Satya Nadella, CEO of Microsoft. The company signed more than US$130 billion of new data centre leases during the June quarter alone. Photo / Sven Hoppe, dpa

For much of the past two years, the artificial intelligence story has been fairly simple. Companies mentioned AI, investors got excited, and share prices went up. That phase may now be ending.

But that does not mean the AI boom is over. In fact, the evidence increasingly points the otherway. Demand for AI infrastructure remains extraordinary. Cloud growth is accelerating. Semiconductor profits are surging. Data centre investment continues to rise. The world’s largest technology companies are still committing hundreds of billions of dollars to the theme.

There are obvious reasons why some investors are becoming more circumspect. Sentiment towards AI has been extremely bullish, with positioning in many leading technology stocks increasingly crowded.

That does not automatically mean a bubble is about to burst, but it does mean the margin for disappointment is becoming much smaller. When expectations are high, even strong results can be punished if they are not strong enough.

That helps explain why investors are becoming more discerning. They are no longer asking which companies are spending the most on artificial intelligence. They are asking which companies can demonstrate that spending is already turning into revenue, earnings and cashflow.

Microsoft was rewarded last week, with its shares soaring 15%, because it gave the market exactly what it wanted: evidence that AI demand is already flowing through the business. Azure growth accelerated to 43%, annual Azure revenue exceeded US$100 billion (NZ$170b) for the first time, and Microsoft 365 Copilot reached more than 30 million paid seats.

Microsoft also disclosed that it signed more than US$130b of new data centre leases during the June quarter alone, taking its total future lease commitments to US$329b. These are largely long-term agreements for AI data centres that have yet to come online. That figure alone provides a powerful reminder of just how enormous the AI infrastructure build-out has become.

The market reaction was extraordinary, with Microsoft adding nearly half a trillion US dollars in market value in a single day.

Amazon delivered a similar message. Its cloud business continues to boom. Amazon Web Services grew 37% to US$42.2b, its fastest growth since 2021, while management lifted capital spending plans to around US$220b.

Normally, that sort of spending increase might have made investors nervous. Instead, Amazon shares rallied 15% because investors could clearly see where the money was going. Cloud growth was accelerating, demand continued to run ahead of capacity, and management said even that level of investment would not be enough to meet customer demand.

That is the new rule of this earnings season. Investors are not worried about companies spending hundreds of billions of dollars. They are worried about companies spending hundreds of billions without proving it is working.

Alphabet, the owner of Google, showed just how fine that line has become. On the surface, its result was impressive. Revenue rose 24% to US$119.8b, Google Cloud revenue jumped 82% to US$24.8b, the cloud backlog reached US$514b, and Gemini now has 950 million monthly active users.

Why? Because investors were focused on the cost of that growth. Alphabet lifted its 2026 capital expenditure guidance to as much as US$205b. Second-quarter capital expenditure alone reached US$44.9b, double the level of a year earlier, with most of that spending directed towards AI infrastructure.

The initial reaction illustrated that the market is no longer giving even the strongest technology companies an automatic pass. Revenue growth still matters. But capital discipline now matters too.

Meta highlighted the same issue from a different angle.

Its advertising business remains exceptionally strong, and AI is clearly improving recommendations, targeting and engagement. But investors are increasingly focused on the cost of the company’s broader AI ambitions.

The pressure point remains Reality Labs, the division behind virtual reality headsets and AI-powered smart glasses. That business lost US$4.6b in the latest quarter, taking cumulative operating losses since late 2020 to more than US$80b. Free cashflow also fell sharply as infrastructure spending accelerated.

The initial market reaction again suggested investors wanted more visible payback.

Then there is Apple, which sits in a slightly different category.

Apple is not trying to win the AI infrastructure race in the same way as Microsoft, Amazon, Alphabet or Meta. It is not spending hundreds of billions of dollars building data centres as aggressively as the hyper-scalers. Instead, Apple is leaning more heavily on its ecosystem, devices, software integration and partnerships.

Its result last week was strong in many respects. Revenue rose 16% to more than US$109 billion, iPhone sales surged 22%, and Mac sales jumped 29%.

At the current run rate, Apple is selling roughly eight iPhones every second – a huge reminder of the extraordinary scale of the business.

But the share price reaction was more complicated. Apple warned that sales growth this quarter would be between 9% and 11%, below market expectations. The issue was not weak demand. It was supply. Tim Cook pointed to memory shortages and limited chip supply constraining production of iPhones, Macs and iPads, describing the disruption as a “100-year flood” for the industry.

This makes Apple an interesting case study in the broader AI debate.

Unlike the hyper-scalers, the market is not worried that Apple is spending too much on AI. The concern is almost the opposite: can Apple move quickly enough to stay competitive as rivals pour enormous sums into artificial intelligence?

For Apple, the issue is not whether it can afford the AI race. It clearly can. The question is whether its more measured, ecosystem-led approach will be enough to drive the next major upgrade cycle across the iPhone, Mac and Services businesses.

So across Big Tech, investors are now asking very different questions.

That range of reactions tells us something important.

The market is not turning against AI.

It is becoming more selective.

Meanwhile, none of this proves AI is in a bubble. It is just that investors are no longer blindly rewarding companies simply because they mention artificial intelligence. Increasingly, they are demanding evidence that AI investment is translating into stronger earnings, cashflow and sustainable competitive advantages.

There will inevitably be those who argue this is simply the dotcom boom all over again. The similarities are obvious. The differences, however, are equally striking.

Semiconductor demand provides another useful comparison. Sales have risen sharply during the AI boom, increasing by around 175% in the current cycle. That is remarkable, but still well short of the roughly seven-fold increase seen during the internet build-out of the late 1990s. Today’s cycle is undoubtedly large. It simply has not yet reached the same level of speculative excess.

Unlike the late 1990s, investors are not simply paying higher and higher multiples for the same profits. In many cases, earnings expectations have risen sharply while valuation multiples have remained flat or even contracted. Across much of the AI complex, share price appreciation has been driven by stronger earnings rather than investors simply paying more for the same dollar of profit.

Investors should hold two ideas in their minds at the same time. AI remains one of the most important long-term investment themes in the world. But not every AI-related investment will be a winner. Identifying the next big investment theme is only part of the challenge. As the market matures, the real opportunity lies in identifying the companies best placed to convert AI investment into sustainable earnings growth. It is entirely possible to believe AI will transform industries while also accepting that the market is right to become more selective.

The market is no longer giving every AI story the benefit of the doubt. The easy phase of the AI trade was about imagination. The next phase could well be about execution. That is not a bad thing. A market that rewards earnings rather than excitement is ultimately a healthier one.

The question is no longer who is spending the most on AI. The question is who can prove it is working.

Generate is a New Zealand-owned KiwiSaver and Managed Fund provider managing over $10 billion on behalf of more than 195,000 New Zealanders.

This article is intended for general information only and should not be considered financial advice. The views expressed are those of the author. All investments carry risk, and past performance is not indicative of future results.

To see Generate’s Financial Advice Provider Disclosure Statement or Product Disclosure Statement, go to www.generatewealth.co.nz/advertising-disclosures/. The issuer is Generate Investment Management Limited.