For the past two years, one question has followed the AI trade everywhere: Are investors watching the next technology revolution, or are they watching another bubble inflate?
Nvidia sits at the centre of that debate.
The company has become the clearest financial winner from the AI infrastructure boom, but its valuation is now creating an interesting contradiction. Nvidia’s share price has surged, yet its forward valuation has fallen as earnings have grown even faster.
That is why DBS Group Chief Investment Officer Hou Wey Fook argues that calling the current AI rally a bubble may be premature.
The Nvidia number that changes the conversation
Nvidia is trading at roughly 17 to 19 times forward earnings, depending on the data source and timing.
That number matters because investors tend to associate Nvidia with extreme valuations. But today’s multiple looks very different from the valuations seen during the dot-com era.
For comparison, Cisco was trading at more than 100 times earnings around the peak of the dot-com bubble.
Nvidia, meanwhile, is generating substantial profits and cash flow while still posting extraordinary growth.
DBS’s Hou puts it simply: if the “poster child” of the AI trade is trading at a mid-teens earnings multiple, it becomes harder to argue that the entire AI rally is simply speculation.
The important distinction is this: investors are paying for earnings, not just a story.
Nvidia’s earnings are doing the heavy lifting
One of the most important developments in the Nvidia story is the gap between its share price and its earnings growth.
When earnings grow faster than the stock price, the valuation multiple falls.
That is exactly what has been happening.
The company has continued to deliver strong results while its forward P/E has compressed. One source in the material puts Nvidia’s current forward P/E around 19 times, compared with a longer-term average close to 32 times.
That means Nvidia can become more valuable as a business without becoming dramatically more expensive relative to its earnings.
This is a very different setup from a classic bubble, where prices typically run far ahead of underlying fundamentals.
But the AI trade is still concentrated
There is another side to the story.
The AI rally may not look like the dot-com bubble on a company-by-company valuation basis, but market concentration is becoming difficult to ignore.
A relatively small group of companies, including Nvidia, Microsoft, Alphabet, Amazon, Meta and Broadcom, is driving a significant share of market returns.
At the same time, these companies are committing enormous amounts of capital to AI infrastructure.
The combined capital spending of some of the largest technology companies has already reached hundreds of billions of dollars, with expectations that annual spending could cross the $1 trillion mark in 2027.
That creates the next big question:
How much AI spending can the economy absorb before investors start demanding proof of returns?
The real risk may not be Nvidia
This is where the bubble debate gets more interesting.
The biggest risk may not be Nvidia’s valuation itself.
It may be the economics of the broader AI buildout.
Hyperscalers are spending aggressively on data centres, chips, networking equipment and energy infrastructure. Some of that spending is being supported by debt.
If AI revenue and productivity gains continue to accelerate, the investment can make sense.
But if spending keeps rising while the financial payoff starts slowing, the market could become much less forgiving.
In other words, investors should not only ask:
“Is Nvidia expensive?”
They should also ask:
“Are Nvidia’s customers getting enough return on their AI investments to keep spending at this pace?”
That may ultimately be the more important question.
Three things investors should watch
1. AI spending versus AI returns
The first warning sign would be continued increases in capital expenditure without a corresponding acceleration in revenue or cloud growth.
If companies keep spending more but the financial benefits start flattening, investors could begin questioning the entire AI investment cycle.
The upcoming earnings season will therefore matter.
Watch not just the headline numbers, but what companies say about 2027 capital spending, AI demand and returns on investment.
2. The debt behind the AI buildout
The AI infrastructure race requires enormous amounts of capital.
As borrowing increases and interest rates remain elevated, investors will become increasingly focused on the cost of financing this expansion.
If AI-related debt starts trading at increasingly wider spreads, it could signal that bond investors are becoming less comfortable with the spending cycle.
That would be an important warning sign even if Nvidia’s own earnings remain strong.
3. The IPO market
The next test could come from the private AI market.
An IPO such as Anthropic’s would give investors another way to judge how aggressively the market is valuing AI businesses outside the established technology giants.
A disciplined IPO would suggest investors are still demanding fundamentals.
A frenzy of huge valuations, aggressive pricing and massive first-day jumps would look much more like the late 1990s.
The Nvidia buyback sends another signal
Nvidia recently increased its share repurchase authorization by $150 billion, taking its remaining authorization to approximately $235 billion.
That is significant for another reason.
A company generally has more incentive to repurchase shares when management believes the stock offers attractive value relative to its expected earnings and cash generation.
Nvidia’s decision to deploy capital this way therefore adds another layer to the valuation debate.
It is not proof that the stock is cheap.
But it does show that Nvidia’s management is confident enough in its long-term cash generation to commit substantial capital to buybacks.
Then there is the next phase of AI
Another interesting part of the Nvidia story is that the current AI cycle may not be limited to today’s data-centre spending.
The next opportunity could increasingly involve physical AI.
That includes robotics, industrial automation and machines capable of interacting with the physical world.
The argument from Nvidia bulls is that this could eventually create another major source of demand for computing power.
If that happens, today’s infrastructure spending could look like an early stage of a much larger technology cycle.
But investors should be careful not to price all of that future growth into today’s valuation.
So, is this 1997 or 2000?
The comparison with the dot-com era is useful, but only up to a point.
There are similarities.
The similarities:
- Market leadership is concentrated in a handful of technology companies.
- Capital spending is accelerating rapidly.
- Investors are paying close attention to a transformative technology.
- Interest rates are becoming a potential headwind.
- Expectations around future growth are extremely high.
But there is also a major difference.
Today’s AI leaders are generally profitable companies with real revenue, cash flow and established customer bases.
Many of the companies that captured investor attention in 1999 and 2000 were valued on expectations rather than meaningful profits.
That does not make today’s AI trade risk-free.
It simply means that calling it another dot-com bubble may be too simplistic.
The biggest risk for investors
The biggest mistake would be to conclude that “not a bubble” means “can’t fall.”
Even Nvidia can experience a major correction if expectations move faster than fundamentals.
The market can also punish a stock trading at a reasonable multiple if earnings growth slows sharply.
And because AI has become such a large part of major indices, weakness in the biggest AI names could have consequences well beyond the technology sector.
A good business can still be a bad investment at the wrong price.
That principle remains just as relevant to Nvidia today as it was to technology stocks in previous cycles.
What should investors actually watch?
Rather than trying to predict the exact moment when the AI trade peaks, investors can watch the underlying signals.
Three questions matter most:
- Are AI investments producing enough revenue and productivity gains to justify the spending?
- Can companies continue financing the infrastructure buildout without putting pressure on balance sheets?
- Are AI valuations in the private and public markets remaining disciplined?
If the answers remain positive, the AI rally could have much more room to run.
If spending continues to accelerate while returns weaken and financing becomes more expensive, the argument changes quickly.
The portfolio lesson
DBS’s Hou is not advocating an all-in bet on AI.
His preferred approach is a “barbell” portfolio, combining growth assets such as technology stocks with investment-grade fixed income, while using hedge funds and gold as additional diversifiers.
That is an important takeaway.
You do not need to decide that AI is either the greatest investment opportunity of the decade or the next bubble.
You can believe in the long-term AI story while still accepting that valuations, interest rates and market concentration create risks.
Bottom line
Nvidia’s valuation makes the AI bubble argument harder to sustain, but it does not eliminate the risks.
The company is growing earnings at an extraordinary pace, and its forward valuation is well below the levels associated with the dot-com peak.
But the broader AI trade is now entering a more demanding phase.
The next leg of the rally will need to be supported by real returns on AI spending, not simply bigger spending forecasts.
That is the line investors should watch.
Because the question is no longer simply “Is AI a bubble?”
The better question is:
“Can the profits keep catching up with the expectations?”