Alphabet (GOOGL) — Q3 FY2026 Earnings Briefing
Fiscal period: Q3 FY2026 · Call date: 2026-10-28
Executive Summary
- A high-octane, high-investment quarter: revenue of $127.6 billion, up 25% YoY in both reported and constant currency — a half-trillion-dollar annual run rate.
- Google Cloud revenue grew 103% YoY, a record quarter powered by GCP (the vast majority of segment revenue) and TPU system sales.
- Demand outruns supply: third-party bridging compute came online only at the very end of Q3 and ramps in Q4, bringing deliberate near-term Cloud margin headwinds.
- The quarter’s tension: operating strength absorbed below the line — negative free cash flow, a SpaceX mark dragging net income and EPS lower, and AI Mode monetization still trailing its usage growth.
- Management held full-year CapEx guidance and anchored the buildout to ROIC “meaningfully above our cost of capital,” framing the AI deployment as a long-range plan with 2027 positioned as a “golden year” of productivity.
Highlights and Lowlights
Highlights
- Google Cloud operating income more than tripled YoY to $11.1 billion, expanding segment margin to 36.2% from 23.7% a year earlier.
- Google Cloud’s backlog reached $561 billion, with just over 55% expected to convert to revenue within 24 months.
- Google Search and other advertising grew 12% YoY to $63.3 billion, with retail driving the largest contribution and finance next.
- First-party model APIs now process nearly 32 billion tokens per minute, up 40% sequentially and more than 4x YoY.
- The Gemini app reached 1 billion users, the fastest-growing product in Google’s history.
- Nearly 100 customers now use TPU systems, with new partnerships signed to supply them to Crux AI and Oracle.
- YouTube Music and Premium subscriptions surpassed 150 million globally (including trials), with subscriptions growing faster than advertising.
Lowlights
- Free cash flow turned negative $9.1 billion in the quarter as capital expenditure, the TPU inventory build, and payment of a 2018 EC fine outweighed operating cash flow.
- Net income and EPS fell YoY as negative $8.2 billion of other income and expense — chiefly an unrealized loss on Alphabet’s SpaceX stake from its Q3 stock-price decline — swamped operating income growth.
- Search decelerated from Q2: management called FX the single largest driver, alongside the EU’s eliminated de minimis exemption hitting APAC retail advertisers and post-FIFA World Cup normalization in media and entertainment.
- Management said Search lapped about 3 points of prior-year growth acceleration in Q3, with a couple more points of lapping expected in Q4.
- Strong AI Mode query growth weighed on Search revenue because monetization on that surface remains lower than traditional search.
- Network advertising grew just 1% YoY to $7.5 billion, by far the slowest of the major ad lines.
- Future purchase commitments stood at $924 billion at quarter-end, joined by $152 billion of pre-commencement lease obligations, hard-coding multi-year infrastructure spend.
1. Consolidated Results
| Line item | Q3 2026 | YoY |
|---|---|---|
| Revenues | $127.6B | +25% (reported and constant currency) |
| Total cost of revenue | $51.4B | +24% |
| — Traffic acquisition costs (TAC) | $16.4B | +10% |
| — Other cost of revenues | $35.1B | +32% |
| Total operating expenses | $35.6B | +20% |
| — R&D | — | +36% |
| — Sales & marketing | — | +22% |
| — G&A | — | −17% |
| Operating income | $40.6B | +30% |
| Operating margin | 31.8% | — |
| Other income and expense (OI&E) | −$8.2B | — |
| Net income & EPS | — | Decreased YoY (driven by OI&E) |
This was Alphabet’s 13th consecutive quarter of double-digit revenue growth. Other cost of revenues rose fastest among cost lines, driven by hardware costs primarily related to TPU system sales, higher depreciation, content acquisition costs, and technical infrastructure operations costs. R&D growth reflects compensation — principally investment in AI talent — plus depreciation; sales and marketing growth reflects investment in Gemini and compensation; G&A declined on lapping higher legal expenses from Q3 last year, partially offset by current-year legal charges. The OI&E swing is treated under the Lowlights; the SpaceX position drove it.
2. Segment Performance
| Segment | Revenue | YoY | Operating income | Margin |
|---|---|---|---|---|
| Google Services | $96.3B | +11% | $39.2B (+17%) | 40.7% |
| Google Cloud | $30.7B | +103% | $11.1B (more than tripled) | 36.2% (vs 23.7%) |
| Other Bets | $372M | — | −$2.0B | — |
| Alphabet-level activities | — | — | −$7.7B | — |
| Revenue line | Q3 2026 | YoY | Drivers named |
|---|---|---|---|
| Total advertising | — | +11% | Growth across all major verticals; FX a slight headwind YoY, more significant sequentially |
| Google Search & other | $63.3B | +12% | Retail the largest contribution, followed by finance |
| YouTube ads | $11.2B | +9% | Brand the largest contribution; notable living-room strength |
| Network | $7.5B | +1% | — |
| Subscriptions, platforms & devices | $14.3B | +11% | YouTube subscriptions (Music and Premium in particular); strong demand for AI plans in Google One |
Google Services growth reflected strong Search and Subscriptions. Google Cloud delivered record growth: GCP — with strength in both AI (infrastructure and solutions) and core GCP — accounted for the vast majority of Q3 revenue, TPU system sales were a meaningful contributor, and the CFO called the Cloud growth rate excluding TPU systems “incredibly strong.” The third-party bridging capacity discussed on the prior call came online only at the very end of Q3, so it had limited margin impact this quarter but ramps in Q4 (see the Outlook). The Other Bets loss reflects the continued expansion of Waymo and investment in key bets; the Alphabet-level activities loss is driven by shared AI R&D expenses.
3. AI: The Gemini 4 Series and Model Velocity
Alphabet is now shipping frontier models on a compounding cadence:
- Argon, launched two weeks before the call, ramped first-week developer-platform usage 3.5x faster than Gemini 3 — despite capacity constraints — and is proving strong in finance, legal, cybersecurity, and science; customers are applying it to untangling legacy codebases, accelerating scientific research, robotics, and domain tasks requiring in-depth reasoning.
- Carbon, announced on the call, is positioned as Alphabet’s strongest frontier model, with state-of-the-art cybersecurity and long-horizon capabilities. It ranks number 1 on the recently updated Artificial Analysis Intelligence Index at 59 and outperforms Argon across almost every domain — full-stack web development, multimodal complex knowledge work, and scientific reasoning. Rollout starts with a subset of trusted cyber defenders before broader access, and it is being tested by the US government.
- Genie 4, announced the day before the call, is the company’s most advanced world model, simulating safe, realistic, interactive worlds with native audio in real time — a building block for training robots and agents.
Internally, teams across Chrome, Ads, and Search have used Argon to ship 35% more code with 30% fewer errors. Sundar Pichai said the company has reached the point where AI increases its own training and launch velocity — expecting “continued acceleration because of RSI” over the next 12 months — with more efficient model versions coming for speed- and cost-optimized deployment across billions of users. Distilled Carbon versions are in the works for billion-user scale across compute, latency, and quality.
On agentic deployment, Alphabet went enterprise-first with the Gemini Agent for Enterprise — chosen because enterprises hold the highest bar for security, governance, and controls — and will bring the same experience to consumers later this year, starting with the Gemini app’s user base (see Highlights) and extending agents as a unifying layer across Chrome, Android, and Workspace.
4. Search: Usage Inflection vs. Monetization Lag
Management framed Search’s AI transition as analogous to the mobile shift: best user experience first, monetization to follow. The usage evidence is unambiguous — the year-over-year query growth rate is the highest it has been in five years, including commercial queries. AI Mode behavior is structurally different: queries average 3x longer than traditional searches, 1 in 5 AI Mode searches globally use voice, images, or documents, and the company crossed 50 billion visual searches per month.
The monetization case is being built deliberately. The majority of AI Mode queries are incremental, many are upper-funnel queries capturing new high-value commercial intent, user research shows people are comfortable with ads on these surfaces and find them relevant, and click-through rates are already similar between sponsored and organic product list items within AI Mode. On that basis Alphabet is scaling ads within AI Overviews and AI Mode to more users and queries — taking its AI-search monetization footprint global for the first time and adding multi-turn ads in AI Mode in the US.
The friction: the strong AI Mode query growth itself weighed on Search revenue this quarter because monetization there remains lower than traditional search (see Lowlights), and Q2-to-Q3 deceleration drivers plus the lapping dynamics run through the outlook (see section 10).
5. Advertising and YouTube
Ads infrastructure. In Q3 Alphabet made the biggest update to Product Listing Ads in a decade, adding AI capabilities that raised satisfaction 19% and cut dissatisfaction by more than half. Personalization-model innovations delivered 18% incremental advertiser value and an 8% increase in user sentiment on Discover ads in 2026. Gemini models helped advertisers generate over 1.5 billion campaign assets in Q3 alone, up over 7x YoY. AI Max and Performance Max continue strong adoption, with AI creative tools now integrated directly into AI Max and a new single AI Max A/B test for budgets and ROI targets across multiple Search campaigns. Telstra’s use of AI Max to capture long-form and multilingual queries drove a 52% increase in conversions and a 36% reduction in cost per order.
Agentic commerce. The Universal Commerce Protocol enables direct checkout for hundreds of thousands of brands across Google — including Chewy, Nike, and Tapestry — now being combined with the Agent Payments Protocol. Universal Cart, launched this quarter, mirrors cross-Google shopping behavior so retailers keep customer relationships while abandoned carts are captured. Philipp Schindler positioned agentic experiences as “clearly additive”: automate the grunt work, enhance the discovery experience, and design time and space for choice — which he argued opens “really interesting, compelling advertising models for agentic use cases,” potentially extending to subscriptions and other payment methods.
YouTube. More than 20 million videos are uploaded daily. Made on YouTube brought over 30 products and features, including Custom Feeds (topic-based tailored home feeds) and Shorts series with seasons, episodes, and sequential playback, alongside generative-AI creation tools in YouTube Studio and a conversational editing assistant in Shorts and YouTube Create. The living room accounted for over one-third of total YouTube ads revenues in Q3. Demand Gen is driving direct response with small and medium advertisers — Wayfair’s Demand Gen campaigns delivered a 2.3x higher return on ad spend than traditional video reach campaigns. The Creator Partnerships program connects brands with over 3 million creators (Lexus working with Colin and Samir drove an 11% lift in brand consideration), more than 1.3 million channels are enrolled in YouTube Shopping, and Amazon is joining YouTube Shopping — US creators can now tag Amazon products directly in videos, Shorts, and livestreams. The subscription ecosystem (see Highlights for the subscriber count) continues to grow faster than advertising, led by YouTube Music Premium, with Premium and Premium Lite giving viewers choice.
6. Google Cloud: Breadth, Backlog, and the Gemini Flywheel
Google Cloud now has 18 product lines with more than $1 billion each in annual revenue. Thomas Kurian attributed momentum to an integrated AI portfolio — chips, models, data, security, and agent platforms designed to work together — with strong diversification across product lines, industries, and customers.
Gemini Enterprise traction. Over the past 12 months, more than 2,600 customers used at least 100 billion tokens and nearly 75 at least 10 trillion tokens. Nearly 90% of the Fortune 100 use Gemini Enterprise, customers who used it in Q3 2025 more than quadrupled their spend YoY, and Alphabet has sold more than 100 million seats for knowledge work across its AI services. The Gemini agent is now deeply integrated across the cloud portfolio — research and answers, writing and running code, and process automation through one agent, one API, and a single prompt window, with enterprise security and cost controls and the widest choice of frontier models.
| Customer | Deployment evidence |
|---|---|
| BNP Paribas | Deploying Gemini Enterprise to more than 65,000 employees with enterprise-grade security and governance |
| Tata Steel | 300+ specialized AI agents in 9 months; manufacturing safety, supply-chain optimization, 50% cut in customer service times |
| Ulta Beauty | Ulta AI shopping assistant drives a 3x higher conversion rate than traditional browsing |
| Deutsche Telekom | Unified data lakehouse across 40+ legacy on-premise systems (BigQuery, Cloud Storage, Knowledge Catalog) |
| Etsy | Migrated 3 petabytes to a unified lakehouse (BigQuery + Iceberg); critical data joins up to 60% faster |
Data, analytics, and security. BigQuery improved query speed over 40% YoY while cutting query processing cost by over 50% YoY, with Knowledge Catalog improving accuracy and lowering token cost and the lakehouse providing zero-copy access across systems. The Gemini Flash Cyber models automate vulnerability discovery across 20 programming languages and patch 2.6 times more reliably than other models — with further gains expected as Argon and Carbon roll out. Gemini in Google Threat Intelligence autonomously analyzes millions of external events daily with 98% accuracy, more than 95% of Wiz customers are adopting AI-powered features, and the security platform protects critical infrastructure including Banorte, Jack Henry, and AT&T.
The flywheel and partnerships. Kurian described a three-part growth dynamic: Gemini generates revenue directly, drives attach of compute, storage, and databases as agents run, and expands the security portfolio — while new-customer adoption accelerates. He acknowledged Alphabet “could obviously serve a lot more demand with capacity.” Strategic partnerships reinforce the platform: an expanded American Express relationship (contact-center modernization plus agentic commerce across Search and Wallet) and Unity’s expanded multi-year Cloud deal to power its next-generation game engine on the full AI stack.
7. The TPU Systems Business
The TPU story matured into a distribution business this quarter:
- Silicon: TPU 8T (training) offers up to 2.7x better training performance and TPU 8i (inference) 80% better inference price performance versus the previous generation; the broader stack spans Axion CPUs, the latest NVIDIA GPUs, and an Agent Compute Engine designed for AI agents.
- Software portability: TorchTPU, built closely with Meta, lets users train, serve, and deploy PyTorch models on TPUs with familiar tools — which Kurian said “opens up a much larger market.”
- Distribution: beyond Alphabet’s cloud, TPU systems are now sold into customer data centers and via other providers — partnerships with Oracle and Crux AI (a joint venture launched along with Blackstone), with “more in the works.”
- Customers: nearly 100 TPU customers (see Highlights), including Anthropic directly and via Broadcom, Apple scaling pre-training of foundation models on the latest generation, frontier model companies Mirandil and Physical Intelligence, capital-markets firms Citadel Securities and Jump Trading, Caltech, and public-sector providers Cassava and Cirruscale.
The financial mechanics (CFO Anat Ashkenazi): a signed agreement first shows up in backlog, then inventory builds (cash out) ahead of revenue, and cash comes in as systems are delivered and billed — “not the traditional kind of CapEx… rather a different cycle of cash flow generation.” Some system sales have been recorded as revenue to date, with more expected through 2027. Kurian added that component procurement happens only after contracts are signed, so all TPU inventory sits behind committed contracts; placing systems in third-party facilities broadens the addressable market (accelerated compute brought to where data cannot move) and offsets the need to pre-build facilities. On margins, management would not break out product-level economics but said it is “very focused on continuing to improve margins as we scale,” citing price-performance versus competitors and the scarcity of accelerated compute; Pichai added that semiconductor economies of scale make dual deployment (Alphabet’s and others’ data centers) “directly translate into better economics.”
One figure circulates only from the buy side: Ross Sandler’s estimate that a gigawatt of TPU full system costs around $35 billion is analyst-provided — management did not confirm or dispute it, declining product-level financials.
8. Capital Intensity, Cash Flow, and ROIC
| Metric | Q3 2026 | Trailing 12 months |
|---|---|---|
| Operating cash flow | $42.3B | $179.6B |
| CapEx | $51.4B | — |
| Free cash flow | −$9.1B | $19.7B |
| Cash & marketable securities | $265.1B (incl. $90.4B marketable equity securities) | — |
| Long-term debt | $125.1B | — |
| Future purchase commitments | $924B | — |
| Pre-commencement leases & other obligations | $152B | — |
| Quarterly dividend | $0.22/share, payable December | — |
The vast majority of CapEx went to technical infrastructure supporting AI. Approximately 65% of Q3 technical-infrastructure investment was servers and networking equipment (generally 6-year useful lives); the remaining 35% was data centres, with useful lives stated on the call as 7 to 40 years. Operating cash flow absorbed the TPU inventory build plus a $5.2 billion cash payment for a 2018 EC fine following the conclusion of the appeal process. The purchase commitments and lease obligations represent maximum contractual obligations; management noted many agreements carry variable pricing and volume flexibility that could reduce actual spend.
On returns, Ashkenazi argued Alphabet is in a unique position to monetize AI investments across products reaching billions of users and enterprises: Cloud has generated tens of billions of dollars of AI revenues year to date, Google One is the fastest-growing subscription product on AI-plan demand, and the company expects an attractive ROIC that remains “meaningfully above our cost of capital” with strong operating cash flow — while explicitly flagging that infrastructure investment and TPU inventory scaling will pressure free cash flows.
9. Long-Term Bets
- Waymo: more than 300 million fully autonomous miles driven; operating in 15 major US cities, with public service recently opened in San Antonio, San Diego, and Tampa and planned expansions into Canada, Germany, Singapore, and Spain.
- Isomorphic Labs: using Argon to support drug-design research.
- Quantum: first-ever demonstration of magic state cultivation.
- Project Suncatcher: TPUs launched into space.
10. Outlook: Q4 2026 and Full-Year 2026
| Item | Guidance from the call |
|---|---|
| FX | Slightly larger headwind to consolidated revenue in Q4 than in the prior quarter (at current spot rates), primarily Search and YouTube ads |
| Search lapping | Continued pressure from lapping the search acceleration that began in Q3 2025 — a couple of points in Q4 |
| TPU systems | Sales continue to scale in Q4, followed by a meaningful revenue ramp in 2027 |
| Cloud margins | Third-party compute capacity usage ramps in Q4; near-term margin headwinds expected |
| FY2026 CapEx | $195–205 billion, unchanged |
| 2027 CapEx | Expected to increase significantly; details promised on the Q4 call |
| Expenses | Higher depreciation and data-center operating costs (such as energy) to keep weighing on the P&L |
Conclusion
This quarter pairs the fastest Cloud growth in Alphabet’s history with a balance sheet now doing the heavy lifting: the company is converting its full-stack position — silicon, models, data, and distribution — into contracted demand faster than it can supply compute, and it is deliberately spending margin and cash flow to close that gap. The result is a genuine tension rather than a clean win: record Cloud profitability, a swelling backlog, and the highest query growth in five years sit against negative free cash flow, an earnings decline from a single investment mark, and an AI Mode monetization gap that management itself quantified in its growth bridge. The investment view now hinges on three executions: converting backlog to revenue as TPU systems ship through 2027, holding Cloud margins through the third-party capacity ramp, and closing the AI Mode monetization gap as ad formats scale globally. Management’s own framing — long-range plans, a disciplined ROIC bar, and inventory built only against signed contracts — is the right discipline to watch as the 2027 CapEx step-up approaches.
Artifacts and Follow-Ups
Key Artifacts — highest priority:
- Cloud margin through the third-party capacity ramp. Baseline: 36.2% in Q3, with third-party bridging capacity online only at the very end of the quarter; management guided to near-term margin headwinds as usage ramps in Q4. Next print confirms or refutes how large that cost is.
- TPU systems revenue conversion. Baseline: some system sales recorded as revenue to date, “more of it to come in through 2027,” against a $561 billion backlog with just over 55% expected to convert within 24 months — watch the inventory-to-billing cash cycle and any new distribution deals.
- Search re-acceleration. Baseline: +12% YoY in Q3, with roughly 3 points of prior-year acceleration lapped and FX the single largest sequential drag; watch whether globally scaled AI Mode/AI Overviews ads plus US multi-turn ads offset both headwinds in Q4.
- CapEx trajectory. Baseline: FY2026 guidance of $195–205 billion held unchanged; the “significant increase” in 2027 CapEx gets detail on the Q4 call — the first hard test of the ROIC framing.
- Model velocity converting to usage. Baseline: nearly 32 billion tokens per minute on first-party APIs (+40% sequentially); watch Carbon’s expansion beyond trusted cyber defenders, US-government testing, and the distilled versions built for billion-user scale.
Secondary Tracking Items:
- Amazon product tagging in YouTube Shopping by US creators — early evidence of direct-sales conversion.
- Waymo’s international expansion (Canada, Germany, Singapore, Spain) after the three new US cities.
- The quarterly dividend payable in December.
- Whether Network advertising returns to meaningful growth.
- Other Bets spend as Waymo scales across more cities and geographies.