Meta and Microsoft signal capex growth rates are decelerating; AI infrastructure investment velocity moderating after sustained acceleration
On the evening of July 29, after market close, Meta and Microsoft each released their latest quarterly earnings. Meta reported Q2 revenue of $60.8 billion, up 28% year-over-year and slightly exceeding expectations; Microsoft reported FY2026 Q4 revenue of $90 billion, with Azure growth of 43% and full-year Azure revenue surpassing $100 billion for the first time. Both reports showed strong results, yet the market's reaction diverged sharply. Meta's stock fell more than 10% after hours, while Microsoft rose nearly 9%.
But what the two companies said about capital expenditures during their earnings calls matters more than single-day stock swings. Investment continues and has actually grown significantly year-over-year. However, the rate of growth in capital expenditure is beginning to show signs of deceleration.
It's important to clarify one point: "slowing growth" and "declining spending" are not the same. Meta's full-year capex guidance is $130 billion to $145 billion; Microsoft's estimated capex for next quarter exceeds $500 billion. The absolute scale continues to grow, and there won't be a cliff in the near term. What this piece examines is the growth rate—for example, potentially declining from 80% to 50% and potentially falling further. This shift is barely visible at present and is not yet settled.
Meta narrowed its full-year capital expenditure guidance from $125 billion to $145 billion to $130 billion to $145 billion. It raised the lower bound while keeping the upper bound unchanged.
Looking at the numbers alone, the $130-145 billion range remains massive, with investment intensity far exceeding last year. But the ceiling didn't move up. This quarter's revenue beat expectations; if AI compute demand is accelerating, the ceiling should rise in lockstep. Instead, it stayed at its original level.
On the earnings call, Zuckerberg spent considerable time defending AI investment. His exact words: "If we simply sold off all our compute capacity in exchange for short-term profits, that would equally be shortsighted." He sketched out multiple monetization pathways—consumer-level agents, enterprise APIs, compute rental—but each remains distant from scale. Wall Street wanted a timetable for returns; he offered a vision instead.
CFO Susan Li provided a telling detail when responding on infrastructure planning, stating the company is committed to "maximizing capacity in 2026 and 2027" while preserving flexibility for server purchasing decisions for 2028 and beyond. 2028 is two years away. A company known for aggressive investment beginning to discuss purchasing flexibility two years out is a choice of language that has been carefully considered.
The true hard constraint is cash flow. Free cash flow plummeted from $8.55 billion in the same quarter last year to $784 million, a staggering 91% decline and the lowest since Q3 2022. Profitability also came under pressure, with net income down 14% to $15.8 billion year-over-year, which included $2.4 billion in legal settlement provisions and $1.18 billion in severance costs—one-time impacts. Excluding these two items, results could have exceeded expectations. Yet even so, operating margin fell from 43% to 31% year-over-year, while capex for the quarter surged 83% to $31.08 billion. The primary culprit behind the cash flow squeeze remains massive AI infrastructure investment; the one-time items merely compound the problem.
Next quarter's revenue guidance midpoint is $62.5 billion, below the average analyst estimate of $63.2 billion. When free cash flow drops from $8.5 billion to below $800 million, sustaining 80-100% capex growth on near-zero cash flow long-term becomes mathematically infeasible. Growth rate moderation here is a mathematical constraint.
Microsoft reported $90 billion in revenue, up 18% year-over-year and exceeding expectations. Azure growth accelerated from 40% last quarter to 43%, with guidance for the coming quarter accelerating further to 45%. Copilot paid seats jumped from 20 million to 30 million in a single quarter. The newly launched Microsoft 365 E7 suite—bundling Copilot, Agent 365, and security products, launched just in May—is starting to contribute ARPU growth. Microsoft also recorded a $3.2 billion unrealized gain on its strategic investment in Anthropic; Anthropic's valuation climbed from roughly $183 billion six months ago to $965 billion, bolstering quarterly profit. Stock rose nearly 9%.
On capital expenditure, Microsoft adjusted its 2026 calendar-year guidance from approximately $190 billion to approximately $175 billion. On the surface, this is a reduction. CFO Amy Hood provided a clear explanation on the call: the change stems from accounting policy adjustments—office buildings and data center useful lives extending from 15 to 25 years, and certain leases reclassifying from finance to operating leases, no longer capitalized. Excluding accounting effects, actual investment expectations remain unchanged. Next quarter's capex is still expected to exceed $500 billion, continuing to grow from this quarter's $41 billion. FY2027 full-year capex is still expected to grow year-over-year.
Based purely on the figures above, "growth deceleration" appears to find no direct evidence in Microsoft's story. Yet several details taken together point in a different direction.
First, consider the timing. Extending useful lives for office buildings and data centers from 15 to 25 years is a reasonable accounting move. But executing it now—precisely when market skepticism over AI capex returns is mounting—objectively signals attentiveness to capex discipline to the market. If this were purely a neutral accounting adjustment, there would be no need to choose this particular moment.
Next, consider the language. Management only said FY2027 capex "will grow," without specific figures or language like "significant growth." If planning to continue accelerating sharply, the phrasing would differ.
Hood also volunteered an exit ramp. When asked by an analyst about return on investment, she noted that Microsoft's capex has increasingly shifted toward CPUs and GPUs—short-lived assets with rapid procurement cycles. "If the demand environment changes, we can directly reduce spending on this largest cost category." This phrasing didn't appear in last quarter's call. At the peak of growth, no company volunteers talk of cuts. Preparing an exit ramp in advance signals management's anticipation that growth rates may flatten.
Nadella repeatedly emphasized operational efficiency on the call, calling it "normalized, deep-rooted work for the long term." Engineering improvements in CPU and GPU utilization allowed newly freed compute to translate into revenue within the quarter. Substantial portions of Azure's 43% growth stem from efficiency gains. Demand still exceeds supply, but Microsoft is choosing to fully leverage current capacity efficiency before deciding the pace of further expansion.
Zooming out a week, Google's parent company Alphabet delivered its own report on July 22. Google Cloud grew 82%, accelerating from 63% last quarter in March, but capex more than doubled year-over-year to $44.9 billion, with full-year guidance raised from $180-190 billion to $195-205 billion. Free cash flow turned negative for the first time—minus $5.9 billion—marking the company's first negative FCF in decades.
The three giants' earnings each had their own emphasis, yet their overall posture converged: all emphasize operational efficiency and return on investment. For Meta, Zuckerberg said "selling compute for short-term profit is shortsighted," while the CFO preserved flexibility for purchasing decisions two years hence. Microsoft's Nadella called efficiency optimization "normalized work," while Hood volunteered an exit strategy for reducing their largest cost item. Google's capex guidance still rose, but free cash flow turned negative.
This consensus itself is a signal. When the world's largest buyers simultaneously discuss efficiency and flexibility on earnings calls rather than only growth blueprints, it usually means the pure expansion phase's peak growth is likely behind us.
The difference between Microsoft and Meta lies in Microsoft's ability to demonstrate how investments convert into revenue—Azure growth still accelerating, Copilot paid seats doubling, E7 suite just ramping—prompting a 9% market rise. Meta lacks such a path for now, falling 10%. But both face shared cash flow pressure; Microsoft's free cash flow also declined 23% to $19.6 billion this quarter. Both show similar flattening growth rate slopes. This holds equally for both.
For the semiconductor hardware supply chain, highly dependent on large cloud provider purchasing, even the faintest signal of deceleration warrants attention.
Global chip stocks experienced violent selling this week. By one accounting, the top 20 global chip companies combined saw market value evaporate by over $1.5 trillion across three trading sessions. NVIDIA, SK Hynix, Samsung Electronics, and TSMC were spared nothing. This selloff began before Meta and Microsoft earnings, with direct triggers including skepticism over NVIDIA's "circular financing" model—NVIDIA recently disclosed over $500 billion in supply chain partnerships with the SK Group and planned compute rental guarantees up to $250 billion for OpenAI, reigniting market doubts over the "circular cash flow" logic of AI capex. Google's negative free cash flow further heightened anxiety. Meta and Microsoft's earnings confirmed this selling pressure from the fundamental level.
"Early signs of growth deceleration" does not equal "spending peaked." Meta's full-year capex guidance still reaches $130-145 billion; Microsoft's actual investment expectations remain unchanged. The semiconductor supply chain won't face demand cliffs in the near term. But if growth rate slopes begin to flatten, an entire sector built on the "AI infrastructure capex" high-growth narrative over the past two years needs to recalibrate expectations.
Amazon's earnings will be released after market close on July 30 local time. Amazon's capex for this year is estimated at around $200 billion, the largest among the three giants. Should Amazon's capex also release similar signals, "early signs" can be upgraded to trend confirmation. Until then, uncertainty facing the semiconductor supply chain has already shifted from "is demand sufficient" to "how much longer can buyer growth rates hold."