First $100B Quarter, $155B Cloud Backlog, and Anthropic Buys a Million TPUs — The Bear Case Has Run Out of Data

Results vs. Consensus

Quality of the Beat

• Revenue: Pure acceleration. Search +15% (vs. +12%), YouTube ads +15% (vs. +13%), Cloud +34% (vs. +32%), Subs/Platforms/Devices +21% (vs. +20%). Every meaningful line item accelerated despite tougher comps.
• Operating margin: The reported 30.5% understates the operational picture by 340bps because the EC fine ($3.5B) was charged to Services in Q3. Excluding it, margin was 33.9% — the best print of 2025 — and Cloud's segment margin alone expanded 660bps Y/Y.
• EPS: Headline $2.87 includes a $12.8B OI&E gain (mostly unrealized non-marketable equity). Stripping that out, "core" EPS was closer to ~$2.20 — still a clean operational beat. Tax law changes (immediate R&D expensing, accelerated depreciation per the July 4, 2025 enacted U.S. tax law) lowered the effective tax rate and aided net income.
• Cloud margin: 23.7% vs. 17.1% Y/Y is a +660bps expansion. Operating income +85% Y/Y. Both materially above what the bull thesis required at this point in the cycle.

Segment Performance

Google Services — Search Reaccelerates Again, Both Volume AND Price Grow

For the third quarter in a row, the bear thesis on AI cannibalization of Search has gotten weaker, not stronger. Search & other revenue accelerated to +15% from +12% in Q2 — a 300bp acceleration on a $54B base, which is operationally enormous. AI Overviews remains at 2B+ MAU and is now scaling ads "below and within the AI's response" to more countries on desktop and mobile. AI Mode reached 75M DAU globally across 40 languages, with queries doubling in the quarter and "100+ improvements shipped in Q3" — and importantly, "AI Mode is already driving total query growth for Search."

The single most important monetization data point of 2025:

"You will see in the 10-Q, paid clicks were up 7% year on year, and CPCs were up 7% year on year." — Philipp Schindler, CBO

Both volume and price are up roughly the same amount — meaning the +15% Search revenue growth is split evenly between more clicks and higher CPC. That is the exact opposite of the cannibalization pattern. AI Max in Search, Schindler said, is now used by "hundreds of thousands of advertisers" and is "the fastest-growing AI-powered search ads product." In Q3 alone, AI Max "unlocked billions of net new queries."

"For AI Overviews, even at our current baseline of ads, below and within the AI's response, overall, we see the monetization at approximately the same rate." — Philipp Schindler, CBO

YouTube ads accelerated to +15% from +13%, helped by the first-ever NFL exclusive global broadcast (19M+ concurrent viewers, all ad inventory sold within weeks) and Living Room interactive direct response ads now at $1B+ annual run-rate. Subscriptions/Platforms/Devices crossed 300M paid subscriptions and grew 21% to $12.9B, driven by YouTube and Google One.

Assessment: Three consecutive quarters of accelerating Search growth alongside AI Overviews / AI Mode scale-up is no longer a coincidence. The bull thesis (AI features expand the addressable query universe, with monetization at parity or better) is now the base case. The bear thesis needs new evidence to remain credible.

Google Cloud — The Defining Quarter of the Cycle

Q3 is the quarter where Cloud transitions from "scaling story" to "moat established." The data:

• Backlog $155B — up $49B in a single quarter (+46% Q/Q, +82% Y/Y).
• $1B+ deals in 2025 YTD — more than the previous two years combined.
• New GCP customer growth — +34% Y/Y (accelerated from +28% Q/Q in Q2).
• Existing customers using AI products — over 70% of all Google Cloud customers.
• 13 product lines at $1B+ annual run-rate — diversification is real.
• Revenue from generative-AI-built products +200% Y/Y.
• ~150 customers each processing ~1T tokens over TTM — power-user concentration.
• 9 of top 10 AI labs choose Google Cloud — explicit competitive claim.
• Anthropic accessing up to 1M TPUs — the most consequential third-party silicon validation in the AI cycle.

"We are excited that Anthropic recently shared plans to access up to 1 million TPUs." — Sundar Pichai, CEO

The Anthropic disclosure is structurally important. Anthropic is one of two frontier AI labs (the other being OpenAI). Their decision to commit to ~1M TPUs is direct evidence that Google's custom silicon is competitively superior at scale for inference workloads — a verdict that NVIDIA's GPU monopoly cannot replicate. Combined with the SSI (Q2) and Physical Intelligence callouts, the pattern is clear: frontier AI labs are choosing TPUs, not just renting GPUs.

Cloud margin expanded to 23.7%, a +660bps Y/Y expansion. Operating income up 85% to $3.6B. Ashkenazi flagged the offset:

"The expansion in cloud operating margin was driven by strong revenue performance and continued efficiencies in our expense base, partially offset by higher technical infrastructure usage costs, which includes depreciation expense and other operations costs such as energy." — Anat Ashkenazi, CFO

Gemini Enterprise, launched earlier in October, has 2M+ subscribers across 700 companies "ahead of GA scale" — a faster monetization curve than anything previously launched in Cloud.

Assessment: This is the quarter where the multi-year Cloud thesis fully materializes. With $155B in backlog, $1B+ deals running 3x prior pace, and Anthropic publicly endorsing TPUs, Cloud is no longer competing for relevance — it's competing for share. Margin trajectory toward 25-30% is now squarely in our base case.

Other Bets / Waymo — 2026 Becomes the Inflection Year

Pichai disclosed Waymo's 2026 expansion roadmap: London (2026), Tokyo (working on it), plus U.S. cities Dallas, Nashville, Denver, Seattle. Approval to operate fully autonomously at SFO and SJC airports. Autonomous testing scaling in NYC. Waymo for Business (enterprise travel offering) launched. Teen accounts in Phoenix scaling.

"Waymo's growth and momentum are strong, and 2026 is shaping up to be an exciting year." — Sundar Pichai, CEO

Pichai also volunteered (in response to Brian Nowak's Waymo question) that Gemini integration with Waymo is on the roadmap for 2026 — multimodal in-car experiences, YouTube integration, profile-driven ride pre-scheduling. This is the first time management has framed Waymo as a Gemini-distribution surface.

Assessment: Waymo is moving from "operates in 4 cities" to "operates in 8+ U.S. cities plus London/Tokyo" in the next 12-18 months. Combined with Gemini integration optionality, this is increasingly material to Alphabet's medium-term value. We're still not modeling explicit Waymo upside but the option premium is climbing.

Key KPIs

Key Topics & Management Commentary

Overall Management Tone: The most confident tone of the year. Pichai opened with "this was a terrific quarter" and proceeded to deliver more product/distribution/capacity callouts per minute than any earnings call we've heard from the company in recent memory. The Anthropic-1M-TPUs line was clearly a deliberate disclosure — they wanted that headline. Schindler's Search monetization framing was direct and confident — "paid clicks +7%, CPCs +7%" is the kind of disclosure you make only when you want it to land. Ashkenazi was the most disciplined of the three on capex/depreciation framing but did not soften the 2026 step-up message.

Anthropic, TPUs, and the Frontier-Lab Endorsement

Pichai disclosed that Anthropic plans to access up to 1 million TPUs. This is unprecedented in the AI infrastructure landscape — frontier AI labs (Anthropic, OpenAI, xAI, SSI, Physical Intelligence, Mistral) are the most discriminating customers in compute, with the ability to choose any chip on any cloud. Their selection of TPUs at this scale is direct, public evidence that Google's silicon strategy is competitively dominant for inference at frontier-model scale.

The economic implication: Anthropic at 1M TPUs is a multi-billion-dollar multi-year contract, and it brings the Anthropic technology stack onto Google Cloud, which means Anthropic-built consumer/enterprise AI products run on Google infrastructure — a meta-distribution win.

Assessment: This is the most important announcement in the call. The TPU narrative was "Google's chips are good for Google's models." It's now "frontier AI labs choose Google's chips for their own models" — an entirely different competitive position.

AI Mode at 75M DAU and the Gen Z Pattern

AI Mode at 75M DAU after one full quarter of global rollout (40 languages) is a faster ramp than ChatGPT achieved at the equivalent stage. Pichai called out specifically that the AI features effect is "particularly encouraging to see this was more pronounced with younger people" — confirming the Q2 Gen Z theme.

AI Mode is "already driving total query growth for Search" — meaning the queries are net additive, not substitutive. This is the data point bears have been demanding for 18 months.

Assessment: The cannibalization debate is functionally resolved on a 9-12 month horizon. AI features are growing the query pie. The remaining question is monetization rate, and Q3's paid clicks +7%/CPCs +7% data answers that too.

The European Commission Fine

$3.5B charge in Q3 G&A from the EC's September 5 decision (adtech competition matter). Fully reflected in Services segment, dragging margin from a hypothetical ~42.5% to reported 38.5%. This is the second meaningful regulatory hit of 2025 (Q2 had a $1.4B legal settlement). The total ~$5B of regulatory charges YTD is a real but contained cost of doing business at Alphabet's scale.

What's notable is what management did NOT say: there was no commentary on the U.S. DOJ Search remedies process or Chrome divestment risk. That continues to be a tail risk that the company is electing not to address publicly on calls.

Assessment: The EC fine is a one-time hit absorbed by the operational strength of the quarter. The DOJ Search remedy is the more important regulatory variable — and management's silence is informative either way. We continue to underwrite some probability of a meaningful structural remedy.

Capex: $91-93B in 2025, "Significant Increase" in 2026

2025 capex raised again: $85B → $91-93B. Q3 alone was $24B. Mix: 60% servers, 40% data centers/networking. 2026 will see "a significant increase, and we'll provide more detail on our fourth-quarter earnings call."

The depreciation curve is fully visible now: Q1 +31%, Q2 +35%, Q3 +41%, with Q4 "accelerating slightly." That's a 10-point acceleration in growth rate over three quarters, on a base that grew $5.6B in Q3 alone. The annualized D&A is approaching $25B.

Assessment: The 2026 capex is now the central forecasting variable. With Q3's "significant increase" framing and $155B of backlog, we now model 2026 capex at $115-130B — meaningfully higher than our prior $95-110B range. The offsetting factor: Cloud margin expansion has consistently outrun depreciation drag, and 2026 should be the year Cloud runs at >$100B annualized at 25%+ margins.

U.S. Tax Law Tailwind

The July 4, 2025-enacted U.S. tax law allows immediate expensing of domestic R&D and accelerated depreciation on eligible capex. Q3 free cash flow benefited materially — $24.5B FCF in Q3 vs. $5.3B in Q2 — partly from these timing changes. This is a real and ongoing benefit, not a one-quarter item.

Assessment: Tax-law-driven FCF tailwind partially offsets the capex-driven FCF compression. Net effect: 2025 FCF likely ends in the $80-90B range, better than the $60-65B we modeled at Q2.

Guidance & Outlook

Updates from the call:

Analyst Q&A Highlights

Search Monetization & AI Mode

• Doug Anmuth (JPMorgan): Asked specifically about clicks per query, conversion rates, and pricing in AI-driven search formats. Schindler delivered the +7%/+7% paid clicks/CPCs disclosure — direct evidence of healthy monetization.
  Assessment: This is the data point bulls have been waiting for since AI Overviews launched. Three quarters into AI scale-up, the monetization metrics keep getting better.
• Brian Nowak (Morgan Stanley): Asked about agentic e-commerce and whether agentic monetization rate will be lower than current Search. Schindler framed it as additive — agentic experiences "help us answer people's tough questions ... and help businesses in the process."
  Assessment: Management is being careful not to commit to a monetization rate for agentic surfaces. We treat agentic monetization as a 2026-2027 question.

Cloud & Capex

• Multiple analysts asked about 2026 capex magnitude. Ashkenazi declined to quantify, deferring to Q4 call. The "significant" framing is itself a signal — typically used when the increase is >15-20%.
• One analyst pressed on whether the $155B backlog should change how investors think about future revenue visibility. Ashkenazi noted the backlog is heavily weighted toward GCP and reflects multi-year customer commitments.
  Assessment: The duration of the $155B backlog isn't disclosed but should now include some multi-year Anthropic commitment. Worth pressing for RPO duration disclosure in Q4.

Productivity & AI Internal Use

• Doug Anmuth (JPMorgan): Asked Ashkenazi about cost-side opportunities to absorb depreciation growth. She noted "nearly half of all code generated by AI" internally now (up from 30%+ at Q1) — an increasing productivity flywheel that helps offset operational expense growth.
  Assessment: AI-assisted coding is now a real efficiency lever. The 50% threshold is industry-leading.

Waymo

• Brian Nowak (Morgan Stanley): Asked when Waymo will integrate with Gemini and user data. Pichai confirmed he's reviewing this in coming weeks and committed to "newer experiences in 2026."
  Assessment: Waymo + Gemini integration is the platform play. The 2026 timeline matches Waymo's city expansion timeline and creates a natural content-and-commerce surface inside the vehicle.

What They're NOT Saying

1. 2026 capex magnitude: Still unquantified. The "significant" framing implies >15-20% step-up, putting it at $105-115B+ minimum. Q4 will be the reveal.
2. RPO duration on the $155B backlog: The number is disclosed but the recognition profile (how much in 12 months vs. 24+ months) isn't. AWS and Azure both disclose this; Alphabet's silence is conservative.
3. Anthropic contract details: Pichai disclosed the 1M TPU access plan, but contract value, duration, exclusivity, and economic structure are not. Anthropic has historically used both AWS and GCP; the new disclosure is a meaningful TPU deepening but not necessarily exclusive.
4. DOJ Search remedy / Chrome: Zero call mentions for the third consecutive quarter. Management is electing not to address. Tail risk persists.
5. Specific Cloud customer concentration: With $1B+ deals running at 3x prior pace, top-customer concentration is rising. Not disclosed.
6. AI Mode / Gemini app monetization roadmap: Schindler said "richer placements" will come for AI Mode and AI Overviews "over time." No specific timeline.
7. Buyback pace: Q3 repurchases were $11.5B vs. $13.6B in Q2 — a slowdown. With $98.5B in cash and continued capex needs, buyback pace is being modulated by capital allocation discipline.

Market Reaction

• After-hours move: Modestly positive in extended trading.
• Next-day reaction: Stock traded up materially as the print landed. Analyst price target raises followed broadly.
• Sentiment summary: The print was widely received as a thesis-validating quarter — first $100B+ topline, $155B Cloud backlog, Anthropic TPU win, paid clicks/CPC data, 2026 expansion roadmap. Capex raise was framed as "investing into demand," not capex panic.

The market reaction is rational given the data. The print resolves several long-running debates favorably, and the only meaningful overhang (2026 capex) is a quantification issue, not a directional one.

Street Perspective

Debate: Is Search structurally re-rated, or is this a temporary AI-driven sugar high?

Bull view: Three consecutive quarters of accelerating Search growth (10% → 12% → 15%) with paid clicks AND CPCs both rising 7% Y/Y. AI Overviews and AI Mode are net-additive to query volume, not substitutive. The structural re-rating is real and durable.

Bear view: Search growth benefits from year-over-year favorable comps (lapping the AI-fear-driven 2024 sentiment) and from product-launch cycles. Once AI Mode reaches AI Overviews scale and replaces classic results in more sessions, the monetization rate will compress.

Our take: Bulls have the better evidence. The +7%/+7% paid clicks/CPC data is hard to reconcile with the bear thesis. We update our base case to assume Search durably grows in the low double digits through 2026.

Debate: Is the 2026 capex "good news" or "bad news"?

Bull view: Capex follows demand. $155B of backlog with $49B added in a single quarter justifies any reasonable capex number. Cloud margin expansion is keeping pace with depreciation, so the FCF profile is still healthy.

Bear view: The capex curve looks structural — $50B → $75B → $91-93B → "significant increase" toward $115B+ in 2026. Eventually the AI demand cycle moderates, and the depreciation curve will outrun revenue. The FCF compression in Q2 was a preview.

Our take: Both partly right. We model 2026 capex at $115-130B. The risk is asymmetric — if demand sustains, the capex pays off in 2027-2028 EBIT growth; if demand softens in late 2026, the depreciation drag is structural for 2-3 years. We accept the asymmetric risk because the demand signal in the Q3 backlog is unusually strong.

Debate: Has the AI-vs-Search debate been resolved?

Bull view: Three quarters of data. AI Overviews 2B MAU. AI Mode 75M DAU. Search +15%, paid clicks +7%, CPCs +7%. Debate over.

Bear view: AI Mode is still under-monetized. Gemini app monetization is years away. The real cannibalization risk shifts in 2026-2027 as those surfaces reach scale.

Our take: The near-term debate (2025-2026) is resolved in bulls' favor. The medium-term debate (2027+) remains open. We're comfortable being long through that horizon and re-evaluating as AI Mode monetization data emerges.

Model Update Needed

Valuation impact: Materially positive. Cloud's path to $100B+ revenue at 25%+ margins is increasingly de-risked. Search resilience supports premium multiple on the Services franchise. The 2026 capex remains the swing factor but is increasingly justified by visible demand.

Thesis Scorecard Post-Earnings

Overall: Thesis materially strengthened across every bull point; the bear case has effectively no remaining near-term data support outside the capex/depreciation arithmetic.

Action: Maintain Outperform. This is the most operationally complete print of the AI cycle so far.