| Jul 22, 2026 | -7.13% | Q2 FY2026 | Read transcript briefingQuarter in one view- Alphabet Q2 FY2026: consolidated revenue $119.8B, +24% reported / +23% cc — an acceleration from Q1's +22%/+19% cc, with only a 1pt FX tailwind (vs. 3pt in Q1). 12th consecutive quarter of double-digit growth. (Transcript says "$119.8 million" — clearly a transcription error for billion.)
- Operating income $40.8B, +30%; margin 34% (down from 36.1% in Q1). Services margin 41.8% (from 45.3%); Cloud margin 35.6% (from 32.9%).
- Net income and EPS "increased significantly," driven by ~$98B of other income, primarily unrealized gains on equity securities (vs. $37.7B in Q1) — the quality-of-earnings distortion more than doubled sequentially. Operating income (+30%) remains the honest growth rate. Exact net income/EPS figures were not stated on the call.
- 2026 CapEx guidance raised again to $195–205B (from $180–190B) on "acceleration in delivery of capacity"; 2027 still guided to "increase significantly," details deferred. Q2 CapEx $44.9B; FCF turned negative: −$5.9B (vs. +$10.1B in Q1) — the FCF compression bear case has now fully arrived.
- Google Cloud $24.8B, +82% (from +63%) — with first-ever TPU system sales into customer data centers recognized in Q2; growth "accelerated meaningfully even excluding TPU system sales." Operating income $8.8B, more than tripled; margin 35.6%. Backlog $514B, +$52B QoQ (a sharp deceleration in sequential adds from +$222B in Q1); majority is standard GCP; just over 50% converts within 24 months.
- Search $63.3B, +17% (decelerating from +19%); YouTube ads $11.1B, +13% (from +11%, World Cup-assisted); Subscriptions/Platforms/Devices $12.9B, +15% (from +19%); Network $7.3B, −1% (improved from −4%).
- Balance sheet transformation continues: cash & securities $242.5B (includes $87.1B marketable equities); long-term debt $98.2B (from $77.5B; ~$16B a year ago per Ashkenazi) — and an equity raise was completed this quarter (first disclosure of equity issuance; an ATM program continues to cover SBC taxes).
What management is focused on- Pichai's framing: "exciting momentum," AI investments "redefining what's possible." The model narrative dominated: new Gemini 3.6 Flash and 3.5 Flash Lite, Gemini 3.5 Flash Cyber (paired with a "code mender" agent for vulnerability find/fix), Gemini 3.5 Pro in testing, and — the headline — Gemini 4 pretraining underway, "our most ambitious pretraining run yet."
- Model-cadence and frontier credibility were the call's dominant theme — three of nine analysts (Anmuth, Sandler, Nathanson) pressed on release pace, coding gaps, and moats. Pichai conceded coding/agentic coding is an acknowledged gap, promised "monthly cadence" iteration, and leaned on Gemini 4 as the frontier answer.
- Supply constraint is now structural policy: third-party capacity will be used as a "bridging strategy" starting Q3, with modest Cloud margin pressure explicitly pre-flagged — a new cost of the constraint beyond lost revenue.
- TPU externalization is now operating reality, not announcement: revenue recognition began, inventory build is hitting operating cash flow, and a Blackstone data-center project was cited as a TPU placement vehicle.
- Capital structure is a new focus area: Ashkenazi laid out a three-source funding framework (operating cash flow → debt → equity) and defended the equity raise as balance-sheet resilience — the first time Alphabet has framed equity issuance as part of the AI funding plan.
- Efficiency proof points continue: AI Mode response cost reduced to its lowest level since launch; Antigravity at 2.4M weekly active users; a Chrome team compressing a 2-year roadmap into 3 months (8x) via model-driven refactoring.
Key numbers and quarter mechanics- Revenue $119.8B, +24%/+23% cc. Cost of revenue $45.9B, +18% (TAC $16.2B, +10%; other cost of revenue $29.8B, +22% — depreciation, inventory costs from TPU system sales (new driver), YouTube content).
- OpEx $33.1B, +27% — R&D +32% (AI talent comp + depreciation); S&M +18% (Gemini app and Search marketing); G&A +24% (compensation and legal/other charges — third straight quarter of elevated G&A).
- Operating margin 34% (from 36.1%). Services operating income $39.5B, +20%, margin 41.8% (down from 45.3% — notable compression). Cloud operating income $8.8B, margin 35.6% (from 20.7% a year ago).
- CapEx $44.9B in Q2 (~60% servers / 40% data centers & networking — same split). H1 CapEx ~$80.6B against a $195–205B guide implies a ~$115–125B second half — another steep back-half ramp.
- FCF −$5.9B in Q2; $53.3B TTM (from $64.4B TTM in Q1, $73.3B FY2025). Operating cash flow $39.1B in Q2; $185.7B TTM. TPU inventory build is now a stated drag on operating cash flow.
- Balance sheet: cash & securities $242.5B; long-term debt $98.2B, +$20.7B QoQ; equity raise completed (size not stated on the call — check the 10-Q); dividend held at $0.22/quarter.
- Other Bets: revenue $382M; operating loss $1.8B (narrowed from $2.1B). Alphabet-level activities loss $5.8B, driven by shared AI R&D.
- Backlog mechanics: $514B, +$52B QoQ; majority standard GCP; TPU systems in backlog; >50% converts within 24 months; TPU revenue "relatively small" in 2026, "vast majority" in 2027.
- Q3 pre-flags: slight FX headwind (vs. +1pt tailwind in Q2), hitting Search and YouTube ads; Search begins lapping the Q3 2025 acceleration; third-party capacity creates modest Cloud margin pressure; Wiz integration headwind continues; depreciation and energy costs rising; FCF "will remain under pressure."
Product and launch scorecard- TPU hardware sales: first revenue recognized in Q2 — the Q1 announcement became a P&L line within one quarter. Small in 2026, ramping into year-end, vast majority in 2027. Margin impact refused twice (Nathanson asked directly; Ashkenazi: "we don't break out margins"). Blackstone project named as a placement vehicle. Inventory build now visible in cash flow.
- Gemini app MAU restored: 950M (from 750M in Q4; metric was withheld in Q1), with DAUs tripling YoY. The metric's return after a one-quarter absence partially validates the Q1 suspicion that it was withheld during a soft patch — or the Q1 omission was noise; either way, +200M in two quarters is strong.
- AI Mode: 1B+ MAU since global expansion last October; AI Overviews and AI Mode merged into "one seamless search experience"; billions of clicks/week sent to websites via AI features.
- Model cadence: 3.6 Flash (+10 points on an agentic-coding benchmark vs. 3.5 Flash in six weeks, more token-efficient), 3.5 Flash Lite, 3.5 Flash Cyber; 3.5 Pro in testing; Gemini 4 in pretraining — Pichai committed to near-monthly release cadence going forward.
- API tokens: ~22B/minute, up from 16B (+~38% QoQ, decelerating from +60% QoQ). 9M+ developers building monthly.
- Gemini Enterprise: ~90% of Fortune 100 using it; Agent Development Kit ~70M downloads; customers PepsiCo, Intel, HSBC, Bell Canada, Macy's, Signal Iduna. Seat/company counts (8M/2,800) still not updated — the metric switch from Q1 is now permanent-looking.
- Cloud demand metrics: existing customers exceeding commitments by >50% (from 45% — accelerating again); new-customer acquisition velocity more than doubled YoY; Marketplace transactions +7x YoY; ~500 customers processed 1T+ tokens (from 330); 2,000+ enterprises at 100B+ tokens.
- Security: 90% of Fortune 100 are Cloud security users; AI-workloads scanned +45% QoQ; new AI threat-defense offering with the cyber model; Morgan Stanley, TELUS, Texas Children's, Atlassian named.
- Omni (video creation, launched at I/O): +40% DAU creating videos on Gemini app since May. Gemma open models: 900M cumulative downloads; Gemma 4 at 300M+ since April.
- Ads/commerce: AI Max at 500K advertisers; AI-powered campaigns deliver +15% conversions/value at similar ROAS; shopping ads relevance +20%; Direct Offers adds IHG; UCP live with Target and Steve Madden; Universal Card announced (cross-retailer single checkout); Buy with Google Pay on CTV (2-click TV purchases).
- YouTube: World Cup — 1.7B unique viewers (Pichai's "1.7 million" is a transcription error; Schindler said 1.7 billion), 550M on TVs, most-viewed World Cup in YouTube history; Ask YouTube: 140M users in June; subscriptions still growing faster than ads.
- Waymo: first 6th-gen Driver vehicle (Ojai) opened to public riders — but no rides/week or city-count update this quarter, breaking the restored-disclosure streak. Wing passed 1M deliveries. Isomorphic Labs raised $2B+ externally.
Sell-side read-through- Nine analysts; tone still constructive, but the questioning center of gravity shifted to model competitiveness — three straight questions on frontier position, release cadence, and moats (Anmuth, Sandler, Nathanson). That is a new anxiety: last quarter it was CapEx and TPU strategy.
- Still no antitrust/DOJ question — fifth consecutive quarter. No question on the ~$98B unrealized-gain distortion, negative FCF, or the equity raise's size/dilution.
- Nowak (Morgan Stanley): how has the GenAI ROIC view changed vs. a year ago — Pichai: "more bullish," early innings, no metrics. CapEx budgeting philosophy — Ashkenazi: invest "as long as we see an attractive return," multiyear view; no hurdle, no number — ROIC deflection streak extends to five quarters.
- Anmuth (JPMorgan): confidence in staying at the frontier; coding gap — Pichai explicitly acknowledged the coding/agentic-coding gap, pointed to 3.6 Flash gains and Gemini 4. Capital structure after equity/debt raises — Ashkenazi gave the cash→debt→equity framework; no plans to return to equity markets except the SBC-tax ATM; debt went ~$16B → ~$100B in a year.
- Sheridan (Goldman): TPU internal/external balance — Pichai: frontier training first, then serving, then external placements (own or third-party data centers). TPU share of backlog and margin impact — Ashkenazi confirmed TPUs in backlog, majority GCP; margin question not answered.
- Sandler (Barclays): model-release speed and the crowded low-cost end — Pichai committed to near-monthly cadence and full Pareto frontier (Flash Lite → Pro). A "third-party compute deal with SpaceX" was referenced by Sandler without management comment.
- Nathanson (MoffettNathanson): moats if models commoditize — Pichai: models are an ingredient in end-to-end solutions (security, data analytics), data confidentiality, orchestration. TPU margin vs. core Cloud margin — refused directly ("we don't break out margins").
- Shmulik (Bernstein): where constraints bind and compute-allocation trade-offs — Pichai restated the hierarchy (AGI training baseline, then Search/YouTube, then Cloud serving) and gave the most candid economics of the call: bridge deals are high-cost for ~6 months but "highly ROI positive" over the multiyear deal life.
- Josey (Citi): Search monetization drivers — Schindler: all verticals contributed, retail largest; Gemini across ads quality/tools/experiences; AI Max unlocking "billions of net new searches." Still no quantification of behavior-vs-tools attribution — unanswered across multiple quarters now.
- Gawrelski (Wells Fargo): 2027 capacity returns vs. 2025/26 given supply-chain inflation — Pichai: input costs factored into pricing, "dynamics look healthier than a year ago." Waymo corporate structure / spin-out conditions — Pichai declined to engage ("focused on scaling"), the most direct Waymo-separation question asked to date.
- Kajuria (Wolfe): does TPU become a merchant silicon business — Pichai notably did not rule it out ("I don't want to project out too far... we will scale up based on the opportunities we see"). YouTube growth drivers — Schindler: CTV, creator tools, Demand Gen, Shorts, shoppable formats.
Management credibility- TPU delivery was fast and as-guided: announced Q1, first revenue recognized Q2, timing profile (small 2026, vast majority 2027) reiterated exactly. On execution, this builds credibility; on disclosure, margins remain refused twice.
- CapEx guide raised for a second consecutive quarter ($175–185B → $180–190B → $195–205B) with a stated reason each time — transparent mechanics, but the pattern is now a ratchet, and 2027 remains an open-ended "significant increase" with details perpetually deferred.
- ROIC deflection streak hits five quarters. New formulation from Ashkenazi: "we will continue to invest as long as we see attractive returns" — still no quantified hurdle, and Nowak's direct "how has your ROIC view changed" got an adjective ("more bullish"), not a number.
- Gemini app MAU returned at 950M after its Q1 absence — the metric-switch concern from last quarter is partially resolved, but the episode confirms metrics are disclosed opportunistically.
- Waymo metrics disappeared this quarter (no rides/week, no city count) right as Gawrelski asked about a spin-out — the same promote-while-growing, withhold-when-convenient pattern seen with Gemini MAU.
- The equity raise was disclosed only in passing, in response to an analyst question, with no size, price, or dilution context on the call. For a company funding with equity for the first time in this cycle, that is thin disclosure.
- ~$98B of unrealized gains inside the print was disclosed but again not contextualized; net income growth was described only as "significant" with no figure given on the call.
- Pichai's candid admission of the coding gap and the honest bridge-deal economics (6 months of high cost for multiyear ROI) were unusually direct answers — credibility positives amid the deflections.
- Services margin compression (45.3% → 41.8%) was not explained on the call beyond the cost lines — worth pressing next quarter.
What changed versus the prior quarter- CapEx guide raised again to $195–205B; Q2 CapEx $44.9B vs. $35.7B; FCF went negative (−$5.9B vs. +$10.1B) — the first negative-FCF quarter of the build-out.
- Funding mix escalated: debt $98.2B (from $77.5B) plus a completed equity raise — equity is now part of the AI funding stack, with an ATM continuing.
- TPU sales moved from announcement to revenue; TPU inventory now a stated cash-flow drag; Blackstone named as a placement partner; merchant-silicon ambition left conspicuously open.
- Cloud accelerated to +82% from +63%; margin 35.6% from 32.9%; but backlog sequential adds collapsed to +$52B from +$222B — the biggest single deceleration in the print.
- Search decelerated to +17% from +19% with a Q3 lap warning; YouTube ads +13% from +11% (World Cup); Subscriptions decelerated to +15% from +19%; Network improved to −1% from −4%.
- Services margin compressed 3.5pts to 41.8%; consolidated margin down to 34%.
- Other income ballooned to ~$98B from $37.7B — earnings quality worsened further.
- Gemini app MAU restored (950M); Waymo metrics withheld; Isomorphic Labs raised $2B+; Wing hit 1M deliveries; Waymo 6th-gen vehicle launched.
- New cost pre-flags: third-party capacity bridge (Q3 Cloud margin pressure) and a slight Q3 FX headwind — both new this quarter.
Bull case- Cloud is compounding at +82% with margin at 35.6% — and growth "accelerated meaningfully" even excluding TPU hardware, so the core franchise, not just chip sales, is driving it. Customers are beating commitments by >50% (accelerating), acquisition velocity doubled, Marketplace +7x.
- TPU externalization is now a real, revenue-generating business one quarter after announcement, with demand from labs, financial services, pharma, and robotics — and management pointedly declined to rule out a full merchant-silicon strategy.
- The demand signal justifies the spend: backlog $514B, >50% converting in 24 months, and management is now paying premium prices for third-party bridge capacity — rational only if multiyear deal economics are strongly positive, which Pichai explicitly asserted.
- Gemini app at 950M MAU with DAUs tripling, AI Mode at 1B+ MAU, and AI Mode response costs at all-time lows — consumer AI scale and unit economics are improving simultaneously.
- Model story has a credible next leg: Gemini 4 pretraining is the "most ambitious" run yet, cadence is committed to near-monthly, and the coding gap is acknowledged with measurable progress (+10 points in six weeks).
- YouTube monetization surface keeps expanding: Buy with Google Pay on CTV, Ask YouTube at 140M users, subscriptions outpacing ads, record World Cup engagement.
- Isomorphic's $2B+ external raise validates the Other Bets playbook: external capital funds the optionality while Alphabet retains upside.
Bear case- FCF is now negative (−$5.9B) with a raised $195–205B guide, a pre-committed 2027 increase, debt at $98.2B (6x a year ago), and — the real escalation — equity issuance. The funding ladder (cash flow → debt → equity) is being climbed in order, and there is still no FCF floor, hurdle rate, or ROIC number after five quarters of asking.
- Earnings quality is the worst on record: ~$98B of unrealized equity gains against $40.8B of operating income. The headline net income number is now predominantly a mark-to-market artifact; any reversal swings violently the other way.
- Backlog momentum broke: +$52B sequential vs. +$222B in Q1. One quarter is not a trend, but the backlog-adds deceleration alongside a capacity-bridge strategy raises the question of whether the best deals are already signed.
- Margin pressure is arriving from multiple directions at once: Services margin −3.5pts QoQ (unexplained), third-party capacity pressuring Cloud margins from Q3, Wiz integration headwind, depreciation and energy costs rising, G&A +24% for a third straight quarter.
- Search decelerated to +17% with management pre-flagging a tough Q3 lap and an FX headwind — the three-quarter acceleration streak is over, and AI-surface monetization remains unquantified (no revenue-per-query data ever provided).
- Disclosure discipline is slipping where it matters: Waymo metrics withheld the same quarter a spin-out question was asked; equity raise size undisclosed on the call; TPU margins refused twice; Services margin compression unexplained.
- The model-competitiveness questions from three analysts signal the market now sees frontier risk as a live issue — Pichai's answer rests heavily on Gemini 4, an unreleased model. If Gemini 4 disappoints, the full-stack premium narrative weakens.
- Regulatory risk remains completely unaddressed — five consecutive calls with no DOJ question or update.
Next-quarter watchlist- CapEx/FCF mechanics: Q3 CapEx vs. the implied ~$115–125B H2 ramp; whether FCF stays negative; debt issuance off $98.2B; equity raise size, pricing, and ATM pace (check the 10-Q); any 2027 CapEx detail; any ROIC quantification (five-quarter deflection streak).
- Backlog: whether sequential adds re-accelerate from +$52B or confirm a slowdown; TPU vs. GCP mix; any color on duration/cancellation terms of the >24-month tail.
- TPU systems: revenue ramp into year-end ("small" in 2026); margin commentary (refused twice — watch for a third); the Blackstone project's structure; any merchant-silicon strategy language.
- Cloud margins: the quantified impact of third-party bridge capacity and Wiz in Q3; whether 35.6% was the peak; GCP vs. segment growth.
- Search: the flagged Q3 deceleration — how much of the +17% holds against the lap and FX headwind; Direct Offers/IHG progress; Universal Card adoption; paid clicks/CPC in the 10-Q.
- Services margin: whether the 41.8% print gets explained and whether it recovers — the largest unexplained margin move of the quarter.
- Gemini 4: training progress signals, release timing, and whether the near-monthly cadence commitment holds; coding-benchmark progress vs. competitors.
- Waymo: whether rides/week and city-count disclosure returns; Ojai/6th-gen ramp; any structural-separation signals after Gawrelski's question went unanswered.
- Other Bets: Isomorphic post-raise trajectory; GFiber/Astound close (expected Q4); whether the $1.8B loss keeps narrowing.
- Legal/regulatory: five quarters of silence — DOJ remedies remain a live, undisclosed risk; G&A legal charges bear watching in the 10-Q.
|
| Apr 29, 2026 | +9.96% | Q1 FY2026 | Read transcript briefingQuarter in one view- Alphabet Q1 FY2026: consolidated revenue $109.9B, +22% reported / +19% cc — an acceleration from Q4's +18%/+17% cc, with a 3pt FX tailwind in the quarter (guided to ~1pt in Q2). 11th consecutive quarter of double-digit growth.
- Operating income $39.7B, +30%; margin 36.1% — a clean quarter with no Waymo-style charge; Services margin 45.3% (from 41.9%), Cloud margin 32.9% (from 30.1%).
- Net income $62.6B, +81%; EPS $5.11, +82% — massively flattered by $37.7B of other income, primarily unrealized gains on nonmarketable equity securities (vs. $3.2B in Q4). This is the largest quality-of-earnings distortion to date; operating income (+30%) is the honest growth rate.
- 2026 CapEx guidance raised to $180–190B (from $175–185B) to include the now-closed Intersect acquisition — and, more consequentially, 2027 CapEx guided to "significantly increase" versus 2026. Q1 CapEx $35.7B; FCF $10.1B (vs. $24.6B in Q4) — the FCF compression predicted last quarter is arriving.
- Google Cloud $20.0B, +63% (from +48%) — first $20B+ quarter; operating income $6.6B, tripled YoY; margin 32.9%. Backlog $462B, nearly doubled sequentially (+$222B QoQ vs. +$85B in Q4) — includes new TPU hardware sales; just over 50% converts to revenue within 24 months.
- Search $60.4B, +19% (from +17%) — third straight acceleration. YouTube ads $9.9B, +11% (from +9%). Subscriptions/Platforms/Devices $12.4B, +19% (from +17%). Network $7.0B, −4%.
- Strategic reversals/milestones: TPUs will be sold as hardware into select customers' own data centers — a direct reversal of Pichai's Q4 non-endorsement; Wiz closed in March (Cloud segment, low-single-digit margin headwind for rest of 2026); Verily deconsolidated after external raise; GFiber to deconsolidate via Astound combination (expected Q4 close); Waymo 500K+ rides/week, 11 US cities; dividend raised 5%; long-term debt jumped to $77.5B from $46.5B.
What management is focused on- Pichai's framing: "terrific quarter," full-stack AI "driving performance across our business." The center of gravity has shifted decisively to Cloud: enterprise AI solutions are now Cloud's primary growth driver "for the first time," with GenAI-model product revenue +~800% YoY (from +~400% in Q4, +~200% in Q3).
- TPU externalization is the quarter's strategic pivot: 8th-gen TPUs (8t training: 3x Ironwood processing; 8i inference: 80% better performance/dollar) announced at Cloud Next, and TPUs will ship to select customers' own data centers (capital markets firms, frontier AI labs cited). Pichai now frames it as "opportunistic" within a Cloud/ROIC framework — a softening of last quarter's refusal.
- Supply constraint remains the binding constraint, now with a P&L consequence admitted: "our cloud revenue would have been higher if we were able to meet the demand." Compute allocation hierarchy disclosed for the first time: GDM/frontier training first, then long-range plans for Search/YouTube/Cloud.
- Efficiency narrative continues with new numbers: cost of core AI responses down >30% since upgrading AIO/AI Mode to Gemini 3; search latency down >35% over 5 years; Antigravity now driving "fully autonomous digital task forces" internally.
- Portfolio pruning in Other Bets is new and deliberate: Verily deconsolidated, GFiber being sold — management explicitly framing Other Bets prioritization while concentrating resources on Waymo.
- Ashkenazi's discipline messaging now leans on results as the ROIC proof: "you're seeing the proof point, the ROIC on that in terms of just the growth rate" — growth itself is now the stated return evidence, still without quantified hurdles.
Key numbers and quarter mechanics- Revenue $109.9B, +22%/+19% cc. Cost of revenue $41.3B, +14% (TAC $15.2B, +11%; other cost of revenue $26.0B, +15% — depreciation, YouTube content, compensation).
- OpEx $28.9B, +24% — R&D +26% (AI talent comp + depreciation); S&M +23% (Gemini app and Search marketing); G&A +21% (legal and other compensation — second straight quarter of elevated G&A, different stated driver than Q4's charitable timing).
- Operating margin 36.1% — no one-off charges this quarter. Services operating income $40.6B, +24%, margin 45.3%. Cloud operating income $6.6B, margin 32.9% (from 17.8% a year ago).
- CapEx $35.7B in Q1 (~60% servers / 40% data centers & networking — same split). Annualized pace (~$143B) is well below the $180–190B guide, implying a steep back-half ramp.
- FCF $10.1B in Q1; $64.4B TTM (down from $73.3B FY2025). Operating cash flow $45.8B in Q1; $174.4B TTM.
- Balance sheet shift: cash $126.8B; long-term debt $77.5B, up $31B sequentially — the build-out is now meaningfully debt-funded. Dividend raised 5%; no buyback figure given on the call (check the 10-Q/press release).
- Other Bets: revenue $411M; operating loss $2.1B — wider than Q4's ex-charge ~$1.5B (inference), with no charge cited this quarter; deconsolidations (Verily done, GFiber pending) will shrink this segment.
- Backlog mechanics: $462B, >50% recognized within 24 months; TPU hardware revenue "small percent" late 2026, "vast majority" in 2027, and will be lumpy quarter-to-quarter.
- Q2 pre-flags: FX tailwind ~1pt (vs. 3pt in Q1); Wiz low-single-digit headwind to Cloud operating margin for remainder of 2026; depreciation pressure continues; continued AI/cloud hiring and AI marketing spend.
Product and launch scorecard- TPU hardware sales (new revenue stream): announced for select customers' own data centers; already embedded in the $462B backlog; revenue mostly 2027. This is the quarter's biggest product-strategy event — and a reversal of Q4's stance, delivered without acknowledging the change.
- Gemini 3.1 series (Pro, Flash, Flash Live audio): shipped with frontier claims; 3.1 Flash Live now powering conversational features in Search and Gemini app; speech-to-text in 70 languages. API tokens 16B+/minute, up from 10B (+60% QoQ — accelerating from +43% QoQ last quarter).
- Gemini app MAU not disclosed — a notable omission after 650M→750M was a headline for two quarters. Management instead cited "strongest quarter ever for consumer AI plans" and 350M total paid subscriptions (from 325M). The metric switch warrants scrutiny.
- Gemini Enterprise: paid MAU +40% QoQ (metric changed from seats/companies — 8M seats/2,800 companies not updated); named customers Bosch, Merck, Mars; partner-sold seats +9x YoY; BigQuery Gemini-powered workflows +30x YoY.
- Cloud AI: GenAI-model product revenue +~800% YoY (accelerating from ~400%); $100M–$1B deals doubled YoY plus multiple $1B+ deals; existing customers outpacing commitments by 45% (from 30%+); 330 customers processed 1T+ tokens each over 12 months, 35 at 10T+.
- Wiz: closed March; "performance so far has exceeded our expectations"; agentic defense customers Deloitte, Priceline, Shell. Low-single-digit Cloud margin headwind for 2026 — integration cost is real but bounded.
- AI Mode monetization: Direct Offers pilot expanded (Gap, L'Oréal, Chewy signed); new retail ad format in test (retailers displayed against organic product recommendations); Ulta Beauty launched full agentic checkout in AI Mode/Search/Gemini app — first live end-to-end commerce. Still no monetization-rate data.
- UCP momentum: Amazon, Meta, Microsoft, Salesforce, Stripe joined the tech council (with Shopify, Etsy, Target, Wayfair) — genuine industry coalescence, including direct competitors.
- AI Max: out of beta; Hilton EMEA 1/3 more clicks at 1/5 of spend, +55% booking value; Etsy +10% search volume, 15% net-new queries; >30% of search advertisers now on AI-enabled campaigns.
- YouTube: ads +11%; 200M+ daily US living-room hours; 10M+ channels publishing Shorts daily; largest-ever quarterly nontrial subscriber adds for Music/Premium; Premium Lite in 23 countries, dozen+ more in Q2; subscriptions growing faster than ads.
- Waymo: 500K+ rides/week (doubled in <1 year), 11 US cities, 6 new cities in 2026 including Nashville; delivery expansion with Walmart/DoorDash; Bay Area plans announced.
- Other: Gemma 4 (50M downloads in weeks; 500M cumulative open-model downloads); Lyria 3 (150M songs); Nano Banana 2 (1B images in half the time of v1); Maps' biggest upgrade in a decade; Pixel 10a launched.
Sell-side read-through- Nine analysts; tone constructive-to-credulous. No antitrust/DOJ question for a fourth consecutive quarter; no question on the $37.7B unrealized-gain distortion, the $31B debt increase, the missing Gemini app MAU, or the TPU-strategy reversal.
- Nowak (Morgan Stanley): compute deployment for Search ROIC — Pichai gave agentic-flows vision, no metrics. TPU pricing philosophy — Pichai reframed as Cloud-first, "opportunistic," ROIC-based, with scale-economics benefit; no pricing answer.
- Anmuth (JPMorgan): 2027 CapEx trajectory vs. backlog — Ashkenazi: demand-driven, "responsible," more clarity on future calls; no number. Search ad-load coverage above the historical ~20% — Schindler: "there is upside in that coverage number" via better intent understanding — a meaningful monetization signal.
- Sheridan (Goldman): competitive positioning of the vertical stack — Pichai: "only provider" with full stack plus security. UCP's services-business implications — Schindler: vision answer, no economics.
- Sandler (Barclays): price/volume evolution of core AdWords in an agentic world — Schindler deflected entirely to user experience; no answer on pricing mechanics.
- Nathanson (MoffettNathanson): capacity-allocation screens — Pichai gave the hierarchy (GDM training first, then long-range plans) and admitted Cloud revenue is being left on the table. Gemini app ads — Schindler: focus is AI Mode, formats would transfer, "not rushing anything" — language identical to Q4; still no timeline.
- Shmulik (Bernstein): can search strength be dimensionalized between user behavior change and advertiser tools — Schindler gave an expansionary-nature list, no quantification; the attribution question remains unanswered across quarters.
- Josey (Citi): cloud margin drivers vs. the "AI revenue is lower margin" thesis — Ashkenazi credited top-line leverage and infrastructure efficiency, acknowledged depreciation headwinds; did not directly address AI-deal margin mix.
- Gawrelski (Wells Fargo): supply-chain inflation in 2026/27 CapEx — Pichai: factored in, scale helps; no specifics. Internal/external compute split update — not answered (Q4's >50%-to-Cloud disclosure not refreshed). Subscription monetization of advanced search use cases — Pichai endorsed tiered subscriptions as the model.
- Post (BofA): TPU opportunity and backlog split — Ashkenazi confirmed TPUs in backlog, majority still GCP, >50% converts in 24 months. Margins on giant GenAI cloud deals — Pichai: no contract comment, "robust ROIC framework" — the fourth consecutive quarter of ROIC language without a number.
Management credibility- Delivered and then raised the 2026 CapEx guide ($180–190B) with a stated reason (Intersect close) — transparent mechanics; but pre-committing to a "significant" 2027 increase before giving any 2027 detail extends the open-endedness.
- The TPU reversal is the quarter's credibility event: in Q4 Pichai "effectively declined" external TPU sales as a revenue stream; one quarter later it's announced, in backlog, and guided. The strategy change may be correct, but it was presented as continuity ("that's the framework with which we think about it") rather than a change — and no analyst called it out.
- ROIC deflection streak extends to four quarters (Post Q2, Nathanson Q3, Shmulik Q4, now Ashkenazi/Pichai repeatedly). New rhetorical move: pointing to revenue growth itself as "the proof point, the ROIC" — growth is not a return metric, and the substitution went unchallenged.
- Gemini app MAU disappeared after being a headline metric at 650M→750M; replaced by subscription-plan strength. Combined with Q4's refusal to give engagement numbers under Sandler's questioning, the consumer-AI disclosure pattern is now: promote metrics while they grow, swap them when convenient.
- Waymo disclosure continues to improve: 500K rides/week (from 400K), 11 cities, doubling in under a year — consistent delivery on the restored metric.
- Cloud backlog disclosure improved materially: TPU inclusion flagged, >50%/24-month conversion given, revenue-timing profile for hardware stated — genuinely useful new transparency.
- The $37.7B other-income gain was disclosed cleanly but not contextualized; an +81% net income headline against +30% operating income demanded more framing than it got.
- Schindler's non-answers were conspicuous this quarter: AdWords price/volume mechanics (Sandler), behavior-vs-tools attribution (Shmulik), and Gemini app ads timing (Nathanson) all received vision statements instead of data.
What changed versus the prior quarter- CapEx: guide raised to $180–190B and 2027 pre-committed to a "significant increase" — the open-endedness bears flagged last quarter is now explicit policy.
- TPU strategy reversed: from Pichai's Q4 non-endorsement to hardware sales in customer data centers, embedded in backlog, revenue mostly 2027.
- Cloud accelerated again: +63% from +48%; backlog $462B (+$222B QoQ) from $240B (+$85B) — sequential adds nearly tripled; margin 32.9% from 30.1%.
- Search accelerated to +19% from +17%; YouTube ads re-accelerated modestly to +11% from +9%; Subscriptions re-accelerated to +19% from +17%; Network worsened to −4% from −2%.
- FCF compressed to $10.1B from $24.6B as CapEx ramped to $35.7B from $27.9B — the bear-case mechanics are now visible in the numbers.
- Long-term debt jumped $31B to $77.5B — funding mix shifted materially in one quarter; dividend raised 5% (buyback undisclosed on call).
- M&A closed: Wiz (March) and Intersect (March) — both pre-flagged, both now in segments/CapEx; Wiz carries a disclosed margin headwind.
- Other Bets rationalization began: Verily deconsolidated, GFiber sale pending — a new portfolio-discipline signal; Other Bets loss $2.1B with no charge cited.
- Earnings quality shifted further below the line: other income $37.7B vs. $3.2B in Q4 — net income growth (+81%) now more than double operating growth (+30%).
- Gemini app MAU metric dropped; consumer AI narrative moved to subscriptions (350M, "strongest quarter ever" for AI plans).
- Waymo: 400K → 500K+ rides/week; 6 → 11 cities; no new funding round or SBC charge this quarter.
Bull case- Demand is compounding faster than capacity: Cloud +63% with a $462B backlog that added $222B in one quarter, deals doubling across size bands, customers beating commitments by 45% (accelerating), and management admitting Cloud revenue is supply-capped — the constraint is physical, not demand.
- Cloud margin reached 32.9% while depreciation accelerates and before Wiz dilution — the operating-leverage proof is now multi-quarter and directly rebuts the "AI revenue is low margin" thesis, at least at the segment level.
- Search has accelerated three straight quarters (+15% → +17% → +19%) with all major verticals contributing, and Schindler explicitly flagged upside to the historical ~20% ad-coverage ceiling — a concrete, previously unstated monetization lever.
- The TPU externalization opens a genuinely new hardware revenue stream with demand from capital-markets firms and frontier labs, already in backlog — and the 8i/8t specs (80% better performance/dollar) support competitiveness.
- Agentic commerce moved from protocol to production: Ulta live checkout in AI Mode/Gemini, UCP council now includes Amazon, Meta, Microsoft, Salesforce, Stripe — ecosystem adoption by competitors is unusual validation.
- Efficiency gains are stacking: >30% cut in core AI response costs since Gemini 3, on top of 2025's 78% serving-cost reduction — unit economics improving while volume explodes (16B tokens/min, +60% QoQ).
- Waymo is scaling (500K rides/week, doubling <1 year, 11 cities) while Other Bets is being pruned (Verily, GFiber out) — the segment's drag should narrow structurally.
- Subscriptions re-accelerated to +19% with AI plans as a stated driver and 350M paid subs — a second consumer monetization engine that doesn't depend on ad formats.
Bear case- FCF is collapsing into the CapEx ramp: $10.1B this quarter against a $180–190B 2026 guide and a pre-committed "significant" 2027 increase; Q1 CapEx annualizes well below guide, so the spend back-loads. Debt already up $31B in one quarter. Still no FCF floor, hurdle rate, or ROIC number — four straight quarters of deflection, now with growth itself offered as the return proof.
- Earnings quality is the worst in this cycle: $37.7B of unrealized equity gains inside an +81% net income print; operating income grew +30% — strong, but the headline overstates it by ~2.7x. Any mark-to-market reversal swings the other way.
- The TPU pivot cuts both ways: selling scarce compute as hardware while admitting Cloud revenue is supply-constrained means either margins on hardware are exceptional (undisclosed — Post's margin question was refused) or allocation is being diluted; revenue is also back-loaded to 2027 and "will fluctuate."
- Backlog quality questions: $222B sequential add includes TPU hardware of undisclosed size; >50% converts in 24 months, but the remainder's duration, cancellation terms, and margin profile are unknown.
- The highest-growth consumer surface lost its metric: Gemini app MAU undisclosed this quarter after being a headline — consistent with a pattern of metric-switching when numbers stop helping (cf. AI Mode basis switch in Q3).
- Monetization of AI surfaces remains unquantified: Direct Offers still a pilot, Gemini app ads still "not rushing," no revenue-per-query data ever provided; Sandler's core AdWords price/volume question — the central bear question on agentic search — was entirely evaded.
- Wiz dilutes Cloud margin for the rest of 2026; depreciation pressure continues; G&A +21% for a second straight quarter with shifting explanations.
- Regulatory risk remains completely unaddressed on four consecutive calls — no disclosure, no questions, no update on DOJ remedies.
Next-quarter watchlist- CapEx/FCF mechanics: Q2 CapEx vs. the implied back-half ramp to reach $180–190B; whether FCF stays positive; debt issuance pace off $77.5B; buyback level (undisclosed on this call — check the 10-Q); any 2027 CapEx detail ("more clarity in future earnings calls" was promised); any ROIC quantification (four-quarter deflection streak).
- TPU hardware: revenue recognition start ("small percent" later this year); backlog split between TPU and GCP; margin commentary on hardware vs. cloud services; customer names beyond the cited verticals.
- Cloud: whether +63% holds against tougher comps and Wiz consolidation; backlog off $462B; margin vs. 32.9% with the disclosed Wiz low-single-digit headwind; GCP growth rate vs. segment (still "much higher"); any update on the >50%-to-Cloud compute split (Gawrelski's unanswered question).
- Search: whether +19% holds as FX tailwind fades to ~1pt; Direct Offers expansion beyond pilot; the new retail ad format's results; any movement on the ~20% ad-coverage ceiling Schindler flagged; paid clicks/CPC in the 10-Q.
- Gemini app: whether MAU disclosure returns at I/O (May 19) — its absence this quarter needs explanation; any ads announcement (Schindler's language unchanged: "not rushing"); AI-plan subscription economics.
- Gemini Enterprise: whether seat/company counts (8M/2,800) are updated or the paid-MAU metric (+40% QoQ) becomes permanent; any ARR or revenue-per-seat disclosure.
- Waymo: rides/week trajectory off 500K; city count vs. the 6-added-in-2026 pace; Bay Area launch; any new capital raise and associated SBC charge recurrence.
- Other Bets: GFiber/Astound close (expected Q4) and the segment's post-deconsolidation shape; whether the $2.1B loss narrows as Verily exits.
- YouTube: ad growth vs. +11%; Premium Lite's dozen-plus new countries in Q2; Brandcast (May 13) upfront signals.
- I/O, GML, Brandcast (all May): product announcements that could reset Search/Gemini monetization narratives before Q2 results.
- Legal/regulatory: four quarters of silence — DOJ remedy proceedings remain a live, undisclosed risk.
|
| Feb 4, 2026 | -0.54% | Q4 FY2025 | Read transcript briefingQuarter in one view- Alphabet Q4 FY2025: consolidated revenue $113.8B, +18% reported / +17% cc (transcript garbles the reported figure as "818%" — +18% is the consistent reading), an acceleration from Q3's +16%/+15% cc. FY2025 revenue $403B, +15% — first $400B+ year.
- Operating income $35.9B, +16%; margin 31.6% — both depressed by a $2.1B stock-based comp charge from Waymo's valuation step-up tied to its new $16B investment round (mostly in R&D). Ex-charge margin would be roughly 33.4% (inference: $2.1B on $113.8B ≈ 1.8pts).
- Net income $34.5B, +30%; EPS $2.82, +31% — again aided below the line by $3.2B other income, primarily unrealized gains on nonmarketable equity securities (smaller than Q3's $12.8B).
- 2026 CapEx guided to $175–185B — roughly double FY2025's $91.4B and far above the "significant increase" pre-flag. This is the quarter's defining disclosure.
- Google Cloud $17.7B, +48% (from +34% in Q3) — a massive acceleration; operating income $5.3B, more than doubled; margin 30.1% (from 23.7% in Q3, 17.5% a year ago). Backlog $240B, +55% QoQ / more than doubled YoY — an $85B sequential add.
- Search $63.1B, +17% (from +15%) — second straight acceleration. YouTube ads $11.4B, +9% (decelerating from +15%, election comps as flagged). Subscriptions/Platforms/Devices $13.6B, +17% (from +21%). Network $7.8B, −2%.
- FCF $24.6B in Q4; FY2025 FCF $73.3B (vs. $91.4B CapEx). Record Q4 operating cash flow $52.4B. Cash $120.8B; long-term debt $46.5B.
- Headline metrics: Gemini app 750M+ MAU (from 650M); Gemini Enterprise 8M+ paid seats across 2,800+ companies (from 2M/700 four months ago); 325M+ paid subscriptions (from 300M); Waymo 400K+ rides/week and 20M+ cumulative trips — the rides metric finally restored after two quarters of silence; Apple partnership: preferred cloud provider, next-gen Apple Foundation models built on Gemini.
What management is focused on- Pichai's framing: "tremendous quarter," Gemini 3 launch as "a major milestone," AI investments "drive revenue and growth across the board." The full-stack thesis now has a flagship external validation: Apple as preferred cloud provider with Apple Foundation models based on Gemini — mentioned by both Pichai and Schindler.
- Efficiency narrative escalated to defend the CapEx doubling: Gemini serving unit costs cut 78% over 2025 via model optimizations and utilization; ~50% of code written by coding agents (reviewed by engineers); agents deployed in treasury/finance operations.
- Supply constraint is now the stated binding constraint: Pichai expects to go through 2026 "in a supply constrained way"; his "what keeps you up at night" answer was capacity — power, land, supply chain. The Intersect acquisition (data center and energy infrastructure) is framed as addressing this.
- Search expansion evidence deepened: Q4 saw "more usage than ever before"; US daily AI Mode queries per user doubled since launch; AI Mode queries 3x longer than traditional; ~1 in 6 AI Mode queries non-text; 250+ product launches in AI Mode/AIO in Q4 alone; Gemini 3 integrated into AI Mode and (in January) AIO.
- Cloud demand framing: new-customer velocity doubled vs. Q1; 2025's $1B+ deals surpassed the prior three years combined (Q3 framing: prior two years); existing customers outpacing initial commitments by 30%+; ~75% of cloud customers use AI products; AI customers use 1.8x more products.
- Ashkenazi's discipline messaging continues: depreciation $21.1B in 2025, +38% / +$5.8B (transcript garbles the dollar increment), with 2026 growth "accelerating in Q1 and meaningfully increasing for the full year"; efficiency framed as continuous, not episodic.
Key numbers and quarter mechanics- Revenue $113.8B, +18%/+17% cc. Cost of revenue $45.8B, +13% (TAC $16.6B, +12%; other cost of revenue $29.2B, +13% — depreciation, YouTube content acquisition, infrastructure ops).
- OpEx $32.1B, +29% — R&D +42% (Waymo SBC charge + AI talent comp + depreciation); S&M +12% (Gemini app and Search marketing); G&A +21% (charitable-contribution timing shift — a mechanical, non-recurring driver).
- Operating margin 31.6% including the $2.1B Waymo charge. Services margin 41.9% (operating income $40.1B, +22%) — back above 40% after Q3's fine-driven break.
- Cloud margin 30.1% — a 6.4pt sequential step, achieved while depreciation accelerates; the segment is now the margin-expansion engine.
- CapEx $27.9B in Q4; $91.4B FY2025, "in line with our expectation" (the $91–93B guide held — first guide to survive the year). Mix ~60% servers / 40% data centers & networking, similar split guided for 2026. Just over half of 2026 ML compute allocated to Cloud — first disclosure of internal/external compute split.
- FCF $24.6B in Q4; FY2025 $73.3B vs. $91.4B CapEx — full-year FCF now below CapEx, a crossover worth noting (Q4 FCF alone was below Q4 CapEx: $24.6B vs. $27.9B).
- Capital return: only $5.5B buybacks + $2.5B dividends in Q4 — buybacks roughly halved from Q3's $11.5B, coinciding with the CapEx ramp. Cash $120.8B; long-term debt $46.5B (debt disclosure is new prominence).
- Other Bets: revenue $370M; operating loss $3.6B — inflated by the $2.1B Waymo charge; underlying loss roughly $1.5B (inference), still wider than Q3's $1.4B. Alphabet funded "a significant portion" of Waymo's $16B round.
- Q1 2026 pre-flags: FX tailwind at spot rates (volatility caveat); normal ad seasonality; cloud growth strong despite tight supply; depreciation growth accelerating in Q1 and "meaningfully" for FY2026; continued hiring in AI and cloud.
Product and launch scorecard- Gemini 3: launched December; "fastest adoption of any model in our history"; Gemini 3 Pro processing 3x the daily tokens of 2.5 Pro since launch; integrated into AI Mode and (January) AI Overviews; credited with a sharp increase in Gemini app engagement per user. This is the strongest model-launch evidence management has offered.
- Gemini app: 750M+ MAU (from 650M); Pichai claims distinct progress in active usage, intensity, and retention across platforms and geographies — but no DAU, retention, or engagement numbers given despite Sandler's direct question.
- API tokens: 10B+ tokens/minute via direct customer API, up from 7B last quarter (+43% QoQ).
- Gemini Enterprise: 8M+ paid seats, 2,800+ companies (from 2M seats/700 companies in ~4 months — roughly 4x seats and companies); managed 5B+ customer interactions in Q4, +65% YoY. 120K+ enterprises use Gemini; 95% of top 20 and 80%+ of top 100 SaaS companies.
- Cloud AI: GenAI-model product revenue +~400% YoY, accelerating from +200% in Q3; ~350 customers each processed 100B+ tokens in December alone; partner-built AI solutions revenue +~300% YoY; top-15 software partner commitments +16x YoY; 14 product lines at $1B+ annual revenue (from 13).
- AI Mode monetization: still "early stages experimenting" — ads below the AI response in test; Direct Offers pilot (sponsored offers matched to ready-to-buy users in AI Mode); agentic checkout "soon" from select merchants. No monetization-rate data.
- AI Max: "hundreds of advertisers" per transcript (likely garbled — Q3 said hundreds of thousands; verify); ~70M creative assets generated via Gemini text customization in AI Max/PMax in Q4; Aritzia 80% incremental conversion-value uplift; L'Oréal 800 campaigns/23 countries, NYX DTC revenue +23%.
- Agentic commerce: Universal Commerce Protocol launched at NRF with retail partners; framed as the interoperability layer for agentic shopping; consumer-facing integration into Gemini/AI Mode promised for 2026.
- AntiGravity (agentic dev platform): 1.5M+ weekly users ~2 months post-launch — new product with real early traction.
- YouTube: FY2025 revenue $60B+ across ads and subscriptions; Shorts 200B+ daily views; US Shorts monetization-per-watch-hour above in-stream (language unchanged from Q3's upgrade); podcasts 700M+ living-room hours in October, +75% YoY; NFL Sunday Ticket highest paid subscribers ever; new YouTube TV genre-specific plans (10+) and a sports tier coming; Premium Lite "popular."
- Waymo: 400K+ rides/week, 20M+ cumulative fully autonomous trips (metric restored); Miami launched (6th market); $16B raise at a higher valuation (triggering the $2.1B SBC charge); airport and freeway public service progressing; US multi-city plus UK and Japan expansion "soon."
- Apple deal: preferred cloud provider + Apple Foundation models on Gemini — no economics, timing, or revenue treatment disclosed.
- Pixel 10a confirmed "very soon"; Project Genie (Genie 3 world model) announced; Chrome Autobrowse; personal intelligence in AI Mode/Gemini app.
Sell-side read-through- Eight analysts; tone constructive. No antitrust/DOJ question for a third consecutive quarter, and no question on the Apple deal's economics despite it being the quarter's biggest strategic announcement.
- Nowak (Morgan Stanley): agentic commerce progress — Pichai: 2025 was foundation-laying (UCP protocol), 2026 is when consumers actually use it. Genie/YouTube integration — Pichai: creators at the center, tools flowing into YouTube "naturally."
- Sheridan (Goldman): the key question — path to closing the compute gap and funding efficiencies. Pichai conceded supply-constrained through the year with lengthening supply-chain horizons. Ashkenazi gave the efficiency litany (data-center self-build, ~50% AI-written code, back-office agents) but no quantified savings target.
- Anmuth (JPMorgan): maintaining Gemini's frontier position — Pichai cited multi-paradigm progress and "relentless cadence," no specifics. TPUs sold outside Google Cloud as a separate revenue stream — Pichai effectively declined, reframing TPUs as part of Cloud's value proposition. A notable non-endorsement of the external-silicon thesis.
- Mahaney (Evercore): YouTube's 9% vs. retail strength — Schindler: election lapping on brand, plus the ads-to-subscriptions shift cannibalizes ad revenue "slightly negative... but positive for our business." "DeepSeek moment" for SaaS customers — Pichai argued Gemini is enabling, cited robust token-usage growth among SaaS customers; no direct rebuttal of seat/pricing erosion.
- Shmulik (Bernstein): investment framework question again — Ashkenazi: "highly rigorous framework," total-envelope then allocation, cash-flow and balance-sheet health considered — still no hurdle rate, ROIC target, or FCF floor. Third consecutive deflection (Post Q2, Nathanson Q3, Shmulik Q4). "What keeps you up at night" — Pichai: capacity constraints.
- Nathanson (MoffettNathanson): UCP rationale — Pichai: user + merchant value prop must both work; seamless action on discovery. CapEx duration split — Ashkenazi: ~60% machines / 40% long-duration (buildings up to 40-year depreciation); >50% of 2026 ML compute to Cloud — the most concrete new disclosure of the call.
- Sandler (Barclays): Gemini app usage/retention and whether 750M MAU is the right competitive metric — Pichai claimed broad engagement/retention progress, gave no numbers, and redirected to AI Mode as a second AI-native surface.
- Gawrelski (Wells Fargo): monetizing conversational AI search where utility is on-platform — Schindler gave vertical color (retail, finance, health led; nearly every major vertical accelerated) and repeated query-expansion claims; no answer on revenue-per-session mechanics or premium-subscription role. Apple partnership alignment (revenue share vs. new model) — went unanswered; the second half of his question was dropped.
- Post (BofA): Gemini app cannibalization of Search — Pichai: "we haven't seen any evidence of cannibalization," expansionary framing. Gemini app monetization — Schindler: free tier + subscriptions today; ads "have always been part of scaling products... we're not rushing anything" — the clearest signal yet that Gemini app ads are coming, on no stated timeline.
- Notable absences: no question on the Waymo charge or its recurrence risk, the halved buyback, the Apple deal terms, Network decline, or FCF turning negative vs. CapEx in 2026.
Management credibility- Delivered the promised 2026 CapEx number ($175–185B) on this call as committed in Q3 — follows through on the disclosure commitment, though the magnitude (~2x) redefines the debate.
- FY2025 CapEx landed at $91.4B, "in line" with the $91–93B guide — the first guide in this cycle to survive a full year, partially rehabilitating guidance credibility after two intra-year raises.
- Waymo rides/week restored (400K+) after two quarters of silence — responsive to a standing watchlist item, and now backed by 20M cumulative trips and a $16B external raise at a higher valuation.
- The ROIC/framework question has now been deflected three straight quarters (Post, Nathanson, Shmulik). Ashkenazi's answer this time was the most detailed version of the same process language — envelope, allocation, balance-sheet health — with zero quantified return metrics. This is now a settled pattern: the framework exists internally but will not be shared in measurable form.
- New efficiency evidence is specific and testable: 78% Gemini serving-cost reduction in 2025, ~50% AI-written code, 60/40 CapEx split, >50% ML compute to Cloud — a real upgrade in capital-allocation transparency even without ROIC numbers.
- The $2.1B Waymo SBC charge was disclosed cleanly with segment placement — but it is a non-cash charge created by a related-party valuation step-up in a round Alphabet itself largely funded; management presented it as a simple negative without discussing the circularity, and no analyst probed it.
- Search acceleration claims are now backed by disclosed vertical breadth (retail, finance, health; nearly every major vertical accelerated) — more verifiable than prior quarters' attribution to AI surfaces alone.
- Gemini app engagement claims ("sharp increase," "distinct progress" in retention) were asserted under direct questioning without a single supporting number — the same pattern as the AI Mode metric-basis switch last quarter.
- Gawrelski's Apple-partnership economics question was simply not answered — an evasion on the quarter's marquee deal.
What changed versus the prior quarter- 2026 CapEx: $175–185B vs. "significant increase" language — roughly double FY2025's $91.4B; the debate shifts from guide credibility to FCF consequences (FY2025 FCF $73.3B already below CapEx $91.4B).
- Cloud re-accelerated dramatically: +48% from +34%; margin 30.1% from 23.7%; backlog $240B (+$85B QoQ) from $155B (+$49B QoQ) — backlog adds accelerating.
- Search accelerated again: +17% from +15%. YouTube ads decelerated to +9% from +15% (election comps, as pre-flagged). Subscriptions decelerated to +17% from +21%.
- Legal-charge pattern broken but replaced: no new fine this quarter; instead a $2.1B Waymo valuation SBC charge — a different species of one-off hitting R&D and Other Bets.
- Gemini 3 shipped (Q3: "later this year") with quantified adoption (3x 2.5 Pro daily tokens) — the promised catalyst landed and is credited for engagement step-ups across surfaces.
- Consumer AI: Gemini app 650M → 750M MAU; subscriptions 300M → 325M; Gemini Enterprise 2M/700 → 8M/2,800 in four months.
- Apple partnership — the largest new strategic fact: preferred cloud provider plus Apple Foundation models on Gemini; economics undisclosed.
- Waymo disclosure restored (400K rides/week) and capital structure changed ($16B external round, Alphabet majority funding).
- Buybacks halved sequentially ($11.5B → $5.5B) as CapEx ramps — capital-return priority visibly subordinated to the build-out.
- Depreciation: FY2025 +38% to $21.1B; 2026 guided to accelerate in Q1 and "meaningfully increase" — harsher language than Q3's "slight acceleration."
- Serving-cost disclosure: 78% Gemini unit-cost reduction in 2025 — new counterweight to the depreciation narrative.
Bull case- Demand is outrunning even a doubled CapEx plan: Cloud +48% with a $240B backlog that added $85B in one quarter, $1B+ deals in 2025 exceeding three prior years combined, existing customers beating commitments by 30%+, and management still expecting to be supply-constrained through 2026.
- Cloud margin hit 30.1% while depreciation accelerates — the segment is demonstrating operating leverage at scale, with GenAI-product revenue growth accelerating to ~400% YoY.
- Search is accelerating (+17%) in the AI era, with breadth evidence (nearly every major vertical accelerated) and monetization surface expansion (AI Max unlocking net-new queries, Direct Offers, agentic checkout coming).
- Gemini 3 is a validated model-leadership event: fastest adoption in company history, 3x predecessor token volume, driving measurable engagement gains across the app and AI Mode.
- The Apple deal is landmark third-party validation of both Gemini models and Google Cloud — potentially a large, durable revenue stream (economics TBD).
- Gemini Enterprise is scaling like a real enterprise product: 4x seats and 4x companies in four months, 5B quarterly interactions.
- Efficiency is quantified for the first time: 78% serving-cost reduction, ~50% AI-written code — evidence the CapEx dollar goes further than bears assume.
- Waymo now has external validation ($16B round), restored usage disclosure (400K rides/week), and multi-continent expansion — Other Bets' anchor asset is de-risking.
Bear case- $175–185B 2026 CapEx against $73.3B FY2025 FCF implies deeply negative FCF in 2026 absent major operating-cash-flow growth — and management offered no FCF floor, no ROIC hurdle, and no ceiling, for the third straight quarter of deflection. Buybacks were already halved; long-term debt ($46.5B) is now part of the picture.
- Depreciation growth guided to "meaningfully increase" in 2026 off a $21.1B base — the P&L headwind compounds precisely as the CapEx doubles; Services margin's 41.9% will be tested.
- YouTube ads decelerated to +9% and management's explanation (election comps, subscription shift) doesn't fully reconcile with claimed retail strength — Mahaney's question on this was only partially answered.
- The highest-growth surfaces remain the least monetized: AI Mode ads still "early experiments," Gemini app monetization explicitly deferred ("not rushing"), and no AIO/AI Mode revenue-per-query data has ever been provided.
- The $2.1B Waymo charge is a related-party valuation artifact — Alphabet funded most of the round that created the charge — and Other Bets' underlying loss (~$1.5B ex-charge, inference) is still widening.
- Quality of earnings: +30% net income again leans on unrealized equity gains ($3.2B other income); G&A's +21% was timing-driven; ex-Waymo-charge adjustments are needed to see the real margin.
- Pichai declined to endorse external TPU sales as a revenue stream — capping one bull thesis on silicon monetization.
- Apple deal economics were evaded under direct questioning — revenue share, TAC implications, and margin treatment are unknown, and the deal could carry its own capacity costs against the constrained supply base.
- Subscriptions growth decelerated (+21% → +17%) even as AI plans supposedly drive demand.
Next-quarter watchlist- CapEx/FCF mechanics: Q1 2026 CapEx vs. the $175–185B ramp ("ramping over the course of the year"); whether operating cash flow growth keeps FCF positive; any further buyback reduction or debt issuance; whether any ROIC/hurdle disclosure ever arrives (now a three-quarter deflection streak).
- Cloud: whether +48% holds or was pull-forward; backlog off $240B; margin vs. 30.1% as depreciation accelerates; Apple deal revenue treatment, timing, and capacity cost — the biggest open disclosure item.
- Depreciation: guided Q1 acceleration and "meaningful" FY2026 increase — track the dollar increment against the +$5.8B FY2025 step and its split between Services and Cloud margins.
- Search: whether +17% holds without election comps; first results from Direct Offers and agentic checkout; any hard AI Mode monetization data; paid clicks/CPC in the 10-Q.
- Gemini app: whether MAU growth continues off 750M post-Gemini 3; any monetization announcement (Schindler's "not rushing" implies a plan exists); engagement metrics to back retention claims.
- Gemini Enterprise: seat growth off 8M/2,800; any revenue-per-seat or ARR disclosure.
- Waymo: rides/week trajectory off 400K; UK/Japan launch milestones; deployment of the $16B round; whether valuation-driven SBC charges recur with future rounds; Other Bets loss ex-charge.
- YouTube: ad growth re-acceleration post-election comps; new YouTube TV plan pricing impact; Shorts monetization language; subscriptions growth stabilization.
- Legal/regulatory: still unasked for three quarters — DOJ remedies and any new EC action remain live risks with no disclosure.
- Intersect acquisition: closing timeline, cost, and how energy-infrastructure ownership changes the CapEx mix and depreciation profile.
|
| Oct 29, 2025 | +2.52% | Q3 FY2025 | Read transcript briefingQuarter in one view- Alphabet Q3 FY2025: first-ever $100B+ quarter — consolidated revenue $102.3B, +16% reported / +15% cc, an acceleration from Q2's +14%/+13% cc. Operating income $31.2B, +9%; operating margin 30.5% — but both include a $3.5B European Commission fine charge. Ex-fine: operating income +22%, margin 33.9%.
- Net income $35B, +33%; EPS $2.87, +35% — flattered by $12.8B of other income, primarily unrealized gains on nonmarketable equity securities. The gap between +9% GAAP operating income and +33% net income is largely below-the-line.
- FCF rebounded to $24.5B (TTM $73.6B) from Q2's $5.3B trough — aided by tax-law changes on R&D expensing/depreciation timing, partially offset by higher CapEx. Cash $98.5B.
- CapEx raised again: FY2025 now $91–93B, up from $85B (which was up from $75B one quarter earlier); Q3 CapEx $24B. 2026 guided to a "significant increase," details promised on the Q4 call.
- Google Cloud $15.2B, +34% (accelerating from +32%); operating income $3.6B, +85%; margin 23.7% vs. 20.7% in Q2 and 17.1% a year ago. Backlog $155B, +46% sequentially / +82% YoY — a $49B sequential add.
- Services $87.1B, +14%; Search $56.6B, +15% (accelerating from +12%); YouTube ads $10.3B, +15% (from +13%); Network $7.4B, −3%; Subscriptions/Platforms/Devices $12.9B, +21%. Services margin 38.5%, down YoY on the EC fine (fully absorbed in this segment).
- Headline disclosures: Gemini app 650M MAU (from 450M), queries up 3x from Q2; AI Mode 75M+ daily active users with queries doubling QoQ; 300M+ paid subscriptions; tokens now "over 1.3 [quadrillion] monthly" per Pichai (transcript garbles the unit — "1.3 quarterly and monthly tokens" — treat as ~1.3 quadrillion, up from 980T in July, but verify against the release); Anthropic planning access to up to 1M TPUs.
What management is focused on- Pichai's framing: "terrific quarter," "AI now driving real business results," first $100B quarter with revenue doubled in five years. The full-stack thesis is now argued with customer commitments (Anthropic's 1M TPUs, 9 of top 10 AI labs on Google Cloud) rather than just usage metrics.
- Search expansion narrative strengthened: query growth accelerated in Q3 (vs. "continuing to grow" in Q2), attributed to AIO and AI Mode; effect "more pronounced with younger people." AI Mode now "driving incremental total query growth" — a stronger claim than Q2.
- Cloud demand evidence escalated: new GCP customers +34% YoY (from +28% QoQ); $1B+ deals in 9M 2025 exceed the prior two years combined (Q2 framing: H1 equal to all of 2024); 70%+ of existing cloud customers use AI products; GenAI-model product revenue +200% YoY; ~150 customers each processed ~1T tokens over 12 months.
- Gemini Enterprise launched as "the new front door for AI in the workplace" — 2M subscribers across 700 companies already (note: "subscribers" here appears to be seats, distinct from consumer subscriptions).
- Ashkenazi's discipline continues: every margin statement paired with depreciation offset; depreciation +41% YoY to $5.6B with growth "expected to accelerate slightly in Q4"; S&M to be weighted to year-end (launches, holiday).
- New internal-efficiency data point: "nearly half of all code generated by AI" — the first refresh of the ">25% of new code" metric from several quarters ago.
- Schindler's monetization message: AI Max in Search is the "fastest-growing AI-powered search ads product," used by hundreds of thousands of advertisers, unlocking "billions of net new queries" in Q3 alone; paid clicks +7% and CPCs +7% disclosed preemptively (10-Q).
Key numbers and quarter mechanics- Revenue $102.3B, +16%/+15% cc. Cost of revenue $41.4B, +13% (TAC $14.9B, +8%; other cost of revenue $26.5B, +16% — YouTube content acquisition, then depreciation and infrastructure ops).
- OpEx $29.7B, +28% — R&D +22% (compensation, AI depreciation); S&M flat; G&A up "meaningfully" on the $3.5B EC fine. This is the second consecutive quarter with a large legal charge ($1.4B settlement in Q2).
- Operating margin 30.5% reported / 33.9% ex-fine. Services margin 38.5% (fine fully in-segment) — breaking the ~40% streak held for four quarters, though ex-fine it would be roughly 42.5% (inference: $3.5B on $87.1B revenue ≈ 4pts).
- FCF $24.5B vs. $5.3B in Q2 — mechanics: strong operating cash flow plus tax changes on R&D expensing and asset depreciation timing, partially offset by CapEx. TTM FCF $73.6B (from $66.7B).
- CapEx $24B; mix back to ~60% servers / 40% data centers & networking (from 2/3–1/3 in Q2). FY2025 guide $91–93B; 2026 "significant increase," details on Q4 call.
- Depreciation $5.6B, +$1.6B YoY, +41% (Q2 was +35%; the guided acceleration happened); slight further acceleration guided for Q4.
- Cloud backlog $155B (+$49B QoQ, +82% YoY) — the justification for the second CapEx raise in two quarters.
- Capital return: $11.5B buybacks + $2.5B dividends; cash $98.5B.
- Q4 pre-flags: FX tailwind at spot rates (with volatility caveat); US election comps hit YouTube ads in Q4; tight cloud supply-demand through Q4 and 2026; S&M weighted to year-end; depreciation growth accelerating slightly.
- Other Bets: revenue $344M, operating loss $1.4B (widened from $1.2B).
Product and launch scorecard- AI Overviews: 2B+ users (unchanged from Q2); monetization "approximately the same rate" repeated again — now the third consecutive quarter of the identical claim with no CTR/revenue-per-query data. Ads in AIO expanding to more countries and surfaces.
- AI Mode: 75M+ daily active users (new metric basis — Q2 gave 100M+ MAU; DAU vs. MAU not reconciled); queries doubled QoQ; rolled out globally in 40 languages; 100+ improvements shipped; ads now being tested in AI Mode (first monetization step); shopping capabilities and agentic checkout added.
- Gemini app: 650M MAU (from 450M), queries +3x QoQ — the strongest consumer AI adoption print yet. Gemini 3 confirmed for release "later this year."
- Tokens: ~1.3 quadrillion/month claimed (transcript wording garbled; verify), from 980T in July; 7B tokens/minute via direct customer API; 13M+ developers (from 9M).
- Veo 3: 230M+ videos generated (from 70M in Q2).
- Subscriptions: crossed 300M paid subscriptions (Google One and YouTube Premium lead) — the first absolute subscriber count management has given, answering a long-standing disclosure gap. Schindler added that Music/Premium subscribers generate "meaningfully higher gross profit" than ad-supported users.
- Shorts: US now earns more revenue per watch hour than in-stream (Q2: "parity... in some countries exceeds") — a small upgrade in language; still no absolute revenue.
- Cloud products: Gemini Enterprise at 2M subscribers/700 companies within weeks of launch; A4X Max instances (NVIDIA GB300) shipping; Ironwood (7th-gen TPU) GA "soon"; Anthropic up to 1M TPUs — the largest external TPU commitment disclosed.
- YouTube: first exclusive global NFL broadcast (Brazil) drew 19M+ viewers, record concurrent livestream, ad inventory sold out in weeks; interactive DR ads on CTV now >$1B annual run rate.
- Waymo: London targeted for 2026, Tokyo in progress; Dallas, Nashville, Denver, Seattle announced; fully autonomous airport operations permitted at San Jose and SFO; NYC testing scaling; Waymo for Business launched. Rides/week metric still not updated — second consecutive quarter of silence.
- Quantum: Willow chip ran an algorithm 13,000x faster than a top supercomputer, "verifiable"; Devoret Nobel noted. No commercial framing.
Sell-side read-through- Eight analysts, tone constructive; again no antitrust/DOJ question despite the $3.5B EC fine being the quarter's biggest margin event — the fine itself went unquestioned.
- Nowak (Morgan Stanley): agentic e-commerce monetization risk — Schindler: "additive," seamless UX plus partner-ecosystem integration, PayPal partnership and agent-to-agent protocols cited; no monetization-rate answer. Waymo-Gemini integration — Pichai: in-car experience opportunity, "newer experiences in 2026," personally reviewing with the team in weeks.
- Anmuth (JPMorgan): clicks/conversions/pricing in AI formats — Schindler gave vertical color, repeated AIO parity, then disclosed paid clicks +7% and CPCs +7% while repeating "we don't manage to paid clicks and CPC targets." Cost base — Ashkenazi: headcount moderation, real estate, data-center build efficiency, AI code generation (~half of code), Shorts' lower revenue share helping gross margin.
- Sheridan (Goldman): custom silicon monetization — Pichai: full-stack differentiation supports cloud margin growth and infrastructure as a growth driver. YouTube ads+subscriptions — Schindler's "twin engine" framing; Premium subs worth more than ad-supported users on gross profit.
- Shmulik (Bernstein): engagement depth across AI surfaces — Pichai: AIO universal/high engagement; AI Mode has a passionate early-adopter core; high satisfaction claimed, no numbers. Search economics vs. serving cost — Schindler repeated parity, noted AI Mode ads still in test, and floated monetizing historically non-commercial queries.
- Nathanson (MoffettNathanson): AI Mode query length and ROAS — Schindler: "too early to tell," declined detail. ROIC question put directly to Ashkenazi — she cited cloud AI revenue ("billions in the quarter") and a "rigorous framework" but offered no hurdle rate, no framework specifics, no quantified returns — the same gap as when Post asked in Q2.
- Sandler (Barclays): can AI Max raise the ~20% commercial-query share — Schindler separated query expansion from advertiser targeting, suggested adjacent-commercial monetization potential; Pichai: commercial applicability follows breadth "over time."
- Gawrelski (Wells Fargo): will engagement growth translate to marketing spend — Pichai: "expansionary moment," commercial journeys follow. Surface convergence — Pichai repeated the Universal Search analogy; two-surface (Search + Gemini) strategy persists with "thoughtful" simplification over time.
- Post (BofA): frontier-model pace — Pichai: pace increasing but each improvement takes slightly longer against more capable baselines; Gemini 3 "later this year." Cloud deal economics — Pichai pointed to full-stack differentiation, no margin or pricing detail on the mega-deals.
- Notable absences: no question on the EC fine, antitrust remedies, Waymo unit economics, Network decline, or the second CapEx raise's FCF consequences.
Management credibility- Delivered on the subscriber-count withhold: 300M+ paid subscriptions is the first absolute number after years of refusal — continues the pattern of disclosing metrics once favorable.
- The AI-code metric was refreshed ("nearly half of all code") after going unupdated in Q2 — responsive to a standing watchlist item.
- CapEx guidance credibility is now the central issue: $75B → $85B → $91–93B in two quarters, with 2026 pre-flagged for a "significant increase." Each raise is demand-justified (backlog +$49B QoQ is hard evidence), but no guide has survived a full year, and the Q4 call now carries a promised 2026 number.
- AIO monetization parity repeated verbatim for a third quarter under direct questioning (Anmuth, Shmulik) — consistent but still data-free. The preemptive paid-clicks/CPC disclosure (+7%/+7%) is new transparency, partially offsetting the opacity critique.
- Ashkenazi's ROIC answer under Nathanson's direct question was process language without metrics — second consecutive quarter an ROIC question yielded no framework. This is now a pattern, not an omission.
- The $3.5B EC fine was disclosed cleanly with ex-fine margins provided — good mechanics — but management gave no context on the fine's origin or whether more is coming, and no analyst asked.
- AI Mode metric basis shifted from MAU (100M+, Q2) to DAU (75M+, Q3) without reconciliation — the numbers are not comparable and management did not bridge them.
- Waymo rides/week unupdated for a second straight quarter despite heavy expansion news — the omission is now conspicuous.
- Pichai's frontier-model candor (improvements take longer against more capable baselines) is a testable, credibility-positive framing ahead of Gemini 3.
What changed versus the prior quarter- Growth re-accelerated across the board: revenue +16% (from +14%), Search +15% (from +12%), YouTube ads +15% (from +13%), Cloud +34% (from +32%). The H2 deceleration setup from Q2's pre-flags did not materialize in Q3.
- CapEx raised a second time: $85B → $91–93B; 2026 language hardened from "further increase" to "significant increase" with a committed Q4 disclosure.
- FCF recovered sharply: $24.5B vs. $5.3B — but partly on tax-law timing changes, not purely operational normalization.
- Cloud margin stepped again: 23.7% vs. 20.7%; backlog growth accelerated to +46% QoQ from +18%.
- Legal charges escalated: $3.5B EC fine vs. $1.4B settlement — two straight quarters of multi-billion-dollar charges, now hitting Services margin visibly (38.5% vs. 40.1%).
- Consumer AI metrics stepped up: Gemini app 450M → 650M MAU with 3x query growth; AI Mode switched to a DAU metric (75M) and began ad testing; subscriptions crossed 300M with an absolute count for the first time.
- Shorts language upgraded from "parity" to "earn more revenue per watch hour than traditional in-stream" in the US.
- TPU external validation escalated: Anthropic's up-to-1M-TPU commitment is the largest named external TPU deal; NVIDIA GB300 instances shipping — the "we offer both" positioning is now backed by named volume.
- Depreciation growth accelerated as guided: +41% vs. +35%, with further slight acceleration guided.
- Other Bets loss widened to $1.4B from $1.2B; Waymo's disclosure shifted further toward geography/business-model news and away from usage metrics.
Bull case- Acceleration with proof of demand: revenue +16%, Search +15%, Cloud +34%, and a $155B backlog that grew $49B in one quarter — the CapEx raises are being pulled by signed commitments ($1B+ deals in 9M25 exceeding two prior years combined), not pushed on speculation.
- Cloud is compounding margin with growth: 23.7% margin, +85% operating income growth, GenAI-product revenue +200% YoY, 13 product lines at $1B+ run rates — the AI spend is visibly converting to profit.
- The AI-cannibalization thesis took another hit: query growth accelerated, commercial queries growing, paid clicks +7% AND CPCs +7% simultaneously, and AI Mode is now driving incremental total query growth with ad testing underway.
- Consumer AI is a real franchise: 650M Gemini MAU with 3x query growth in a quarter, 300M+ paid subscriptions, and AI Pro/Ultra plans feeding the +21% subscriptions segment.
- Anthropic's 1M-TPU commitment plus 9-of-10 top AI labs on Google Cloud is third-party validation of the TPU stack that competitors cannot easily replicate.
- FCF recovered to $24.5B even at $24B CapEx — the business still self-funds the build-out at current levels.
- Gemini 3 landing "later this year" with management explicitly confident sets up a potential model-leadership catalyst into Q4.
Bear case- The CapEx guide has now been raised twice in two quarters ($75B → $85B → $91–93B) with a "significant increase" promised for 2026 and still no ROIC framework, hurdle rate, or ceiling — Ashkenazi's second non-answer on ROIC confirms the absence is deliberate or the framework doesn't exist in shareable form.
- Reported operating income grew only +9% and margin fell to 30.5%; the +33% net income print leaned on $12.8B of unrealized equity gains — quality of earnings this quarter is weaker than the EPS headline.
- Legal charges are becoming recurring: $1.4B then $3.5B in consecutive quarters, with no explanation of the EC fine's scope or whether the regulatory charge run-rate is now structural. Services margin broke its 40% streak.
- AIO monetization parity remains an unverified assertion three quarters running; AI Mode ad monetization is only now entering testing — the highest-growth surface is the least monetized, and management declined every request for economics detail ("too early to tell").
- Depreciation +41% and guided to accelerate further; supply remains tight through 2026 — cloud growth is still capacity-gated, and the mega-deals' pricing/margin terms went unaddressed when Post asked directly.
- FCF's recovery relied partly on tax-timing changes; strip that out and the underlying FCF trajectory against a rising CapEx base is less comfortable than the $24.5B print suggests.
- Waymo's usage metric has now disappeared for two quarters while Other Bets losses widen to $1.4B — expansion announcements are substituting for operating evidence.
- AI Mode's metric switch (100M MAU → 75M DAU) without reconciliation leaves consumer-AI engagement trends unauditable quarter-over-quarter.
Next-quarter watchlist- 2026 CapEx number on the Q4 call — the single biggest pending disclosure; size, phasing, and any accompanying return framework. Whether management finally offers an ROIC/hurdle-rate articulation after two straight deflections.
- Cloud: growth vs. +34% with capacity still constrained; backlog off $155B; margin vs. 23.7% as depreciation accelerates; economics of the mega-deals (any pricing/margin disclosure); Ironwood GA timing and Anthropic TPU deployment progress.
- Gemini 3: launch timing ("later this year"), benchmark reception, and whether it drives another step-up in Gemini app MAU (650M) and API token volume.
- Search: whether +15% holds against Q4 election comps (YouTube explicitly flagged); AI Mode ad-test results and any expansion; first hard data behind the AIO parity claim; paid clicks/CPC trends in the 10-Q (+7%/+7% this quarter).
- Legal: details behind the $3.5B EC fine, appeal posture, and whether Q4 carries another charge; any DOJ remedy developments — the question remains unasked for two quarters but the risk is live.
- FCF mechanics: how much of Q3's recovery was tax-timing vs. operational; Q4 CapEx vs. the implied ~$43B remaining in the $91–93B guide (inference: $91–93B minus ~$70B spent through Q3 — verify actual YTD spend).
- Subscriptions: whether the 300M count gets segmented (YouTube vs. Google One) and whether +21% growth holds; Gemini Enterprise seat growth off 2M/700 companies.
- Waymo: whether rides/week returns; London/Tokyo 2026 execution milestones; Waymo for Business traction; Other Bets loss trajectory vs. $1.4B.
- Depreciation: guided "slight acceleration" in growth rate off +41% — the pace of this line against Services/Cloud margins is the key P&L tension for 2026.
|
| Jul 23, 2025 | +1.02% | Q2 FY2025 | Read transcript briefingQuarter in one view- Alphabet Q2 FY2025: consolidated revenue $96.4B, +14% reported / +13% cc — a slight deceleration from the prior quarter's +15%/+16% cc pace (per prior summary). Operating income $31.3B, +14%; operating margin 32.4%. Net income $28.2B, +19%; EPS $2.31, +22%.
- FCF collapsed to $5.3B in the quarter (TTM $66.7B) on a "sizable sequential increase in CapEx" ($22.4B, up from ~$13B/quarter in the prior period) plus Q2's dual federal tax payments. Cash and marketable securities: $95B.
- The headline: 2025 CapEx guidance raised to ~$85B from $75B, with a "further increase" flagged for 2026 — justified by cloud demand, a $106B backlog (+38% YoY, +18% sequentially), and a tight supply environment lasting "into 2026."
- Google Services $82.5B, +12%; operating margin 40.1% flat YoY despite a $1.4B legal settlement charge. Search $54.2B, +12% (retail and financial services/insurance lead); YouTube ads $9.8B, +13% (direct response first, then brand); Network $7.4B, −1%; Subscriptions/Platforms/Devices $11.2B, +20% (subscription-led, YouTube then Google One).
- Google Cloud $13.6B, +32%; operating income $2.8B; margin 20.7% vs. 11.3% a year ago; annual run rate >$50B.
- Major disclosure upgrades: Gemini app 450M MAU (the number refused two quarters ago), Shorts monetization parity in the US ("earn as much revenue per watch hour as traditional in-stream... in some countries it now exceeds"), AIO at 2B+ monthly users, AI Mode at 100M+ MAU (US/India), 980T monthly tokens processed (doubled since May).
- Other Bets: revenue $373M, operating loss $1.2B. Waymo: 100M+ autonomous miles on public roads, Atlanta launched, Austin territory doubled, testing in 10+ cities including New York and Philadelphia. No rides/week update this quarter.
What management is focused on- Pichai's framing: "standout quarter," AI "positively impacting every part of the business," full-stack thesis unchanged but now backed by usage metrics at every layer (tokens, developers, MAU, enterprise counts).
- Search expansion narrative: AIO drives >10% more queries globally for query types that show them, with growth "continuing to increase over time"; overall and commercial queries growing YoY; AI features cause users to search more, especially younger users. This is the core defense against the AI-cannibalization thesis.
- Cloud demand evidence stacked deliberately: deals >$250M doubling YoY; $1B+ deals signed in H1 2025 equal to all of 2024; new GCP customers +28% QoQ; 85,000+ enterprises building with Gemini (35x usage growth YoY); "nearly all GenAI unicorns" on Google Cloud; named TPU customers Safe Superintelligence and Physical Intelligence.
- Agentic AI as the next frontier, with unusual candor on limitations: agents are "a bit slow and costly... sometimes brittle"; latency and cost compound across chained events; 2026 named as the year agentic experiences go broad. Internal agentic coding workflows now rolling out to Google engineers.
- Ashkenazi's message discipline: every margin/efficiency statement paired with the depreciation offset; depreciation +35% YoY to $5B with growth "expected to accelerate further in Q3"; headcount additions increasing in Q3 (new graduates); Pixel launch expense flagged for Q3.
- Schindler quantified ad-product AI claims: AI Max in Search campaigns +14% conversions; Smart Bidding Exploration +19% conversions; 2M+ advertisers using AI asset generation (+50% YoY).
Key numbers and quarter mechanics- Revenue $96.4B, +14%/+13% cc. Cost of revenue $39B, +10% (TAC $14.7B, +10%; other cost of revenue $24.3B, +10% — YouTube content acquisition, then depreciation).
- OpEx $26.1B, +20% — the biggest driver was the $1.4B legal settlement charge; R&D +16% (compensation, depreciation); S&M +5%.
- Operating margin 32.4% absorbed both the legal charge and the depreciation ramp; Services margin 40.1% flat YoY — the 40% level now held for a fourth quarter (per prior-summary context).
- FCF $5.3B vs. $17.6B in the prior quarter's period — mechanics: CapEx nearly doubled sequentially and Q2 carries two quarters of federal tax payments. TTM FCF $66.7B (improved from the $55.8B trough noted in the prior summary).
- CapEx $22.4B; mix shifted to ~2/3 servers, 1/3 data centers/networking (from the previously disclosed 60/40). FY2025 guide ~$85B (from $75B); 2026 guided to a further increase, details deferred to a future call.
- Cloud backlog $106B, +38% YoY, +18% sequentially — the hard evidence behind the CapEx raise.
- Depreciation $5B, +$1.3B YoY, +35%; growth rate guided to accelerate in Q3.
- Capital return: $13.6B buybacks + $2.5B dividends; cash $95B (stabilized vs. the $93B low in the prior summary).
- H2 pre-flags: lapping 2024 financial-services strength; US election spend comps hit YouTube in H2; Q3 FX tailwind at spot rates; Q3 headcount additions up sequentially; Pixel launch expense in Q3; cloud supply tight into 2026 with capacity additions weighted to the back half of the year — and Ashkenazi explicitly warned cloud growth is not linear quarter-to-quarter given capacity-delivery timing.
Product and launch scorecard- AI Overviews: 2B+ monthly users, 200+ countries, 40 languages; now on Gemini 2.5 ("fastest AI responses in the industry"); drives >10% query growth on covered query types; monetization "approximately the same rate" reiterated by Schindler under direct CTR questioning from Sandler — still no CTR or revenue-per-query data behind the parity claim.
- AI Mode: launched US and India; 100M+ MAU already; Deep Search and personalization coming. First hard adoption number for the end-to-end AI search experience.
- Gemini app: 450M MAU, daily requests +50% vs. Q1 — the consumer usage number management refused to give two quarters ago, now volunteered. Veo 3: 70M+ videos generated since May; photo-to-video rolling into Google Photos.
- Tokens processed: 980T/month, doubled from 480T at I/O in May — the cleanest aggregate AI-usage metric disclosed.
- Subscriptions: Google One AI Pro/Ultra plans driving "accelerated traction" since 2.5 Pro launch; subscriptions/platforms/devices +20% with paid-subscription growth the biggest driver. Still no absolute subscriber counts.
- Shorts: 200B+ daily views (from 70B in the prior summary's period — note the metric basis may differ; treat the jump cautiously); US revenue-per-watch-hour parity with in-stream achieved, exceeding it in some countries — the three-quarter-old monetization question finally answered with a rate claim, though still no absolute revenue figure.
- Lens: +70% query growth YoY; majority incremental; Circle to Search on 300M+ Android devices (from 150M+).
- Cloud products: Agentspace with 1M+ subscriptions booked pre-GA; open-source Agent Development Kit 1M+ downloads in <4 months; Anywhere Cache (−70% inference latency); BBVA saving ~3 hours/employee/week, rolling to 100K employees.
- Waymo: 100M+ autonomous miles cumulative; Atlanta launched; Austin territory doubled; LA/SF Bay Area +50%; teen accounts in Phoenix; testing in 10+ cities incl. NYC and Philadelphia. No rides/week figure this quarter — the 150K/week metric from the prior period went unupdated.
- OpenAI confirmed as a Google Cloud customer (Sandler's question; Pichai: "super excited... look forward to investing more in that relationship") — a notable validation of the open-platform pitch.
Sell-side read-through- Nine questions; tone constructive, no antitrust question this quarter (a reversal from the prior period where DOJ remedies surfaced).
- Sheridan (Goldman): volume vs. monetization in changing consumer behavior — Pichai: "we will lead with the organic experience," monetization follows on new surfaces (Gemini app explicitly near-term unmonetized). YouTube ads-vs-subscriptions: Schindler's "double tier strategy," no mix targets.
- Anmuth (JPMorgan): pressed the compute-access paradox ($10B more CapEx yet still supply-constrained) — Pichai: both true, investment lag. Paid clicks: Schindler disclosed paid clicks +4% YoY (10-Q) and warned against drawing conclusions from click/CPC metrics — a deflection with a number attached.
- Nowak (Morgan Stanley): agentic technical hurdles — got the quarter's most candid answer (latency/cost compounding, brittleness, 2026 timeline). Internal AI efficiency: agentic coding now rolling out internally, but no update to the ">25% of new code AI-generated" metric from two quarters ago.
- Nathanson (MoffettNathanson): glasses/hardware (Pichai: phones central for 2–3 more years); AI Mode vs. Gemini app differentiation (information-grounded vs. assistant/companion use cases).
- Shmulik (Bernstein): AI talent war — Pichai claimed retention and inbound metrics "healthy" without numbers; Ashkenazi tied talent cost to the OpEx line with efficiency offsets.
- Sandler (Barclays): CTR durability under AI SERP formats — Schindler repeated the "approximately the same rate" parity claim, no new data. OpenAI cloud relationship — confirmed, expansion hinted.
- Mahaney (Evercore): H2 ad environment — Schindler declined ("only a few weeks into Q3"); two-surface search strategy — Pichai defended with the Universal Search analogy, implying eventual convergence.
- Gawrelski (Wells Fargo): cloud H2 outlook — Ashkenazi walked back linearity expectations, capacity weighted to back half, tight into 2026. Agentic web structure — Pichai: business models must be solved, not just technology.
- Post (BofA): Gemini consumer subscriptions ("accelerated traction" since 2.5 Pro, no numbers) and CapEx ROI — Pichai cited low churn, high satisfaction, margin trajectory as ROI evidence; no ROIC framework or hurdle rate offered.
- Notable absences: no antitrust/DOJ question, no Waymo unit-economics question, no Network question, no challenge on the FCF collapse.
Management credibility- Delivered on the two longest-standing withheld disclosures: Gemini app users (450M MAU, after refusing the ChatGPT-comparison question two quarters ago) and Shorts monetization rate (US parity, after three-plus quarters of "improving" without quantification). The pattern suggests metrics get disclosed when they turn favorable — favorable for trust in the numbers, cautionary for what remains undisclosed.
- CapEx credibility cuts both ways: the prior "increase but smaller step-up" signal was honored in direction, but the guide moved from $75B to $85B in one quarter — a 13% raise that validates demand claims while weakening confidence in any given guide's durability. The 2026 "further increase" pre-flag at least removes surprise.
- The cloud supply narrative shifted: prior commentary pointed to constraints easing toward end of 2025; now "tight demand supply environment going into 2026." Ashkenazi addressed the change head-on under Gawrelski's question but the goalpost moved.
- AIO monetization parity was repeated verbatim under a direct CTR question — consistent, but still an assertion without CTR, impression, or revenue-per-query data after multiple quarters of requests.
- The $1.4B legal charge was disclosed cleanly in prepared remarks with its margin impact explained — better than the unflagged $3B EC fine payment in the prior period.
- Pichai's agentic candor ("slow, costly, brittle") is credibility-positive: specific limitations plus a dated expectation (2026) create a testable marker.
- No reconciliation offered of the prior period's 40x/14x Gemini API discrepancy; the metric has been replaced by "35x growth in Gemini usage YoY" among enterprises — a cleaner but different basis.
- Waymo's rides/week metric (150K+ previously) was not updated despite expansion news — a quiet omission on a metric that had been stepping up hard.
What changed versus the prior quarter- CapEx regime change: from ~$13B/quarter to $22.4B, FY guide raised $75B→$85B, and 2026 pre-flagged for another increase. The "smaller percent step-up" framing from the prior period is dead.
- FCF compressed sharply: $5.3B vs. $17.6B, on CapEx plus tax timing; TTM FCF recovered to $66.7B from $55.8B.
- Cloud margin stepped again: 20.7% vs. 17% (and 11.3% a year ago) — the durability question from the prior period is being answered affirmatively, though depreciation headwinds are explicitly accelerating.
- Consumer AI disclosure opened up: Gemini app 450M MAU and +50% daily-request growth disclosed after prior refusal; AI Mode 100M MAU; tokens doubled to 980T.
- Shorts monetization: from unquantified "improving" to US parity with in-stream, exceeding it in some countries.
- Search growth held at +12% but the lead vertical flipped back to retail first, then financial services (insurance called out); YouTube ads ticked to +13% with direct response leading; Network −1%.
- Subscriptions/platforms/devices +20% — now subscription-led (YouTube, Google One AI plans) rather than hardware-pull-forward-led as in the prior period.
- Antitrust went silent: no analyst question, no management comment — versus the DOJ-remedy exchange in the prior period. The risk didn't disappear; the airtime did.
- New named relationships: OpenAI as a cloud customer; PayPal partnership (cloud + payments across Google products).
- New H2 headwind pre-flags: US election comps on YouTube, financial-services lap, Q3 headcount and Pixel expense, accelerating depreciation growth.
Bull case- Demand is outrunning supply with proof: $106B backlog (+38% YoY), $1B+ deals in H1 equal to all of 2024, >$250M deals doubling, new GCP customers +28% QoQ, and management raising CapEx $10B mid-year because customers are asking for it. Cloud margin at 20.7% shows the spend converts to profit, not just revenue.
- The AI-cannibalization bear case keeps losing ground: AIO drives >10% incremental queries on covered types, overall and commercial queries growing, Search +12% with all verticals contributing, and monetization parity reiterated.
- The two biggest disclosure withholds resolved positively: Gemini app at 450M MAU with +50% daily-request growth is a real consumer AI franchise; Shorts at US monetization parity removes a multi-year drag narrative.
- Token volume doubling in two months (480T→980T) plus 9M developers and 85K enterprises on Gemini shows platform-level adoption compounding across the stack.
- Services margin held 40.1% while absorbing a $1.4B legal charge and 35% depreciation growth — underlying efficiency gains are real.
- Subscription flywheel diversifying revenue: +20% segment growth led by paid subscriptions, with AI Pro/Ultra plans creating a consumer AI revenue line that didn't exist a year ago.
- OpenAI choosing Google Cloud is third-party validation of TPU/infrastructure competitiveness that no marketing claim could match.
Bear case- FCF of $5.3B against $16.1B of capital return means the quarter's buyback was effectively balance-sheet funded; CapEx is guided up again in 2026 with no ceiling, no ROIC framework, and depreciation growth explicitly accelerating — the FCF trajectory is now hostage to data-center delivery schedules.
- The CapEx guide lasted one quarter before a 13% raise; if demand signals soften or AI monetization disappoints, the depreciation base is locked in. Ashkenazi's own warning that cloud growth won't be linear sets up possible deceleration prints against a +32% comp.
- Monetization parity on AIO remains an unverified assertion — Sandler asked directly about CTR and got the same sentence as two quarters ago. Paid clicks +4% with management actively discouraging click-based analysis leaves the Search monetization mechanism opaque.
- H2 is pre-loaded: election comps on YouTube, financial-services lap, Pixel expense, headcount growth, accelerating depreciation — management has built a wall of excuses in advance, which invites the question of what they're seeing.
- The supply-constraint timeline slipped from "easing late 2025" to "tight into 2026" — either demand keeps surprising (bullish) or planning keeps missing (execution risk); either way, cloud growth is capacity-gated, not demand-gated.
- Waymo's ride metric went unupdated; Other Bets losses widened to $1.2B; the agentic future management is investing toward is self-described as "slow, costly, brittle" with a 2026 timeline.
- Antitrust silence this quarter is not resolution — the DOJ remedy overhang on Search distribution (Apple ISA, Android pre-installs) remains undisclosed and unaddressed.
Next-quarter watchlist- Cloud: growth vs. +32% given the non-linearity warning; whether back-half capacity delivery materializes; margin vs. 20.7% as depreciation accelerates; backlog trajectory off $106B; any auditable AI-revenue figure.
- CapEx/FCF: Q3 CapEx run-rate vs. the ~$85B FY guide; whether FCF recovers as tax timing normalizes; first 2026 CapEx details (promised on "a future earnings call"); depreciation growth rate vs. the guided acceleration.
- Search: whether +12% holds against financial-services laps; any hard AIO monetization data (CTR, revenue per query) behind the parity claim; AI Mode expansion beyond US/India and its monetization treatment; paid-click trends in the 10-Q.
- Gemini: MAU growth off 450M; whether AI Pro/Ultra subscription traction gets quantified; Deep Search and personalization feature launches; any monetization move on the Gemini app (management said organic-first "for the near term").
- YouTube: election-comp impact in H2; Shorts parity extending beyond the US; NFL Brazil game execution; CTV conversion momentum (1B+ conversions claimed).
- Subscriptions: whether +20% holds without a hardware pull-forward; any absolute subscriber counts (still never given).
- Waymo: whether the rides/week metric returns and at what level; NYC/Philadelphia testing progress; any unit-economics disclosure; Other Bets loss trajectory vs. the $1.2B print.
- Legal/regulatory: details behind the $1.4B settlement; any DOJ remedy developments — the question will return.
- Agentic timeline: evidence toward the 2026 "broad agentic experiences" claim; internal agentic-coding results; whether the >25%-AI-generated-code metric gets refreshed.
- Talent: whether the "healthy retention" claim survives the AI talent war without a visible compensation spike in R&D.
|
| Apr 24, 2025 | +1.68% | — | Transcript briefing unavailable |
| Feb 4, 2025 | -7.29% | — | Transcript briefing unavailable |
| Oct 29, 2024 | +2.82% | Q3 FY2024 | Read transcript briefingQuarter in one view- Alphabet Q3 FY2024: consolidated revenue +15% reported / +16% cc — a reacceleration from Q2's +14%/+15%. Operating income $28.5B, +34%; operating margin 32% (flat vs. Q2's 32% but up YoY). Net income $26.3B, +34%; EPS $2.12, +37%. FCF $17.6B in Q3 (up from Q2's $13.5B) but TTM FCF fell to $55.8B from $60.8B — the flagged tax-timing drag plus a new $3B EC shopping-fine cash payment.
- Cash and marketable securities fell again: $93B vs. $101B in Q2 — despite $17.6B FCF, the company returned $15.3B in buybacks + $2.5B dividends in the quarter (~$70B TTM).
- Google Services $76.5B, +13%: Search and Other $49.4B, +12% (decelerating from +14%; vertical order flipped to financial services/insurance first, then retail); YouTube ads $8.9B, +12% (brand first, then direct response — same ordering as Q2); Network $7.5B, −2% (improved from −5%); Subscriptions, Platforms and Devices $10.7B, +28% — a sharp reacceleration from +14%, driven by subscriptions plus the pulled-forward Made by Google launch. TAC $13.7B, +9%.
- Services operating income $30.9B, +29%; margin 40% for a third consecutive quarter — the level Porat declined to endorse last quarter held.
- Google Cloud: $11.4B, +35% (accelerating from +29%), operating income $1.9B, margin 17% (from 11%) — a six-point sequential margin jump, attributed to AI products, core GCP, Workspace, and efficiency.
- Other Bets: revenue $388M, operating loss $1.1B (flat vs. Q2). Waymo metrics stepped up hard: 150K+ paid rides/week (from 50K+), 1M+ fully autonomous miles/week; Uber partnership in Austin/Atlanta, new Hyundai partnership, 6th-gen system with "significantly reduced unit costs."
- CapEx $13B again; Q4 guided to "approximately the same level"; first 2025 commentary: an increase, but not the same percent step-up as '23→'24 — detail deferred to the Q4 call.
- Ashkenazi's debut as CFO: reaffirmed the cost-reengineering program, promised to "accelerate" and "pivot" where needed, and gave the first 2025 CapEx signal.
What management is focused on- Pichai framed the "full stack" AI thesis (infrastructure, research, global reach) as now "operating at scale," with new hard efficiency evidence: AIO machine cost per query down >90% in 18 months while doubling the custom Gemini model size — a major upgrade from Q2's "flat cost per serve" framing.
- Internal AI productivity quantified for the first time: more than a quarter of all new code at Google is AI-generated (engineer-reviewed) — a direct answer to Q2's "very early stages" posture.
- Organizational velocity as a theme: Gemini app team moved into DeepMind; "new synapses" restructuring metaphor; Notebook LM cited as proof small teams can ship; third-generation Gemini model "progressing well"; Project Astra experiences targeted "as early as 2025."
- Cloud demand framed through five usage patterns with quantified proof points (LG AI −50% inference time/−72% cost; Gemini API calls ~40x in 6 months; BigQuery ML operations +80% in 6 months; Mandiant threat-detection adoption 4x over 6 quarters; Snap My AI engagement 2.5x).
- Schindler: ads now live inside AIO for US mobile users (the Q2-announced test executed); shopping ads rolled out above/alongside Lens results; upfront commitments +20% YoY; Shorts monetization "improved again" — still no rate quantification.
- Ashkenazi's doctrine: build on the reengineering program, "evaluate where we might be able to accelerate work and where we might need to pivot to free up capital for more attractive opportunities" — continuity with an explicit willingness to reallocate.
Key numbers and quarter mechanics- Revenue +15%/+16% cc. Cost of revenue $36.5B, +10%; other cost of revenue $22.8B, +11% — content acquisition (YouTube), depreciation from technical infrastructure, and hardware costs from the Made by Google pull-forward (the pre-flagged Q3 headwind landed as advertised).
- OpEx $23.3B, +5%: driven by $607M facilities charges (office-footprint optimization, booked in Alphabet-level activities), then depreciation; offset by lower legal charges. R&D +11% (compensation, depreciation); S&M +5% (Made by Google, AI/Gemini marketing); G&A −10% (lower legal charges).
- FCF $17.6B; YoY comparison hit by the 2023 deferred-tax lap (as pre-flagged) and a $3B cash payment on the 2017 EC shopping fine — a new, unflagged drag item.
- TAC $13.7B, +9% — still below Search growth (+12%); mix shift toward Search and away from high-TAC Network remains margin-supportive.
- Cloud margin bridge: 11% → 17% in one quarter, on revenue acceleration plus efficiency; Ashkenazi flagged ongoing depreciation/infrastructure headwinds against it.
- CapEx mix disclosed for the first time: ~60% of technical-infrastructure spend to servers (TPUs + GPUs), ~40% to data centers/networking. Q3 data-center announcements >$7B, ~$6B in the US. Nuclear: first corporate agreement for SMR power, up to 500MW.
- Q4 pre-flags: ad revenue YoY growth still pressured by H2 2023 APAC-retailer strength; subscriptions/platforms/devices faces a Q4 headwind from the hardware pull-forward (mirror of Q3's benefit); depreciation/infrastructure expense keeps rising, partially offset by lower Q4 device cost of revenue.
- Capital return quantified: $15.3B buybacks + $2.5B dividends in Q3; ~$70B TTM.
Product and launch scorecard- AI Overviews: rolled out to 100+ new countries this week; >1B monthly users; ads inside AIO now live for US mobile (test executed on schedule); monetization "at approximately the same rate" as traditional search per Schindler — the first monetization-parity claim, still unaudited; cost per query down >90% over 18 months. Engagement claims remain qualitative ("one of the most positive user satisfaction launches," growth compounding over time).
- Circle to Search: 150M+ Android devices; a third of users who tried it now use it weekly — a real retention metric; higher 18–24 engagement.
- Lens: ~20B visual searches/month, 1 in 4 with commercial intent; product search and shopping ads launched alongside results — a new monetizable surface with hard volume attached.
- Gemini: API calls ~40x in 6 months (Pichai later said 14x — an internal inconsistency worth flagging); now on GitHub Copilot; all seven 2B+ user products on Gemini (Maps newly crossed 2B); Gemini app momentum claimed but no user number given despite Sandler's direct ChatGPT-comparison question.
- Cloud: customer engagement suite launched (Volkswagen myVW); Gemini for Workspace — 75% of daily users say it improves work quality; Vertex/BigQuery/security adoption metrics above.
- YouTube: $50B+ combined ad+subscription revenue over trailing four quarters — first time; Nielsen US streaming #1 maintained; creators earning majority of revenue on TV screens +30% YoY; Olympics: 12B views, 850M unique viewers, 40B minutes, 35% on TV; Shorts: 70B daily views, 70% of uploading channels upload Shorts, 3-minute Shorts launched, first-position Shorts ads in ~40 markets; Veo coming to Shorts later this year.
- Waymo: 150K+ paid rides/week (3x Q2's 50K), 1M+ autonomous miles/week, Uber (Austin/Atlanta) and Hyundai partnerships, 6th-gen unit-cost reduction — LA re-entered the narrative ("you've seen us move into L.A.").
- Pixel 9 series launched with "strong demand" and awards; the pull-forward drove the +28% subscriptions/platforms/devices print.
Sell-side read-through- Nowak (Morgan Stanley): which GenAI search products drive durable growth and what constrains scaling — Pichai named AIO/Circle/Lens, promised significant 2025 search evolution, and said he sees no real constraint at 1B users given latency/cost progress.
- Anmuth (JPMorgan): TPU advantage and CapEx trajectory — extracted the quarter's two biggest disclosures: the 60/40 server/data-center CapEx split and the first 2025 CapEx signal (increase, smaller percent step-up, details on Q4 call). Pichai pointed to Flash pricing as external evidence of TPU efficiency.
- Sheridan (Goldman): Waymo city-scaling learnings (Pichai: consumer adoption "surprised us on the positive"; per-city launches getting easier; multiple go-to-market models being tested) and Shorts consumption vs. monetization (Schindler: 70B daily views, gap narrowing, still no rate number — third straight quarter without quantification).
- Sandler (Barclays): two sharp probes. (1) Gemini usage vs. ChatGPT's 250M weekly actives — declined to give a number, pivoted to 1B AIO reach and API growth. (2) The first antitrust question in three quarters: DOJ remedy risk to Apple ISA/Android pre-installs — Pichai declined to speculate, vowed to "vigorously defend," called DOJ proposals "far reaching" with "unintended consequences." No contingency plan disclosed.
- Post (BofA): AIO engagement for the 1B-user base and monetization of previously unmonetized informational queries — got the parity claim ("approximately the same rate") and an affirmative that under-monetized queries are an opportunity. The most concrete AIO monetization statement to date, still without data.
- Nathanson (MoffettNathanson): organizational changes behind AI velocity — Pichai's "new synapses" answer; third-gen Gemini in progress.
- Mahaney (Evercore): Cloud margin path vs. peers' ~30% — Ashkenazi credited scale + efficiency, flagged investment headwinds, "more to come," no target. Also asked about enterprise-wide cost efficiency — she made it a personal priority.
- Ju (UBS): cloud GenAI sales cycles (Pichai: customers leaning in, real proof points, momentum into 2025) and PMax upper-funnel drift (Schindler redirected to Demand Gen — a soft non-answer).
- Structural notes: nine questions; antitrust finally surfaced (Sandler); still no questions on Network, capital return sizing, or Waymo unit economics despite the 6th-gen cost-reduction claim.
Management credibility- Delivered on pre-flags: Q3 hardware pull-forward hit cost of revenue exactly as warned; depreciation headwind materialized; the tax-timing FCF drag recurred as promised; ads-in-AIO launched on the Q2 timeline; AIO country expansion executed (100+ countries vs. "through the course of the year").
- The 40% Services margin Porat wouldn't endorse held for a third quarter — the Q2 deflection now reads as conservatism, not concealment.
- New quantification where assertions previously stood: AIO cost per query −90% (vs. Q2's vague "flat"), >25% of new code AI-generated (vs. Q2's "very early stages"), Waymo 150K rides/week. Disclosure cadence on contested items improved this quarter — except Shorts monetization rate (third quarter unquantified) and Gemini app users (refused under direct fire).
- Inconsistency logged: Pichai cited Gemini API growth as "~40x in 6 months" in prepared remarks and "14x in the past 6 months" answering Sandler — unreconciled on the call.
- The Q1 "YouTube + Cloud $100B+ combined run rate exiting 2024" claim was again unreferenced — but YouTube's $50B TTM disclosure plus Cloud's $45.6B annualized pace implicitly substantiates it; management let the math speak without claiming the win.
- Ashkenazi's debut was substantive: 2025 CapEx direction, CapEx mix, Q4 guidance, and a personal cost-efficiency mandate — more forward disclosure in one call than the prior two quarters provided. The EC fine payment, however, was not pre-flagged.
- Cloud margin jump (11%→17%) was explained with specific drivers rather than hand-waved; the "billions YTD" AI revenue framing from Q2 was dropped in favor of adoption metrics — still no auditable AI revenue figure.
What changed versus the prior quarter- Growth reaccelerated: consolidated +15% (from +14%), Cloud +35% (from +29%), subscriptions/platforms/devices +28% (from +14%, hardware-assisted). Search decelerated to +12% (from +14%) with the lead vertical flipping to financial services/insurance; YouTube ads roughly stable at +12%; Network improved to −2% (from −5%).
- Cloud margin stepped from 11% to 17% — the "young profitability story" aged several notches in one quarter.
- AIO moved from test-announcement to live monetization (US mobile ads in AIO) plus a 100-country expansion and a monetization-parity claim; cost disclosure upgraded from "flat per serve" to "−90% over 18 months."
- Waymo tripled its disclosed ride base (50K→150K/week) and added Uber and Hyundai partnership paths; LA re-entered the narrative after going unmentioned in Q2.
- 2025 CapEx silence broken: increase coming, smaller percent step-up, details on Q4 call. CapEx mix (60/40 servers/data centers) disclosed for the first time.
- Capital return re-entered the disclosure set ($15.3B + $2.5B; ~$70B TTM) after a Q2 call with none; cash fell to $93B.
- Antitrust moved from absent to asked: Sandler's DOJ-remedy question is the first in three quarters; management's answer was defense posture only, no mitigation plan.
- CFO transition completed: Ashkenazi reaffirmed the program and added "accelerate/pivot" optionality; the $607M facilities charge signals the reengineering continues under new ownership.
- New drag items: $3B EC fine payment; Q4 subscriptions/platforms/devices headwind pre-flagged (pull-forward mirror).
Bull case- Reacceleration with margin: +15%/+16% cc revenue, +34% operating income, 32% margin, Services at 40% for a third straight quarter — the reengineering program is compounding while AI investment ramps.
- Cloud is the clearest AI monetization proof in the portfolio: +35% growth, 17% margin (from 11%), GCP above segment growth, 40x API growth, and named customer outcomes with hard numbers (LG −72% cost, Hiscox days→minutes, Snap 2.5x engagement).
- The AIO margin-dilution thesis took another hit: cost per query −90% in 18 months, monetization "approximately the same rate," 1B+ users, ads live inside overviews — the bear case's core mechanism is being dismantled with data.
- Search's surface area is expanding into monetizable territory: Lens at 20B monthly visual searches with 25% commercial intent now carrying shopping ads; Circle to Search with a 33% weekly-retention metric.
- YouTube crossed $50B TTM ad+subscription revenue, holds Nielsen #1, upfronts +20%, and CTV creator economics are scaling (+30% majority-TV-revenue creators).
- Waymo is scaling like a real business: 3x rides in a quarter, 1M autonomous miles/week, unit costs down on gen-6, and capital-light expansion paths via Uber/Hyundai.
- Internal AI productivity (>25% of new code) is a direct offset to the CapEx/depreciation ramp, and the new CFO is signaling more efficiency extraction, not less.
Bear case- Search decelerated to +12% with comps (APAC retail strength) flagged as a continuing H2 headwind; the reacceleration is Cloud- and hardware-assisted, not Search-led.
- The most-wanted consumer AI number was refused: Sandler's Gemini-vs-ChatGPT usage question got deflection, and the 40x/14x API discrepancy undercuts confidence in the usage-growth narrative.
- AIO monetization parity is still an assertion — no CTR, no revenue-per-query data, three quarters of requests without audited figures; Shorts monetization rate likewise unquantified for a third quarter.
- CapEx rises again in 2025 with no ceiling and no ROIC framework; depreciation headwinds are explicitly ongoing; TTM FCF compressed to $55.8B (from $69.1B two quarters ago) while cash fell to $93B and capital return runs at ~$70B/year.
- Antitrust is now a live line of questioning: DOJ remedies against the Apple ISA and Android pre-installs strike at Search's distribution core, and management offered no mitigation strategy — only litigation posture.
- Q4 is pre-loaded with headwinds: subscriptions/platforms/devices faces the pull-forward mirror, ad comps stay tough, depreciation keeps climbing.
- The $3B EC fine payment surfaced unannounced — regulatory cash costs are recurring and unpredictable.
- Cloud margin's six-point jump invites the question of how much is one-off efficiency vs. durable; Ashkenazi herself flagged investment headwinds against it.
Next-quarter watchlist- The Q4-call 2025 CapEx guide: magnitude of the increase, whether a framework or ceiling accompanies it, and depreciation trajectory against the FY margin-expansion commitment.
- Ashkenazi's "accelerate and pivot" mandate: first evidence of reallocation (headcount, footprint, Other Bets); whether the facilities-charge cadence continues.
- AIO: any hard monetization/CTR data behind the parity claim; ads-in-AIO expansion beyond US mobile; whether cost-per-query keeps falling as coverage grows.
- Gemini usage disclosure: whether management ever puts a consumer number against ChatGPT's 250M weekly actives; reconciliation of the 40x/14x API figures; third-gen Gemini launch timing; Project Astra "as early as 2025."
- Cloud: growth vs. +35%, margin vs. 17% (is the jump durable?), and any auditable AI-revenue figure replacing adoption anecdotes.
- Search: whether +12% holds against APAC comps; financial-services vertical durability; Lens shopping-ads traction.
- YouTube: Shorts monetization quantification (fourth chance); Veo-in-Shorts launch; NFL Sunday Ticket renewal economics; upfront conversion into revenue.
- Subscriptions/platforms/devices: magnitude of the flagged Q4 hardware headwind; refreshed subscriber counts (still none given).
- Waymo: ride growth off 150K/week; Austin/Atlanta Uber launch execution; any unit-economics disclosure behind the gen-6 cost claim; cadence of the $5B commitment.
- Antitrust: DOJ remedy filings and any management framework for distribution risk (Apple ISA, Android pre-installs) — now that the question is on the table, silence gets harder.
- Cash and capital return: whether cash stabilizes near $93B or the buyback keeps drawing it down alongside the CapEx ramp.
|
Corbanu