Cloud Demand Hardens, Capex Ratchets Higher — and the 2026 Spend Is Already a Tell

Results vs. Consensus

Quality of the Beat

• Revenue: Genuine acceleration — total +14% (+13% cc) vs. Q1's +12% (+14% cc), with Search +12% (vs. +10%), Cloud +32% (vs. +28%), YouTube ads +13% (vs. +10%), Subs/Platforms/Devices +20% (vs. +19%). Every reportable line in Services accelerated despite tougher comps.
• Operating margin: 32.4% headline understates underlying performance. The $1.4B legal settlement charge is a discrete one-time hit (~150bps); excluding it, consolidated margin was ~33.9% — flat Q/Q despite a 35% Y/Y increase in depreciation. Underlying operating leverage is intact.
• EPS: $2.31 vs. $2.17 is a clean operational beat. OI&E this quarter was $2.66B (vs. $11.2B in Q1, which had the $8B private-co unrealized gain) — closer to a normalized run-rate, so the quality of EPS is higher than Q1's headline implied.
• FCF: $5.3B in Q2 looks ugly in isolation but is mostly mechanical: capex stepped up to $22.4B from $17.2B, and Q2 includes federal tax payments for both Q1 and Q2. TTM FCF of $66.7B is the right number, and it's down from $74.9B last quarter — the capex cycle is starting to compress trailing FCF.

Segment Performance

Google Services — Search Accelerates Despite (Because of?) AI Overviews

The Q1 thesis — that AI Overviews monetizes at parity with classic Search — got a stronger second data point in Q2. AI Overviews' user base scaled from 1.5B to 2B+ MAU across 200+ countries and 40 languages, AI Mode rolled out to U.S. and India and reached 100M+ MAU in one quarter, and Search & other revenue actually accelerated to +12% from +10%. Pichai disclosed a particularly important new metric: queries showing AI Overviews are growing 10%+ faster than overall query volume.

"AI Overviews are now driving over 10% more queries globally for the types of queries that show them, and this growth continues to increase over time." — Sundar Pichai, CEO

Schindler delivered the most pointed line of the quarter on monetization mechanics, addressing the bear thesis head-on:

"Some of the product and policy changes we make actually drive better monetization at the expense of paid clicks. You will see in the 10-Q paid clicks were up 4% year on year, but a number of factors affect these metrics... It's really important when it comes to paid clicks and CPCs to avoid drawing overly broad conclusions solely based on these metrics." — Philipp Schindler, CBO

Translation: paid clicks +4% with revenue +12% means CPC is rising — a classic positive monetization mix. Schindler is pre-empting the bear narrative that AI is reducing clicks; the data shows clicks are growing modestly while revenue per click is rising sharply, which is consistent with AI features steering users toward higher-intent commercial queries.

YouTube also accelerated: ads +13% (vs. +10% in Q1), Shorts at 200M+ daily views, Shorts now monetizing at parity with in-stream watch hour in the U.S. (a milestone the company has been pursuing for 3+ years). 12.8% of total U.S. TV viewing share per Nielsen — a record. Subscriptions/Platforms/Devices +20%, with Google AI Pro and Ultra plans getting their first call-out as material subscription drivers.

Assessment: Two quarters in a row, the cannibalization thesis has gotten weaker rather than stronger. Schindler's "paid clicks up 4%" disclosure with 12% revenue growth is a clean rebuttal to the CTR-compression argument. Subscriptions/Platforms/Devices growing 20% on a $11B base is increasingly meaningful — this segment alone is now a ~$45B annualized business growing at a hyperscale rate.

Google Cloud — The Demand Signal Is Now the Story

Cloud revenue accelerated to +32% from +28% in Q1, but the more important data is the demand signature behind that growth:

• Backlog of $106B, +18% Q/Q and +38% Y/Y — translation: customers have committed to $106B of future Cloud revenue.
• Number of $250M+ deals doubled Y/Y.
• H1 2025 saw the same number of $1B+ deals as all of 2024 — i.e., the largest enterprise deals are running 2x last year's pace.
• New GCP customers up ~28% Q/Q.
• 85,000+ enterprises building with Gemini; Gemini API usage 35x Y/Y.
• Cloud at $50B+ annualized run-rate.

Cloud operating margin expanded to 20.7% from 11.3% a year ago — a ~940bps expansion, with operating income up 141% to $2.83B. Ashkenazi was explicit that capacity remains the bottleneck:

"We expect to remain in a tight demand-supply environment going into 2026." — Anat Ashkenazi, CFO

Pichai walked the analyst question of "you raised capex by $10B and demand is still tight" cleanly:

"There's obviously a time delay between this additional investment and how it will play out in future years. And so that's why both of them are true at the same time." — Sundar Pichai, CEO

Assessment: The Q1 read on Cloud — "the operating leverage thesis is real" — was confirmed and amplified. With $106B of backlog, the revenue trajectory is now substantially de-risked through 2026. The marginal question is no longer "will Cloud grow?" but "how fast can capacity be brought online?" — which is a manufacturing/supply-chain question, not a competitive/demand question. That's a fundamentally better risk profile than 12 months ago.

Other Bets / Waymo — Commercial Traction Compounds

Waymo crossed 100M cumulative autonomous miles on public roads (up from earlier disclosed milestones), launched Atlanta, more than doubled Austin service territory, expanded LA and SF Bay Area territories ~50%, launched teen accounts in Phoenix (14-17 year olds), and is testing in 10+ cities including New York and Philadelphia. The pace of city launches is accelerating quarter-over-quarter.

Ashkenazi noted explicitly that "we're allocating more resources to businesses like Waymo, where we see opportunities to create additional value" — the first time on a recent earnings call that management has signaled internal capital reallocation toward Waymo. That's a meaningful tone shift.

Assessment: Waymo continues to be the most underweighted optionality in Alphabet's stack. With 100M autonomous miles, public service in 4+ metros, and the only commercial-scale robotaxi operation in the U.S., the eventual monetization path (operate, license, or hybrid) is becoming less hypothetical. We don't yet model upside, but the tail option keeps thickening.

Key KPIs

Key Topics & Management Commentary

Overall Management Tone: Confident and increasingly emphatic. Pichai opened with "Q2 was a standout quarter" and the metrics-per-minute density of his prepared remarks was higher than Q1 — a tell that they have more to show. Schindler's pre-emptive defense of paid-click metrics shows management is anticipating the cannibalization debate but not retreating from it. Ashkenazi was the most explicit she's been about 2026 capex stepping up — a signal that she wants to manage Street expectations into the back half rather than surprise them at Q4.

Capex: $85B in 2025, Higher in 2026

The single most important new disclosure in the quarter is not the $10B raise to 2025 capex — it's the explicit pre-announcement that 2026 will be higher again:

"Looking out to 2026, we expect a further increase in CapEx due to the demand we're seeing from customers as well as growth opportunities across the company. We will provide more details on the 2026 CapEx outlook in a future earnings call." — Anat Ashkenazi, CFO

This matters for two reasons. First, the company is now signaling that this is not a one-year capex bulge but a multi-year investment cycle. Second, the demand signal embedded in the $106B backlog is what's driving it — capex follows demand, not vice versa. The depreciation drag will worsen accordingly: Q2 D&A grew 35% Y/Y to $5B, and Ashkenazi explicitly said it accelerates in Q3.

Assessment: The 2026 capex pre-announcement is the new variable in the model. We treat it as net positive — it implies management has visibility into demand far beyond the current backlog — but it does cap near-term margin upside. The right framing: Cloud margin expansion is the offset to depreciation drag; if Cloud margin keeps expanding faster than depreciation grows in absolute dollars, the math works. Q2's print supports that.

AI Mode and AI Overviews — Gen Z Adoption

Pichai called out two things specifically about AI features and Gen Z: AI features "cause users to search more as they learn that search can meet more of their needs," and that effect is "especially true for younger users." Multimodal search adoption (Lens + Circle to Search + AI Overviews) is "most pronounced among younger users."

This is the demographic Alphabet has historically been most worried about — younger users have been migrating from Search to TikTok, Instagram, and ChatGPT for queries. The AI Overviews/AI Mode/Lens combination appears to be reversing that pattern, at least for now. The virtual try-on shopping feature, the new agentic calling capability for local businesses, and the photo-to-video Gemini app feature are all designed to deepen Gen Z engagement.

Assessment: If the Gen Z reversal is real and durable, this is a massive long-term positive — younger users were the most vulnerable to search disintermediation. Worth tracking specifically in coming quarters.

Cloud's Enterprise AI Moat

Beyond the financials, the qualitative Cloud story is increasingly about enterprise AI agents:

• Wayfair, Mattel, Target, Capgemini, BBVA all named as Gemini-on-Cloud customers with specific use cases.
• Agent Development Kit: 1M+ downloads in <4 months.
• AgentSpace: 1M+ subscriptions booked ahead of GA.
• BBVA rolling Gemini/Workspace to 100K employees globally.

Pichai also named "leading AI research labs like SSI [Safe Superintelligence] and Physical Intelligence" as TPU-specific customers — a meaningful tell that frontier AI labs are choosing Google's custom silicon for inference, even when GPUs are available. SSI is Ilya Sutskever's startup; their selection of TPUs is a strong third-party validation.

Assessment: Cloud's competitive position has materially strengthened. The fact that frontier AI labs (which can use any chips on any cloud) are choosing TPUs specifically is a strong signal that Google's custom silicon strategy is paying competitive dividends. This is hard to replicate in 12-24 months.

The $1.4B Legal Settlement

Q2 included a $1.4B charge for a settlement in principle of certain legal matters (not specified in the call). It pulled Services operating margin down ~150bps and reduced consolidated operating margin from a hypothetical 33.9% to the reported 32.4%.

Management did not name the matter. The most likely candidate based on the timing and size is a privacy-related class action settlement, though the specific case isn't disclosed.

Assessment: One-time charges of this size are not new for Alphabet, and the absence of any meaningful disclosure suggests the settlement is concluded rather than ongoing. We treat it as a non-recurring item but note that legal/regulatory settlements have been a fairly persistent ~$1-2B annual headwind for the past 3 years.

Guidance & Outlook

Alphabet doesn't provide formal guidance, but management's color was substantive:

Implied Q-over-Q ramp: Q3 should see slightly tougher Y/Y comp on YouTube ads (lapping 2024 election money) and Search (lapping financial-services strength), partially offset by FX tailwinds. Cloud growth should be sustained or higher as capacity comes online, particularly weighted toward Q4. We expect Q3 consolidated growth in the +13-14% Y/Y range, with Q4 reaccelerating into mid-teens as Cloud capacity adds materialize.

Analyst Q&A Highlights

Search Monetization & AI Behavior

• Doug Anmuth (JPMorgan): Asked Schindler about paid clicks vs. pricing growth within the 12% Search growth. Schindler delivered the carefully constructed answer: "We don't manage to paid clicks and CPC targets. Some of the product and policy changes we make actually drive better monetization at the expense of paid clicks. Paid clicks were up 4% year on year." Translation: revenue per click is up materially, paid click count is up modestly — exactly the opposite of the bear thesis pattern.
  Assessment: This is the single most important Q&A exchange of the quarter. Bears now need to argue that 4% paid-click growth on +12% revenue growth equals "click degradation," which is a hard case to make.
• Eric Sheridan (Goldman): Asked Pichai about evolving consumer behavior and click monetization implications over 3-5 years. Pichai punted: "We will lead with organic experience" — meaning Gemini app and AI Mode will be ad-light initially, with monetization to follow once habits are established.
  Assessment: This is the right strategy but creates a multi-year drag on AI-feature monetization. The Q1 framing of "monetization at approximately the same rate" applies to AI Overviews specifically; AI Mode and Gemini are explicitly NOT monetized at parity yet.

Cloud Capacity & Capex

• Doug Anmuth (JPMorgan): How is Alphabet still capacity-constrained even after $10B more capex? Pichai: time delay between investment and capacity online — "both are true at the same time."
• Multiple analysts pressed on 2026 capex. Ashkenazi gave the most she would: "further increase ... more details on a future call."
  Assessment: The non-quantification is intentional. Best-case interpretation: management has flexibility to ramp into demand. Worst-case: 2026 capex could be $100B+, which would meaningfully change the FCF profile.

Subscriptions & Pricing Power

• Multiple questions on AI Pro / AI Ultra subscription monetization. Pichai noted these are getting "a boost" from AI Pro and Ultra plans; specific subscriber counts not disclosed but framed as "strong growth."
  Assessment: AI subscriptions are emerging as a meaningful new revenue line. Apple, Microsoft, and OpenAI all have subscription-based AI products at $20+/month; Alphabet finally pricing in. Worth tracking for 2026.

Waymo

• Light Waymo questioning given the operational milestones. Pichai didn't volunteer business-model framing this quarter but noted resource reallocation toward Waymo from Other Bets.

What They're NOT Saying

1. 2026 capex magnitude: Pre-announced higher, but no number. The Street will model $90-110B; the company is buying optionality by not committing.
2. The $1.4B legal settlement: Not specified. Investors should assume it's a privacy/competition matter; the lack of detail suggests resolution.
3. AI Mode / Gemini app monetization timeline: Pichai said "near-term focus on organic experience" but offered no roadmap to commercial layers. This is a multi-quarter (likely multi-year) drag on incremental AI revenue.
4. Cloud customer concentration: $250M+ deals doubling and $1B+ deals matching all of 2024 in H1 implies meaningful concentration risk among top customers — undisclosed.
5. RPO disclosure: Backlog of $106B is now disclosed (good), but the duration profile (% recognizable in next 12 vs. 24+ months) is not. AWS and Azure both disclose this.
6. DOJ Search remedy: Q2 call had zero mentions of the ongoing remedies phase. The Chrome divestment threat remains unaddressed publicly.
7. Headcount in Cloud: Total headcount is up modestly; Cloud-specific hiring intensity is undisclosed but likely well above the company average.

Market Reaction

• After-hours move: +1-2% initially, with shares moving up ~2% in early next-day trading.
• Read on the print: Generally positive — Cloud backlog and margin expansion welcomed; capex raise framed as "investing from a position of strength" rather than panic spending; 2026 capex pre-warning absorbed without major selling pressure.

The stock reaction is muted relative to the operational beat — likely because the capex raise creates near-term FCF concerns that offset some of the bullish operational data. With $106B of backlog, however, the FCF concern is more about timing than terminal value.

Street Perspective

Debate: Is the capex raise a sign of strength or a sign that returns are eroding?

Bull view: Backlog +38% Y/Y to $106B is the demand signal that's pulling capex higher. Companies don't raise capex by $10B mid-year unless customers are committing dollars. This is investing from strength.

Bear view: Capex is rising faster than revenue and FCF; 2026 will see another step-up. Eventually the depreciation curve catches up to revenue growth and margins compress structurally. The "tight supply" framing is also classic late-cycle hyperscaler behavior.

Our take: Bulls have the edge here, primarily because the backlog disclosure quantifies the demand. As long as Cloud margin keeps expanding (now 20.7% vs. 11.3% a year ago), the math works even with elevated depreciation. The risk we're pricing is asymmetric — if demand softens in 2026, the capex commitments are sunk.

Debate: Is Search monetization at risk from AI?

Bull view: Two consecutive quarters of accelerating Search growth (10% → 12%) with AI Overviews scaling from 1.5B to 2B MAU. Paid clicks +4% with revenue +12% means CPCs are rising, not falling. The cannibalization thesis lacks data support.

Bear view: AI Mode and Gemini app are explicitly NOT monetized at parity yet, and as users shift to those surfaces, "approximately the same monetization rate" will not hold. Q1+Q2 are a transition period; the real test is when AI Mode reaches AI Overviews scale.

Our take: Bears are right that AI Mode/Gemini monetization is the real test. But that test is 3-6 quarters out, and meanwhile Search keeps accelerating with AI Overviews scaling. The momentum strongly favors the bulls in the medium term. We update our prior on cannibalization downward again.

Debate: Cloud margin trajectory — peak or pause?

Bull view: 11.3% → 17.8% → 20.7% over two quarters is too steep to be a one-time mix benefit. As scale ramps and Q4 capacity comes online, margin should keep climbing — albeit lumpy.

Bear view: Cloud margin Q2 over-earns from the same factors as Q1 (mix, utilization). New Q4 capacity will bring depreciation without revenue, compressing margin in Q3-Q4 before resuming expansion in 2026.

Our take: Both partly right. We expect Q3 Cloud margin to compress 100-200bps from the Q2 print, then resume expansion through 2026. The longer-term trajectory toward 25-30% margin remains intact.

Model Update Needed

Valuation impact: Net positive on operational data, modestly negative on FCF compression. The 2026 capex framing is the swing factor — if Q3 quantifies it at <$100B, the math is comfortable; at $110B+, near-term FCF pressure becomes material.

Thesis Scorecard Post-Earnings

Overall: Thesis materially strengthened on Cloud, modestly strengthened on Search, with one new variable (2026 capex magnitude) entering the model.

Action: Maintain Outperform. The operational case is now stronger than at initiation; the new risk (2026 capex) is real but proportionate to demand.