Alphabet and Microsoft Earnings Test Whether AI Spending Will Pay Off

The bar isn't 'beat estimates.' It's 'show me the payoff.'

Alphabet and Microsoft report earnings after the bell Wednesday. Together they're 11.5% of the S&P 500, so if you own a plain index fund, you're more exposed to Wednesday night than you might think.

Here's the odd part: a good quarter may not be enough. Analysts expect revenue growth of 25.2% at Alphabet and 16.7% at Microsoft, which is huge for companies this size. Meta just showed how little a beat buys you right now. It topped revenue expectations and gave stronger-than-expected fourth-quarter guidance. Its stock still slid 7% Thursday and another 1% Friday. Demand for AI looks as strong as ever. Investors are now asking a different question: will the giant spending needed to meet that demand turn into profit and cash?

Investors would like a roadmap to profitability on the other side of the AI infrastructure spend. Even a crudely drawn set of directions on a napkin might do.

What's on the table

The good news first. Google Cloud revenue jumped 82% year over year to a record $24.8 billion in the second quarter. Microsoft's Azure cloud business grew 43% to $37.8 billion last quarter, and Yorkville Ives analysts expect it to hit the high end of management's 44%-to-45% growth guidance. Microsoft 365 Copilot, its AI assistant, passed 30 million paid users, up 50% from the prior quarter. Chip suppliers Micron, ASML and TSMC have all reported strong results recently.

Now the worrying part. Alphabet used to avoid debt. It has now started borrowing, and it reported negative free cash flow of $5.9 billion in the second quarter. That means it spent more cash than its business brought in. Gartner estimates AI providers will spend about $6 trillion on infrastructure from 2027 to 2030. It warns that AI revenue "will be insufficient to support profitable returns on all planned AI capital expenditures." Capital expenditures, or capex, means spending on long-lived assets like data centers and chips.

The so what: demand clearly isn't the problem. Supplier results and cloud growth both prove customers are showing up. What's at risk is the math. If revenue doesn't grow fast enough to cover the spending, the money keeps going out faster than it comes in, and these stocks get valued less like cash machines and more like big construction projects.

The numbers that will move the stocks

Cloud backlog: This is the bulls' best card. Backlog is revenue that customers have already committed to but that hasn't been booked yet. Yorkville Ives says Alphabet's backlog exceeds its capacity. It estimates hyperscalers (the giant cloud operators) have a $2 trillion backlog across the group that's growing faster than sales. If that's still true Wednesday, the companies can say their spending fills orders that already exist. They aren't building and hoping customers show up. If backlog flattens, that defense gets much weaker.

Capex plans: A bigger spending number isn't automatically bad. Yorkville Ives argues the hyperscalers have plenty of room to spend more because they're short on capacity, not customers. The risk is what happened to Meta: higher spending, negative free cash flow and a falling stock, even after a strong quarter. Watch whether rising spending comes with a clear line to the revenue it's supposed to produce.

Cloud margins: These are the profit the cloud businesses keep on each dollar of sales, and analysts will watch them closely. Healthy margins say AI customers are paying enough to justify the buildout. Shrinking margins would mean the bill is arriving before the payoff.

Hard AI numbers: Copilot's 30 million paid users is the kind of specific, countable figure the market wants. The more AI revenue the companies can point to directly, and the less they fold it into broad categories, the easier it is to believe the spending will pay off.

The bigger picture: AI money is getting more expensive

The weakest link in the AI chain is the neoclouds. These are companies whose entire business is renting out computing power. Most aren't profitable, and they've borrowed billions to grow fast. Interest on CoreWeave's bonds climbed past 15% this month. Its credit default swaps, which work like insurance against a company failing to pay its debts, implied a better-than-even chance of default within five years. Gartner predicts 80% of neocloud providers could stop operating independently by 2030. Alphabet and Microsoft aren't in that position, thanks to their huge, diversified businesses. Still, the cost of insuring Meta's debt hit a record last week. Even the giants are getting asked whether they're borrowing too much.

Private markets are adjusting too. Anthropic may lower its IPO valuation target from $2 trillion to $1.5 trillion, amid worries about slowing revenue growth and an AI bubble. It hopes to raise $60 billion to $100 billion, mostly for computing power. OpenAI reportedly first pitched a $1.4 trillion valuation and is now close to a round at $1.2 trillion. At least five firms have filed for AI bond ETFs. One analyst, Seaport's Jay Goldberg, expects yields on AI debt to start rising because so much of it is coming to market. When yields rise, borrowing to build data centers costs more.

What to watch: If Alphabet and Microsoft show backlog still growing, steady cloud margins and concrete AI revenue, they can argue they're the safe end of the AI trade. If they raise spending without much to show for it, expect the pressure that hit Meta and CoreWeave to spread to the biggest stocks in the market. Those are also the stocks that make up the biggest slice of your index fund. And with Anthropic's IPO likely next month, a rough Wednesday could make public investors even pickier about what they'll pay for AI.