Alphabet just reported the strongest quarter in Google Cloud’s history. Revenue came in at $119.8 billion, up 24% year over year. Cloud grew 82% to $24.8 billion and blew past analyst estimates. The Cloud backlog hit $514 billion. Nearly 90% of the Fortune 100 is using Gemini Enterprise. By most definitions, that is a blowout quarter.
The stock fell 6.5% the next morning. Capital expenditures came in at $44.9 billion for a single quarter. Free cash flow turned negative. Most of the net income surge came from a one-time gain on the Anthropic stake. Strip that out and investors were left looking at a company spending at a rate that makes even strong revenue growth feel like it may not be enough.
Microsoft (MSFT) reports July 29 and Meta reports July 30. The next week is effectively a live test of whether the AI trade’s math actually works.
What Alphabet’s Q2 results reveal about the AI trade’s biggest risk
The Alphabet (GOOGL) reaction captures the problem in one quarter. Cloud revenue grew faster than at any point in the company’s history. Investors sold the stock anyway, CNBC reported.
The issue isn’t whether AI is generating revenue. It’s whether the capital required to generate that revenue is sustainable, and whether the returns will ever justify the scale of investment.
More Wall Street:
- Wall Street sends strong 4-word verdict on the stock market
- Wall Street’s $200 billion IPO wave threatens sell-off
- Wall Street flees software plays for triple-digit chipmaker boom
Forty-four billion dollars in quarterly capex is not a small number. Annualized, that’s close to $180 billion from Alphabet alone.
When you add Microsoft, Meta (META), and Amazon (AMZN), the combined spending for 2026 is running toward $725 billion, with analysts projecting it could cross $1 trillion in 2027, CNBC reported.
The question the market is now pricing into every print is how long before the revenue catches up, as TheStreet reported ahead of Alphabet’s earnings.
The gap between AI spending and AI revenue that investors are watching
The capex-to-revenue gap is the central tension in the AI trade right now. Sequoia analyst David Cahn has calculated that there is roughly a $600 billion annual gap between what hyperscalers are spending on AI infrastructure and what the AI ecosystem generates in actual sales, Forbes reported.
Goldman Sachs has noted that to justify the scale of investment, hyperscalers would collectively need to generate more than $1 trillion in annual profits, more than double current consensus estimates, as TheStreet reported.
According to Allianz Research, the divergence between AI capital spending and revenue growth is running at 46%, already wider than the 32% divergence seen during the 2001 telecom cycle that preceded years of pain in tech stocks.
Michael Heinrich, co-founder and CEO of 0G Labs, which builds decentralized AI infrastructure, described the dynamic plainly in an interview with TheStreet:
“When the capital going into a technology outruns the revenue coming out of it by that margin, valuations are pricing perfection.”
Alphabet’s results were exceptional. And still, free cash flow went negative. That’s what “pricing perfection” looks like in practice: a quarter that would have been a strong earnings beat in any other sector, and a stock that still dropped because the bar for AI spending to produce proportional returns keeps moving higher.
Michael/Getty Images
How the AI rally compares to the dot-com era and where it diverges
The comparison to the late 1990s is now coming from serious voices.
JPMorgan CEO Jamie Dimon said earlier this month that AI spending may not “pay off the way you expect and the timetable you expect.”
He drew a direct parallel to the internet boom, where the technology proved transformative but the timeline disappointed nearly everyone who priced it in early.
Heinrich sees both the parallel and where it breaks.
“The similarity is the reflexive bidding up of anything with the label attached, well ahead of proven business models. The difference is that the underlying technology this time is already generating real usage and real cash flows, so this is less a fiction problem and more a physics and economics problem,” he added.
The dot-com era was full of companies with no path to revenue. AI has actual enterprise customers paying for actual products. Google Cloud at 82% growth is not a fiction.
The risk isn’t that the technology doesn’t work. It’s that the cost of delivering it at scale may not produce returns proportional to the capital being committed, at the speed the market has priced in.
What Microsoft and Meta need to show for the AI test to pass
Microsoft’s July 29 report will be the next data point. Azure guided for 39% to 40% growth in constant currency. If it delivers at or above that, the market will read it as confirmation that cloud AI spending is translating into revenue acceleration. If it misses, questions about the return on $190 billion in annual capex get louder fast, as TheStreet reported.
Meta reports July 30 against its own complicated backdrop. The company has already cut 8,000 jobs this year and moved thousands of employees into AI roles, then acknowledged at an internal meeting that AI-agent progress has not accelerated as expected.
The question on Meta’s call is whether $125 billion to $145 billion in AI spending this year is producing the kind of product traction that justifies it.
Three things will tell investors whether the AI trade is facing a healthy correction or something more serious:
- Whether the gap between AI infrastructure spending and AI revenue is narrowing;
- If AI is moving from assistant to agent, meaning systems that complete tasks and get paid for outcomes rather than just answering questions, and
- Whether the unit economics of running AI inference are falling fast enough to make applications viable at scale.
The next two earnings reports will give investors more data on all three than any single quarter has provided yet.
Related: Scott Bessent sends unprecedented warnings to China on AI models
