Alphabet Raises 2026 Capex Guidance to $205B: Why Google Stock Dropped 5%

$205 billion is the new top end of Alphabet's 2026 capex guidance, up from $190 billion just one quarter ago, announced alongside Q2 2026 earnings on July 22, 2026. That's the short answer. The longer answer is that Google Cloud grew 82% and beat every estimate on the table — and the stock still fell 5% anyway.

Alphabet just delivered one of the strongest quarters in company history and got punished for it. Revenue hit $119.8 billion, up 24% year-over-year and well ahead of the $117 billion Wall Street expected. Google Cloud posted 82% growth. The backlog swelled to $514 billion. None of that mattered as much as one number: a fresh $15 billion increase to the top of the 2026 capex range, with management warning 2027 spending "will increase significantly" on top of that. Here's what actually happened and what it means for the AI capex arms race.

Why did Alphabet raise its 2026 capex guidance to $205 billion

Alphabet lifted its full-year 2026 capital expenditure guidance to a $195-205 billion range during its Q2 2026 earnings call on July 22, up from the $180-190 billion range set just one quarter earlier — a roughly $15 billion increase at the midpoint. CFO Anat Ashkenazi told analysts the company is accelerating capacity deliveries because demand for Google Cloud AI infrastructure continues to exceed available compute, and cautioned that supply constraints will persist into 2027, when spending is expected to rise again.

Analysts polled by Visible Alpha had modeled around $188 billion for the year — meaning even the low end of Alphabet's new range beats what Wall Street was pricing in. That's the pattern that's defined 2026 for the entire hyperscaler cohort: guide high, then guide higher, because the cost of under-building AI capacity (losing cloud customers to a rival with available GPUs and TPUs) is judged worse than the cost of over-building.

Why Google stock dropped 5% on a beat-and-raise quarter

GOOGL fell roughly 5% in after-hours trading on July 22 despite beating on revenue, cloud growth, and backlog — because investors weighed the capex raise more heavily than the operational beat. The math is straightforward: a $205 billion capex year, layered onto a warning that 2027 will be higher still, compresses free cash flow at exactly the moment the market wants proof that AI infrastructure spending converts into durable margin, not just top-line growth.

Alphabet spent $44.9 billion in Q2 alone on technical infrastructure — about 60% on servers (including custom Ironwood, Google's seventh-generation TPU) and 40% on data centers and networking. That single-quarter number, annualized, tracks almost exactly to the new full-year guidance, which is the clearest signal that management isn't pacing spend conservatively into the back half of the year; it's front-loading capacity as fast as supply chains allow.

This is the same tension we flagged in our Big Tech AI capex breakdown: every hyperscaler now fears being structurally behind on compute more than it fears a capex-driven stock sell-off, and Alphabet's Q2 print is the cleanest evidence yet that the fear of under-investing has fully won internally, even as public markets push back.

Google Cloud's 82% growth and the $514B backlog

Google Cloud revenue jumped 82% year-over-year to $24.8 billion in Q2 2026, up from $13.6 billion a year earlier, blowing past the $22.4-22.46 billion analysts expected and marking the fastest growth rate among the major cloud providers this quarter. Cloud backlog — contracted future revenue not yet recognized — hit $514 billion, which is the strongest forward-demand signal Alphabet has ever reported and the core justification management used for the capex raise.

On the consumer side, the Gemini app now has 950 million monthly active users, up sharply from prior quarters, and is processing 22 billion API tokens per minute, up from 16 billion just one quarter earlier. Nearly 90% of the Fortune 100 is using Gemini Enterprise. Those usage curves are the demand-side story management is betting $205 billion will keep pace with — see our AI valuations dashboard for how usage growth like this feeds into how AI-native and AI-infrastructure companies get priced.

Alphabet's 2026 AI capex guidance vs. Microsoft, Amazon, and Meta

Alphabet's raised guidance puts it near the top of the hyperscaler pack, just behind Amazon and roughly in line with Microsoft, while comfortably ahead of Meta. The table below lines up each company's 2026 capex plans as of their most recent guidance.

Figures are 2026 guidance as reported across Q2 2026 earnings calls and coverage by CNBC, Yahoo Finance, GuruFocus, and Converge Digest. Prior guidance ranges reflect each company's most recent earlier-2026 disclosure; ranges vary slightly by source and reporting quarter.

What Ironwood (TPU v7) buys Alphabet that Nvidia GPUs don't

A meaningful share of the new capex is going toward Ironwood, Alphabet's custom seventh-generation TPU, which became generally available to Google Cloud customers in May 2026. Ironwood scales to 9,216-chip superpods delivering 42.5 FP4 exaflops and 1.77 petabytes of shared HBM memory, and it's now the chip powering Gemini 3.5, Search AI Overviews, YouTube's recommendation engine, and Gmail's smart features internally. Vertical integration on custom silicon is Alphabet's structural cost advantage over cloud rivals still fully dependent on Nvidia allocation.

That's also why Alphabet's capex dollar buys something qualitatively different from a straight GPU purchase: every dollar spent on Ironwood capacity serves both external Cloud customers and Alphabet's own AI products, which is a form of internal demand no other hyperscaler except possibly Amazon (with Trainium) can fully replicate. For more on how this factors into equity pricing, see our Big Tech earnings dashboard.

What this means for investors and startups building on Google Cloud

For public investors, the $205 billion guidance raise is a bet that 82% cloud growth and a $514 billion backlog justify near-term free cash flow pain — and the market's 5% sell-off says that bet isn't fully underwritten yet. For startups and funds evaluating AI infrastructure exposure, the signal is more straightforward: compute supply constraints that have limited GPU and TPU availability through 2026 are unlikely to ease before 2027, which keeps pricing power with the hyperscalers and specialized chip suppliers rather than with buyers.

I've been tracking this cycle closely because it changes how we underwrite AI-native startups: a company whose margins depend on falling compute costs is making a bet against what all four hyperscalers are telling you directly on earnings calls. Alphabet just told the market, in writing, that 2027 capex "will increase significantly" on top of $205 billion. Plan accordingly.

How fast Alphabet's capex has grown since 2023

Alphabet's capex trajectory shows how quickly the AI buildout escalated. Full-year 2023 capex ran around $32 billion. That roughly doubled to about $52.5 billion in 2024 as the first wave of generative AI demand hit. 2025 capex then jumped again to roughly $91-93 billion, and now 2026 guidance sits at $195-205 billion — meaning Alphabet's annual capital spending has grown more than 6x in three years. Few capital-intensive businesses in history have scaled spending this fast this consistently, which is itself a data point for how unusual this cycle is.

That growth curve also explains why depreciation is becoming a bigger swing factor in Alphabet's reported earnings. Servers and TPUs depreciate over a shorter useful life than office buildings or fiber networks, so a $205 billion capex year flows through the income statement as materially higher depreciation expense within 12-24 months — pressuring reported operating margin even if the underlying cash generation from Cloud and Search stays healthy. That's part of why analysts reacted to the guidance raise instead of just the trailing quarter's beat.

Is Alphabet's $205B AI capex bet paying off yet

The clearest evidence the spending is working sits in the Cloud segment itself: Google Cloud's operating margin has been expanding even as revenue growth accelerates, a combination that's genuinely rare for a business scaling capacity this fast. An 82% growth rate paired with a $514 billion backlog means demand isn't the constraint — supply is, which is precisely the argument Ashkenazi made for spending more, not less, despite the market's skepticism.

The harder question is return on invested capital over a 5-7 year horizon, not the next two quarters. A $205 billion capex year assumes AI compute demand keeps compounding at something close to the current rate through the depreciation life of the hardware being bought today. If Gemini usage growth or enterprise AI adoption decelerates meaningfully before that hardware pays for itself, Alphabet — and every hyperscaler making the same bet — will have built capacity for a demand curve that didn't show up. That's the actual risk being priced into the 5% sell-off, more than any single quarter's numbers.