Amazon and Microsoft are each spending roughly $200 billion in 2026 building out AI data centers — a combined $400 billion bet that's testing investor patience like never before. With earnings due this week for both companies, Wall Street is asking a question that would have seemed absurd a year ago: when does all this spending start paying off?
Alphabet kicked off earnings season with a warning shot. The Google parent raised its 2026 capital expenditure forecast to $195-205 billion, up from $180-190 billion, to meet what CFO Anat Ashkenazi described as "supply-constrained" demand for AI cloud capacity. Investors responded by sending Alphabet's stock down 7% on Thursday — the worst single-day drop in nearly three years.
The sell-off spread across the hyperscaler group. Amazon, Meta, and Microsoft all fell too, as the market reassessed the sheer scale of infrastructure investment the industry has committed. The combined data center lease commitments among the largest cloud providers now exceed $850 billion, with Meta and Microsoft leading the charge.
"The trade-off is worthwhile given AWS's re-acceleration and Amazon's expanding platform advantages," Wedbush analysts wrote in a note Thursday, recommending buying Amazon stock. But the tone has shifted from the unqualified enthusiasm of 2025.
The $400 Billion Question
Amazon in February guided to $200 billion in capex for the year, the highest among the hyperscaler group — until Alphabet pushed its top end to $205 billion. Andy Jassy, Amazon's CEO, told investors in April that the company's "plan is largely the same." Analysts now expect Amazon to lift its forecast again when it reports earnings, with Visible Alpha consensus creeping up to $207.4 billion after Alphabet's report.
Microsoft guided $190 billion in April, including $25 billion attributed to higher component prices as AI chip demand consumes memory supply. Analysts expect the company to hold at $190.1 billion when it reports, but Cowen analyst Derrick Wood told CNBC that a further increase "based on what we saw in the reaction of Google, it's probably going to lead to selling pressure in the stock."
The spending is not optional. Alphabet's Ashkenazi was blunt on the earnings call: "While we have increased our capacity quite a lot over the past three years, the demand still outpaces that investment. We are, just like the rest of the industry, working in a supply-constrained environment."
Cloud Revenue Growth Remains Red-Hot
Despite the capex anxiety, the cloud business itself is booming. Google Cloud revenue jumped 82% year over year to $24.8 billion in the second quarter — the fastest growth since at least 2020. It was the fourth consecutive quarter of acceleration, and Google's cloud business, once dismissed as a distant third, is now nearly half the size of AWS.
Amazon Web Services, which still leads the market, grew 28% in the first quarter, with analysts expecting nearly 32% in Q2. Azure and other Microsoft cloud services expanded 40% in Q1, with Q2 consensus at 39%. None of the three shows any sign of slowing.
Gartner, the research firm, raised its 2026 global IT spending forecast to $6.37 trillion on Monday, citing surging AI infrastructure demand as the primary driver. The 14.2% year-over-year increase was an upward revision from Gartner's April forecast of $6.31 trillion. "AI infrastructure growth remains rapid despite concerns about an AI bubble, with spending rising across AI-related hardware and software," said John-David Lovelock, a distinguished VP analyst at Gartner.
The full story on Gartner's latest numbers is covered at CFO Dive.
Data Center Leases Hit $850 Billion — And Climbing
Behind the headline spending numbers is a less visible infrastructure boom: long-term data center lease commitments. Future lease obligations among the largest cloud and tech companies have passed $850 billion, according to Bloomberg, with Meta and Microsoft accounting for the biggest additions.
Meta alone added $79 billion in new lease commitments — a 76% increase — even though the company does not operate a conventional cloud business. Instead, Meta is reportedly exploring a plan to sell excess computing capacity to third parties, effectively becoming a cloud provider by accident as its AI infrastructure buildout outpaces internal demand.
The scale of these commitments is reshaping the data center construction industry. Technology companies are pre-leasing capacity years in advance, locking in power and cooling contracts and driving up prices for everyone else. Enterprises that do not plan their data center capacity 18-24 months ahead are finding increasingly limited options.
Free Cash Flow Turns Negative for the First Time
The most worrying signal for investors may be the cash flow picture. Amazon's free cash flow flipped negative in the first quarter of 2026 for the first time since 2022, when the company was doubling its warehouse footprint in response to a pandemic e-commerce surge. Analysts expect it to stay negative for the full year.
Alphabet turned cash flow negative in Q2 2026 for the first time, after raising $98 billion in long-term debt — a 111% increase over six months. Microsoft's free cash flow is projected to go negative in Q4, which would be the first time since at least 2001, according to FactSet estimates.
These numbers represent a structural shift for companies that have been cash-generation machines for years. The bet is that the AI infrastructure paid for today will generate returns for decades. But the debt is real, and the interest payments are already hitting income statements.
Tiffany Wade, a fund manager at Columbia Threadneedle which held positions in all three hyperscalers at the end of June, told CNBC that "patience is required for these names because I do think that these will be AI winners over the medium and longer term."
Growing AI Fatigue Among Investors
The risk of an AI spending backlash is not hypothetical. Janus Henderson Investors found in a May study that nine out of 10 investors have at least some concerns about AI, with 28% citing fears that it may not live up to expectations. Two-thirds of investors surveyed said they were worried about a potential AI bubble or market correction.
Corporate boards are also tightening the screws. A Cloudzero study released in June found that 66% of boards are conditioning additional AI funding on proof of return. Nearly 90% of finance leaders said they feel pressured to tie AI spending to business outcomes within the next year, but only 22% have already achieved that goal.
Jake Dollarhide, CEO of Longbow Asset Management, summed it up: Amazon could struggle to impress investors "in this environment of growing AI fatigue, the sudden questioning of meteoric capex budget increases and Silicon Valley and the Mag 7 taking on noticeable levels of debt in order to fund the massive data center buildout."
What It Means for Cloud Customers
For enterprises buying cloud services, the capex boom is a double-edged sword. On one hand, hyperscalers are building capacity at a pace that will eventually drive down compute costs — the fundamental economics of scale that have made cloud computing cheaper than on-premises infrastructure for most workloads.
On the other hand, the current supply-constrained environment means that the biggest AI workloads — training large models, running real-time inference pipelines — command premium pricing. AWS, Azure, and Google Cloud have all introduced premium instance tiers for GPU and custom AI accelerator servers that carry steep markups over conventional compute.
The shortage of high-bandwidth memory and AI accelerators is a key bottleneck. Microsoft attributed $25 billion of its capex increase to higher component prices, reflecting the global scramble for HBM memory modules and advanced packaging capacity.
The Edge Computing Thread
The $400 billion hyperscaler buildout is primarily about centralized mega-data centers — the kind that house tens of thousands of GPUs and consume 100+ megawatts of power. But the infrastructure wave is also pulling edge computing along.
Equinix, the world's largest data center REIT, published analysis this week arguing that the best AI infrastructure strategy is one you can change — meaning a mix of centralized cloud, distributed edge nodes, and on-premises deployments. With 34% of Kubernetes clusters now estimated to run on edge devices, from Raspberry Pis to NVIDIA Jetson boards, the edge-to-cloud continuum is becoming a practical reality for enterprises that need low-latency AI inference.
The edge computing segment of the market is expanding fast, and the same companies building hyperscale data centers are also deploying smaller nodes closer to users. For a deeper look at how these trends connect, check our prior coverage of how Google's $30 billion SpaceX deal is reshaping the cloud compute market.
The Bottom Line
The $400 billion AI infrastructure buildout is the biggest capital deployment in the history of the technology industry. It will reshape the cloud computing market for a decade or more. But the market is no longer giving these companies the benefit of the doubt — investors want results, not just promises.
The semiconductor supply chain, AI model developers, and enterprise cloud customers all have a stake in whether the hyperscaler bet pays off. The next 12 months will tell us whether $400 billion was the investment of a generation or the biggest overbuild since the dot-com era.
As Gartner's Lovelock put it: "Driven by the expansion of AI workloads and demand for high-performance computing, hyperscalers and enterprises are rapidly scaling next-generation data center capacity." The only question is how long the market's patience will last.