Alphabet Q2 2026: Cloud +82%, CapEx $205B и gap Gemini 3.5 Pro
Кто и какую проблему решает? Mac-разработчики и инвесторы, которым нужно декодировать GOOGL after-hours, Google Cloud backlog и Gemini stack после отчёта 22 июля 2026 — с primary source, не headline CNBC. Что внутри: метрики $119,8B revenue, Cloud +82%, CapEx guidance $195–205B, задержка Gemini 3.5 Pro, ASIC Frozen v2, матрица pain points, 7-step validation pipeline. Структура: TOC, таблицы метрик/CapEx/Frozen v2/spending, Q1/Q2 compare, Q3 checklist, FAQ×6, Mac rental CTA.
Содержание
Контекст: GPT-5.6 release, DeepSeek V4 GA, Kimi K3.
Featured Snippet
22 июля 2026 Alphabet (NASDAQ: GOOGL/GOOG) опубликовал Q2: revenue $119,8B (+24% YoY, beat), Google Cloud $24,8B (+82% YoY). After-hours −4% (~$327) из-за поднятия full-year CapEx guidance до $195–205B — growth priced in, infrastructure bill repricing начался.
01 · TL;DR (hard numbers)
- Revenue beat: $119,8B vs consensus $116,9B; diluted GAAP EPS $9,11 (вкл. ~$99B unrealized gains по stakes Anthropic/SpaceX).
- Cloud inflection: +82% YoY acceleration (Q4 +48%, Q1 +63%); op margin 35,6%; backlog $514B (+$50B QoQ).
- Ads intact: Search $63,3B (+17%), YouTube Ads $11,1B (+13%), total ads $81,6B.
- CapEx shock: Q2 $44,9B (+100% YoY); 3rd guidance raise; FCF −$5,9B vs +$5,3B YoY.
- Model gap: Gemini 3.5 Pro delayed; Frozen v2 = long-run inference cost lever.
02 · Pain matrix (decision engineering)
Три structural conflicts в одном 10-Q — без декомпозиции получаете wrong trade / wrong stack bet.
| Pain point | Affected | Signal | Mitigation |
|---|---|---|---|
| Beat vs −4% | Traders, PMs | CapEx path, FCF | Read CapEx timeline before EPS |
| GAAP EPS distortion | Model builders | $9,11 vs ~$2,88 adj | SEC Exhibit 99.1 line-by-line |
| Single-vendor Gemini | Mac devs, GCP | 3.5 Pro no GA date | Multi-model fallback + isolated node |
- Narrative split: retail смотрит EPS beat, institutions — CapEx trajectory + FCF inflection.
- GAAP vs adjusted: ~$99B mark-to-market по equity stakes ломает YoY EPS compare; adjusted ~$2,88 — in-line.
- Stack lock-in: backlog $514B bullish для GCP, но coding/agent frontier до 3.5 Pro GA требует GPT-5.6 / Claude Fable 5 / Kimi K3 routing.
03 · Key metrics table (Q2 2026)
| Metric | Q2 2026 | Consensus | Q2 2025 | YoY |
|---|---|---|---|---|
| Total revenue | $119,8B | $116,9B | $96,4B | +24% ✅ |
| Diluted EPS | $9,11* | ~$2,88 | $2,31 | Beat ✅ |
| Google Cloud | $24,8B | $22,4B | $13,6B | +82% ✅ |
| Google Search | $63,3B | $63,4B | $54,2B | +17% |
| YouTube Ads | $11,1B | — | $9,8B | +13% |
| Operating income | $40,8B | — | $31,3B | +30% |
| Cloud op margin | 35,6% | — | 20,7% | +14,9 pp |
| Q2 CapEx | $44,9B | — | $22,5B | +100% |
| Free cash flow | −$5,9B | — | +$5,3B | ⚠️ negative |
| Cloud backlog | $514B | — | ~$464B | +$50B QoQ |
*GAAP EPS incl. ~$99B unrealized equity gains; adjusted ~$2,88 — within consensus.
Hard metric #1: backlog >5× Cloud quarterly run-rate — abnormal revenue visibility среди hyperscalers (source: Alphabet Q2 release).
04 · Google Cloud: +82% YoY acceleration
Growth curve: Q4 2025 +48% → Q1 2026 +63% → Q2 2026 +82%. Operating income $8,8B (prior $2,8B), margin 35,6% (+1490 bps YoY).
Drivers (verified on call)
- Enterprise AI infra: TPU/GPU capacity leasing spike
- Gemini Enterprise: ~90% Fortune 100 deployed
- Core GCP workloads: steady lift
- First TPU system hardware sales — customer DC delivery
- Wiz integration: security pipeline expansion
Sundar Pichai: «Q2 was remarkable — Alphabet +24% YoY, Google Cloud accelerated to 82%, driven by AI infrastructure and AI solutions demand.»
Hard metric #2: Gemini API throughput 22B tokens/min (Q1: 16B); Antigravity 2,4M WAU — product-side cross-check Cloud revenue curve.
05 · Search & YouTube: AI не cannibalize ads (yet)
- Search & Other: $63,3B (+17% YoY)
- YouTube Ads: $11,1B (+13%)
- Total ads: $81,6B
AI product stats: AI Mode >1B MAU; Gemini App 950M MAU; Pichai: AI search features drive billions of clicks/week to publishers — zero-click thesis not confirmed in Q2 dataset.
06 · Beat + −4% after-hours: CapEx dominates repricing
6.1 CapEx guidance — 3rd raise
| Timestamp | FY2026 CapEx guidance |
|---|---|
| Early 2026 | $175B – $185B |
| Q1 earnings (Apr) | $180B – $190B |
| Q2 earnings (Jul) | $195B – $205B |
Q2 CapEx $44,9B (+100% YoY). Split: ~60% servers/TPU, ~40% DC/network. Management: 2027 CapEx «materially higher» than 2026.
6.2 FCF inflection
FCF −$5,9B vs +$5,3B prior year at record revenue — classic «growth priced, bill arriving».
6.3 Gemini 3.5 Pro overhang
CFO Anat Ashkenazi: ROI framework + stronger demand signals vs year ago; market compares vs OpenAI/Anthropic/xAI July model drops.
07 · Gemini 3.5 Pro delay — technical postmortem
Target GA: June 2026. Status: delayed, no public ship date.
- Bloomberg (Jul 16): coding/agent performance below internal bar — months slip
- Late-June training data refresh for code — insufficient lift
- Model exists; partner testing 3.5 Pro + upgraded Flash
- No GA date; Raymond James: perception shift leading → trailing edge
- Jul 16 rumor: ~$2T market cap wipe (−4,4% single day)
Competitive set (Jul 2026): GPT-5.6, Claude Fable 5, Grok 4.5 shipped. Google counter: Gemini 4 R&D, ~monthly model releases, Gemini 3.5 Flash Cyber (security vertical) live.
08 · Frozen v2 — ASIC spec table (The Information)
| Attribute | Spec |
|---|---|
| Architecture | Partial Gemini model graph embedded in silicon |
| Efficiency | Est. 6–10× vs current TPU (tokens/watt) |
| Scope | Gemini-specific; not general-purpose TPU replacement |
| Positioning | TPU complement; low initial volume |
| P&L impact | TPU hardware revenue recognition ~2027; Frozen v2 = margin story 2027+ |
Hard metric #3: at 6–10× inference efficiency, structural COGS advantage vs pure Nvidia-GPU competitors on Gemini serving — ties to $514B backlog margin sustainability. If Google ships Frozen v2 at scale in 2027, unit economics on Gemini Enterprise inference could decouple from external GPU spot pricing — watch for any TPU/Frozen utilization metrics in future 10-Q footnotes.
09 · Spending table — capital allocation map
| Line item | Evidence |
|---|---|
| AI datacenters | ~40% CapEx facility/network |
| AI compute hardware | ~60% CapEx servers/TPU |
| Equity raise (Jun 2026) | $49,6B stock + preferred |
| Debt (Q2) | $20,3B senior unsecured notes |
| Cash position | $242,5B cash + securities post-raise |
| Third-party compute Q3 | Cloud margin short-term compression |
Open question for FY2026–2027: does $200B+/yr CapEx generate equivalent incremental AI revenue slope?
10 · Q1 vs Q2 2026 YoY compare
| Segment | Q1 YoY | Q2 YoY | Trend |
|---|---|---|---|
| Total revenue | +22% | +24% | ⬆️ accelerate |
| Google Cloud | +63% | +82% | ⬆️ strong accel |
| Search | +19% | +17% | ⬇️ slight decel |
| YouTube Ads | +11% | +13% | ⬆️ reaccel |
| Operating margin | 36% | 34% | ⬇️ investment pressure |
11 · Q3 2026 watchlist (7 signals)
- Gemini 3.5 Pro GA date — equity catalyst
- Search comp base — CFO: tougher YoY in Q3
- Cloud margin — third-party compute bridge
- 2027 CapEx guidance — «materially higher» language
- Competitive releases — Kimi K3, GPT-5.6, Fable 5
- Frozen v2 production timeline — any leak = major signal
- External TPU revenue recognition — 2027 acceleration
12 · 7-step validation pipeline (earnings window)
- Isolated node: Gemini API baseline — long-doc summary, code completion, multi-file refactor; log p50/p95 latency + $/task
- Antigravity (2,4M WAU) vs Cursor + DeepSeek V4 on same repo — SWE-bench proxy tasks
- GCP billing drill-down: AI infrastructure line items vs CapEx expansion — backlog growth ≠ unit price drop
- Wire multi-model fallback (OpenRouter/custom router) until 3.5 Pro GA
- Primary sources: Q2 PDF + SEC 99.1
- API keys off production Mac — ephemeral Apple Silicon or daily rental; zero Keychain pollution
- Stakeholder ROI memo: Cloud +82% vs CapEx +100% — documented decision artifact
13 · Summary: growth story vs ROI story
Bull case: AI monetization measurable — Cloud +82%, Search growing, Gemini ~1B-user scale. Full-stack (model + ASIC + cloud + consumer) producing quantifiable financial output.
Bear case: frontier cost scaling superlinear — $200B CapEx, negative FCF, flagship model gap. Market reprices bill before rerating growth.
Q2 2026 = 2026 AI commercialization inflection: growth narrative proven; ROI narrative still under test. Next binary events: 3.5 Pro GA date, Q3 Cloud margin after third-party compute bridge, any 2027 CapEx quantification on October call.
14 · FAQ
Q: Key Q2 2026 numbers?
A: Revenue $119,8B (+24%), Cloud $24,8B (+82%), beat $116,9B consensus.
Q: Why stock down on beat?
A: CapEx $195–205B, Q2 $44,9B, FCF −$5,9B, Gemini delay.
Q: Gemini 3.5 Pro delayed?
A: Yes — code bar missed, partner testing, no GA date.
Q: Frozen v2?
A: Gemini ASIC, 6–10× efficiency vs TPU, complementary.
Q: Cloud vs AWS/Azure Q2?
A: +82% leads hyperscalers; $514B backlog — peer Q2 compare separately.
Q: Gemini-only or multi-model?
A: Deepen GCP/Gemini Enterprise; keep GPT/Claude/Kimi fallback for coding until 3.5 Pro GA. Log routing decisions with model version, token cost, and error rate — minimum viable observability for multi-vendor agent stacks.
15 · Mac rental CTA — isolated Gemini stack validation
Antigravity + Xcode + macOS Keychain = real-world Google AI stack eval surface. Post-earnings CapEx/model cadence shifts API pricing/latency — ephemeral Apple Silicon node runs Gemini API + Antigravity without polluting production Keychain or burning local RAM on 1M-token contexts.
Linux VPS handles pure API scripts; cannot validate macOS agent IDE integration. Daily rental fits 1–3 day earnings-window acceptance — pricing: M-series compute rates. Capture p50/p95 latency, tokens/min quota headroom, and $/1M tokens in a reproducible benchmark sheet — required input for any production routing decision during the 3.5 Pro gap window.
16 · Sources (primary first)
- Alphabet Q2 2026 PDF
- SEC Exhibit 99.1
- CNBC Live (Jul 22, 2026)
- Bloomberg (Jul 16, 2026): Gemini 3.5 Pro delay
- The Information: Frozen v2 reporting
Updated: 2026-07-23 | Data: Alphabet Q2 2026 earnings release