely underpin economic activity for decades, whoever ends up owning it.

Unlike some past manias, a large part of this boom rests on companies with real revenue, real customers and real cash flows today, not just a compelling presentation slide deck.
 

Risks - read this part twice

Here is where honesty matters most, so let me be direct about what could go wrong with AI-related investments:

• Valuation risk. Even a genuinely transformative technology can be a poor investment if you overpay for it. Prices that assume perfection leave no margin for the ordinary disappointments that every industry eventually delivers.
• Concentration risk. Because so much market value in AI now sits in so few AI companies, an investor who thinks they're diversified across the index may in fact be heavily exposed to a single theme. That's worth checking.
• The 'right theme, wrong company' trap. For every enduring winner of a technology boom, there is a graveyard of companies that had the correct idea and still went to zero. Picking the theme is easy. Picking the survivors is brutally hard.
• Circular financing and hype. When the same handful of companies are simultaneously each other's suppliers, customers and investors, revenue can look more robust than it is. Healthy skepticism about where the money is really coming from is warranted.
• Political risk. The question on most people’s minds, but with no clear-cut answers, is ‘What is the social and economic cost of AI and how deep will it go?’ This is something the world’s governments are grappling  with as they seek answers from the AI juggernauts on the social and economic consequences of their inventions, with very little substance provided in return thus far. So, there is a risk that political overreach through kneejerk actions could see a bottleneck in the very technology touted to improve mankind, akin to man discovering fire, which could  stifle growth and deployment of the technology and act as a drag on company metrics.
• Timing and drawdowns. As Stein's Law reminds us, “If something can't go on forever, it will stop.” That doesn't tell you when - trends can run far longer and further than sceptics expect. But it does mean anyone in this theme should be prepared for sharp, stomach-testing falls along the way, and should never invest money they can't afford to see halved.
   

Conclusion

So, AI boom or bust? Almost certainly both at some stage, and probably more than once. That's how transformative technologies have always arrived - real progress, highlighted by episodes of wild over-enthusiasm and painful corrections, with the durable value only obvious in hindsight.

None of this is a reason to avoid the theme, and none of it is a reason to pile in blindly. It's a reason to do the unglamorous work - understand which layer of the stack you're buying, know what you're paying for the growth you're promised, size your position so a bad year doesn't derail your plan, and hold a timeframe long enough for the technology to actually matter.

The investors who do well out of AI over the next decade won't be the ones who were most excited. They'll be the ones who stayed clear-eyed while everyone else was choosing between euphoria and panic.
 

From ASX

ASX Content On-Demand has information on latest market and sector trends.