Amazon and Microsoft report this week with one uncomfortable question hanging over both calls: investors want proof that the AI buildout can turn back into cash.
The number is almost too large to hold in your head. Amazon is expected to spend more than $200 billion on capital expenditures in 2026. Microsoft has told investors it expects roughly $190 billion for the calendar year. Put them together and you get close to $400 billion from two companies in one year. That is the test. The AI story no longer wins just because the spending is large. You now have to show when the spending earns its keep.
As Fortune reported on July 26, earnings from Microsoft are due Wednesday, July 29, and Amazon follows Thursday, July 30. The timing is brutal because Alphabet has already gone first. Google's parent reported Q2 revenue of $119.8 billion, up 24%, and capital expenditures of $44.9 billion, according to recent market reports. It also raised its full-year capex forecast to as much as $205 billion. Free cash flow turned negative for the first time since Alphabet's 2004 IPO, at about minus $5.9 billion. The stock fell more than 7%. That was the warning shot.
Here's the thing: nobody serious is arguing that AWS and Azure are weak businesses. They aren't. Amazon said AWS sales rose 28% year over year to $37.6 billion in Q1, its fastest growth in 15 quarters, and Andy Jassy said the business had reached an annualized run rate of about $150 billion. S&P Global Market Intelligence's Visible Alpha preview put Q2 AWS consensus revenue around $40.5 billion, with AWS margin expected at 33.8%. BofA Securities, according to Investing.com, raised its AWS growth estimate to 33% year over year.
Those are strong numbers. They just sit beside a much uglier cash-flow line. Amazon's own Q1 release showed trailing 12-month free cash flow fell to $1.2 billion from $25.9 billion a year earlier, mainly because purchases of property and equipment rose by $59.3 billion. That is not a small wobble. If you own the stock, you're being asked to accept a near-term cash squeeze because the company says the capacity will be monetized later.
The cash question is now louder than the growth story #
Microsoft has the cleaner cloud narrative, but it faces the same math. In its fiscal Q3 call, Microsoft said Azure and other cloud services revenue grew 40%, or 39% in constant currency. It guided Q4 Azure growth to 39% to 40% in constant currency and said capital expenditures would rise to more than $40 billion for the quarter. Amy Hood also told investors Microsoft expects roughly $190 billion in calendar 2026 capex, including about $25 billion from higher component pricing.
That last detail matters. Chips, power and data center leases are no longer background costs - neither is memory. They are the story. Microsoft also said demand continued to exceed available supply, which is a good problem until investors start asking how long the company has to keep writing checks before that shortage shows up as free cash flow. Growth is real. So is the bill.
Fortune's Bloomberg piece captured the turn in market mood through Jason Lemire, chief investment officer at Bold Wealth Partners, who said investors used to see more capex as better and now see less as better. That's a sentiment reversal. It won't rewrite the long-term AI thesis by itself, but sentiment reversals move stocks long before spreadsheets catch up.
Amazon does have a concrete answer it can point to. In February, OpenAI and Amazon announced an expanded AWS agreement worth $100 billion over eight years, on top of an earlier $38 billion multi-year agreement. OpenAI also committed to consume about 2 gigawatts of Trainium capacity through AWS infrastructure. That is real demand, not a slide-deck hope.
But you still have to care about timing. Amazon's Q1 materials said much of its 2026 AWS capex will be installed in future years and that cash is often laid out 6 to 24 months before billing starts, depending on the component. That is the part investors should watch closely on July 30. A backlog is useful only if it turns into revenue on a schedule that keeps depreciation and financing costs from eating the upside.
The risk cuts both ways #
There is a perfectly good argument that Amazon and Microsoft have no real choice. If you under-invest in AI infrastructure while Alphabet, Meta, Oracle, OpenAI and Anthropic keep reaching for more compute, you don't just lose a little cloud share. You risk missing the next computing platform. Nvidia has been the obvious winner from that fear, with hyperscaler demand soaking up GPUs and pulling cooling and power suppliers along with it.
Still, the risk is asymmetric in a way management teams rarely say plainly. If enterprise AI demand arrives on schedule, these companies look disciplined and early. If it arrives more slowly - or if cheaper model architectures reduce the need for brute-force compute - they will be sitting on very expensive assets while depreciation grinds through the income statement. Not disaster. Just a slow, dull margin problem.
That is why this week's earnings calls matter. Amazon and Microsoft don't need to convince you that AI is important. That case has been made. They need to show that the revenue curve is catching up with the spending curve before investors decide the biggest infrastructure bet in tech history has become too expensive to applaud on faith.
Also read: Google Gemini has 900 million users and barely any of them chose it • LG Electronics wins Korea's first Nvidia liquid cooling certification as AI data centers hit a heat wall • OpenAI connected 300 million weekly health queries to real medical records and the implications are enormous