Decoding Google: AI Overviews Is Saving Search, $462B Cloud Backlog, and the $40B Anthropic Bet

OurAlpha Scorecard

Composite score 6.9/10 — fundamentals and the "full-stack AI" narrative are compelling, but the stock is no longer in deep-value territory. The setup is about narrative execution pace, not margin of safety.

Q1 2026 Earnings: 22% Growth + Record Margins, AI Bear Case Dismantled Point by Point

Google Q1 2026 (quarter ended 3/31/2026) results^{[Alphabet Q1 2026 8-K · SEC]}:

Two diverging narratives:

• Profitability — The 36.1% operating margin is an all-time high on a publicly comparable basis. Three forces combined to drive it higher: declining AI inference costs (per media accounts of management commentary, AI serving costs fell further after the Gemini 3 upgrade, though Google did not disclose specific figures in its earnings tables^{[Search Engine Journal]}), accelerating Search ads, and growth in higher-margin YouTube ads.
• Cash flow — TTM FCF of $64.4B is down 14% YoY, but this is not primarily an OCF problem — OCF is still growing in the double digits. The culprit is capex TTM growth far outpacing OCF. The capex-OCF scissors gap is unlikely to close near-term.

Market reaction: GOOGL rallied post-earnings (vs. MSFT falling 4–5% after announcing $190B capex in the same quarter)^[heygotrade]. Both raised capex — so why did the market reward GOOGL but punish MSFT? Because GOOGL paired its capex with a ROIC narrative via Cloud +63%, while MSFT had only Azure +40% (itself a sustained acceleration) but lacked the visual punch of a record-shattering beat.

II. Search Accelerates to +19% for Five Straight Quarters: AI Cannibalization Is Actually Reigniting Search

This is the most consensus-defying part of this earnings report.

Historical growth rates are derived from Google Services – Search & other figures disclosed in Alphabet's quarterly earnings releases^{[Alphabet IR · Historical Earnings Releases]}.

Five consecutive quarters of acceleration, with the pace widening each quarter. On the earnings call, Pichai attributed the trend to AI experiences — AI Overviews and AI Mode — driving usage and query growth^{[Sundar Pichai · Q1 2026 Earnings Remarks]}. Management also stated that Q1 Search queries hit an all-time high.

That said, two layers of nuance are worth keeping in mind:

1. Company narrative — Pichai credits the acceleration to AI experiences. This is the official framing from the Q1 2026 earnings call.
2. Third-party data — SEO industry research shows^{[QuickSEO 2026 Data]}: AI Overviews appear in roughly 20% of search results (up to 37.2% in some markets); when AI Overviews are present, CTR on top organic links falls by approximately 34.5%–64.4%; and 93% of AI Mode sessions generate no outbound clicks.

These two sets of data are not contradictory — they are two sides of the same story:

• Third-party SEO data captures pressure at the publisher / organic traffic level — the CTR decline is real, but it does not directly translate to monetization pressure on Google Search ad revenue;
• Even as AI Overviews compress some outbound clicks, total query volume growth + ad product adjustments + retention of commercial queries have collectively offset that pressure (the all-time high query claim comes from Pichai on the earnings call);
• A useful analytical framework — Search ad revenue ≈ per-query monetization × total query volume — can guide the analysis, but Google has not disclosed per-query monetization or the standalone contribution of AI Overviews / AI Mode to Search revenue; "slight price compression × large volume gain" is an analytical framework, not a factual conclusion.

This is why the market has been willing to re-rate GOOGL. But whether this equilibrium holds depends on how many categories can continue supporting paid ad placements alongside AI-generated answers. Google is already testing ad units within AI Overviews and AI Mode^{[Search Engine Journal]} — a deliberate effort to offset CTR losses through monetization. Results look promising so far, but this has only run for five quarters and has yet to be tested through a full ad cycle.

III. AI Overviews + AI Mode: Two Different Stories

These two products are often lumped together, but they are two distinct curves:

Note: The user figures above are primarily from 2025 reporting. Q1 2026 earnings did not include a standalone breakdown for AI Overviews or AI Mode — which is itself part of the disclosure gap.

The two products are at very different stages of monetization:

• AI Overviews is already running ads pilots — ads have appeared on U.S. mobile, but Google has not disclosed its standalone contribution to Search revenue. It is better understood as an explanatory variable for usage/query growth rather than a primary monetization engine.
• AI Mode is still in testing — third-party observations suggest ads appear in roughly 25% of AI Mode results^{[DigitalApplied]}, but Google has not publicly disclosed AI Mode's standalone monetization rate or CTR.

This is the single largest disclosure gap in the GOOGL bull/bear debate — the company has offered two data points (all-time high total query volume, Search revenue +19%), but has not provided a monetization breakdown between AI Overviews vs. traditional Search, or AI Mode vs. Overviews. The result:

• Bulls can say "AI is the growth engine" — and they'd be right;
• Bears can say "AI is diluting monetization density" — and they could also be right;
• Neither side can prove their case with company-disclosed data. The real test comes when AI Mode's user base is large enough that Google has no choice but to report it separately.

IV. Google Cloud +63% and $460B+ Backlog: The Bottleneck Is Supply, Not Demand

Source: Alphabet historical earnings releases / Statista^{[Alphabet IR]}. Note: Google Cloud revenue includes GCP + Workspace + other enterprise services; the Q1 2026 acceleration to +63% was driven by enterprise AI solutions and infrastructure within GCP.

What a $460B+ Backlog Actually Means:

Alphabet's official Q1 release stated that "Google Cloud backlog nearly doubling quarter on quarter to over $460 billion"^[TechCrunch]. Against an estimated Q4 2025 backlog of ~$240B, that implies ~$220B+ of net new bookings in a single quarter — one of the largest single-quarter backlog jumps in cloud history. A few definitional caveats apply^[Tikr]:

• The majority reflects standard GCP contracts;
• TPU hardware sales contracts are included — a scope difference versus AWS and Azure that needs to be broken out for apples-to-apples comparisons;
• The company expects just over 50% of the backlog to be recognized as revenue within the next 24 months.

Why This Is a Supply Constraint, Not a Demand Problem

Pichai was explicit on the earnings call: "Our cloud revenue would have been higher if we had more capacity"^[TechCrunch]. This marks the first time in four years that Google Cloud has officially identified capacity — not demand — as the ceiling on growth.

That distinction is central to the re-rating thesis. The old framework was "Google Cloud is a permanent third-place also-ran." Q1 data dismantles it: Google Cloud +63% vs. Azure +40% vs. AWS +28% (AWS figures pending second-round verification)^{[Mirror Review]} — Google Cloud is now the fastest-growing of the three hyperscalers, and the gap is widening.

V. Anthropic's $40B Closed Loop: Equity + 5GW Compute + 1M+ Ironwood = Full-Stack AI Finally Proven End-to-End

On 2026/4/24 (five days before Q1 earnings), Google and Anthropic jointly announced the deepest industrial tie-up in Mag 7 history^[CNBC]:

What is the Ironwood chip: Google's seventh-generation TPU, GA'd in 2025/11, delivering 4.6 PetaFLOPS FP8 / 192GB HBM3e / 7.4 TB/s bandwidth per chip^{[ServeTheHome]}. Gemini 3 / Gemini 3.1 Pro are already running production inference on Ironwood.

Why this closed loop matters:

• Google is simultaneously writing Anthropic an equity check, selling Anthropic compute through GCP, and selling Anthropic chips through Ironwood — a triple payoff of investment returns + cloud revenue + internal chip demand validation.
• This is the first Mag 7 example where "full-stack AI" has actually been made to work end-to-end: model (Gemini), silicon (TPU), cloud (GCP), and model customers (Anthropic + its own Gemini products) — all four layers self-reinforcing.
• Compared with Microsoft's restructured OpenAI relationship (27% economic interest, exclusivity downgraded to non-exclusive), Google's Anthropic relationship is deeper and more durable — Anthropic is more tightly locked into Google's compute than OpenAI is into Azure (5 GW + 1M chips is physical hard-binding; OpenAI has since gone multi-cloud).

Counterbalancing constraints (must be read alongside the bull case):

• The $30B tranche of the $40B is milestone-based — it has not been spent; it only gets triggered if Anthropic hits specified performance targets.
• Separately from Google, Anthropic has already signed a large Trainium deal with Amazon AWS. This means the compute lock-in is non-exclusive — Anthropic runs a multi-cloud, multi-silicon stack.
• Claude Enterprise competes directly with Gemini Enterprise^{[The Next Web]} — Google is simultaneously Anthropic's investor, supplier, and commercial rival.
• Circular financing concern: The "investment → cloud contract → valuation uplift → long-term compute commitment" loop between Big Tech and AI startups is self-reinforcing across the board (Google → Anthropic → GCP/TPU; MSFT → OpenAI → Azure; AMZN → Anthropic → AWS Trainium). Markets may eventually demand a clearer distinction between genuine external demand and intra-ecosystem demand — otherwise both backlog quality and capex justification face a credibility discount.

This is the deepest vertical integration story among all current Mag 7 names, but it should not be read as "Anthropic is a Google subsidiary", nor should ecosystem-internal demand be conflated with external market demand.

VI. Three-Front Antitrust Battle: Mehta Ruling + Dual Appeals + Ad Tech Retrial

Search Antitrust Timeline^{[CRS Legal Sidebar · LSB11362]}:

Why the market hasn't cracked:

• Judge Mehta's September 2025 ruling landed on behavioral remedies rather than structural breakup — the single biggest reason GOOGL rallied ~8% that day^[NPR].
• The data-sharing provisions came in narrower than feared — the market had priced in a potential requirement to share ranking signals; the final order covers only partial sharing of search index and user-interaction data.
• Apple TAC is no longer "exclusive," but Apple can still choose Google as its default search engine — it simply cannot sign an exclusivity contract to that effect.

Key uncertainties over the next 18 months:

• If the DOJ appeal succeeds, Chrome divestiture comes back on the table — but even then, executing a structural breakup would likely take another 12–24 months.
• The Ad Tech case — a separate antitrust track on which Google already received an adverse ruling covering its ad server and ad exchange businesses (2025) — still has remedy hearings underway, with a ruling possible within 2026. This is an independent overhang the market has not fully priced in.

OurAlpha take: Antitrust is not a binary blow-up risk for GOOGL's valuation — it is a long-tail overhang that will intermittently pressure the stock as the appeals process unfolds, but the probability of a core structural breakup (Chrome / Android) before 2027 remains low.

七、Waymo：从 250K 到 1M rides/week 的隐性估值

Waymo's revenue is not broken out separately in Q1 earnings (still reported under Other Bets), but its operating metrics have entered a new order of magnitude. Pichai explicitly disclosed in the Q1 2026 earnings release that Waymo has surpassed 500,000 fully autonomous rides per week^{[Sundar Pichai · Q1 2026 Earnings Remarks]}.

City expansion^{[TechCrunch · Feb 24, 2026]}:

Waymo is now offering or testing driverless rides across multiple cities. Phoenix, San Francisco Bay Area, Los Angeles, and Austin are its core commercialized markets; Houston / Dallas / San Antonio / Orlando are in select-rider rollout; Atlanta / Miami / Washington D.C. and others are advancing via partnerships or planned expansion. City count should be broken down by status: full commercial ops, select riders, testing, and coming soon. On the manufacturing side, Magna's Arizona facility has scaled to 2,000+ vehicles/year (including the joint factory with Waymo).

Contrast with TSLA Robotaxi's 25-vehicle fleet:

OurAlpha's previous TSLA deep dive^{[OurAlpha · TSLA Deep Dive]} put Austin's active fleet at ~25 vehicles. Waymo's 500K rides/week represents a two-order-of-magnitude gap—and Waymo operates fully driverless, with no safety driver. This is something everyone in the industry knows, yet it remains largely absent from GOOGL's valuation.

Illustrative valuation contribution scenarios (Type D — not a transaction valuation; not investment advice):

• 500K rides/week × 52 weeks × assumed $20/ride = $520M/year in cash operating revenue
• If 1M rides/week is achieved at $25/ride = $1.3B/year
• At a 25x SaaS multiple → $32B implied valuation (OurAlpha simplified derivation)
• In practice, Waymo would not be valued on a SaaS P/E basis — it commands a premium as the only L4 robotaxi operator at scale. OurAlpha's scenario range is ~$80–150B — this is not a current funding-round valuation, but rather a common reference point across third-party and investor models; a gap exists between public funding rounds and this estimate.

In most GOOGL sum-of-parts models, Waymo is ascribed only a low single-digit percentage of total enterprise value — making it a potential re-rating catalyst, contingent on Waymo actually pushing toward 1M rides/week and executing its tiered city expansion.

VIII. Valuation & Consensus: $402 Current Price, $408 Consensus, Forward P/E 28x

As of ~5/15/2026^{[Stock Analysis][MarketBeat]}:

Note: Forward P/E multiples vary across sources depending on which EPS year is used (FY26 vs. FY27 forward) and the reference share price; figures above reflect Trefis as of 5/14/2026.

A few nuances worth unpacking:

1. Median vs. mean price target — The gap between the $408 and $427 figures (35 vs. 52 analysts) is largely driven by the higher price targets issued more recently (May), pulling the broader average up.
2. The outliers — DA Davidson's $310 bull case rests on an antitrust + capex double-hit thesis; Citizens' $515 reflects a Cloud + Waymo re-rating story. Both are coherent framings. At $397–402, the market is pricing in a base case and not fully pricing in either tail.
3. What the 28x forward P/E actually tells you — If FY26 EPS lands at $14.22 and FY27 consensus holds at +12–15% growth (the last year before capex depreciation meaningfully kicks in), then $397–415 implies a forward P/E of 26–28x. Whether that's expensive comes down to your view on how the capex cycle will affect future ROIC.

IX. The Next 3 Quarters: 5 Hard Numbers That Could Reprice the Stock

Additional watch items (Scenario D):

• DOJ appellate argument scheduling——if oral argument dates are set in H2 2026, expect a fresh round of overhang repricing;
• Ad tech remedy hearing ruling——a separate overhang that could resolve within 2026;
• Waymo 1M rides/week milestone——if hit ahead of schedule in Q3, it would trigger a sum-of-parts revaluation.

十、反向 setup：什么情况会让 thesis 翻车

OurAlpha's "bear case" checklist — scenarios that would break the above narrative (Category D hypotheticals, not forecasts):

• Search growth re-accelerates to +19% then decelerates to +12%: signals AI Mode is driving zero-click acceleration faster than ads monetization can keep up — the "AI re-accelerates Search" narrative takes a haircut
• Cloud snaps back to ~+40% growth shortly after capacity expansion — if backlog conversion also slows, the market will question whether Q1's +63% was driven by one-time mega-deals or TPU contracts, undermining the "capacity-constrained" narrative
• Cloud backlog reveals a single oversized contract broken out in disclosure — if one customer accounts for a disproportionate share of the $460B+, the quality of that backlog gets repriced
• DOJ appellate court overturns Mehta's "behavioral remedy" ruling, reintroducing the Chrome divestiture option — this is a binary risk, low probability but high payoff if it materializes
• AI Mode monetization data leaks and disappoints — if analysts independently model AI Mode CTR well below traditional Search, market confidence in Search's long-term monetization path erodes
• Capex guidance raised to $200B+ while FY27 EPS consensus is cut — implies the market believes ROIC realization on this capex cycle is slower than expected, calling into question the forward P/E of 28x
• Anthropic misses milestones and the follow-on $30B investment is deferred — would dent the "full-stack AI flywheel" narrative, though this is a long-tail risk

十一、Back to Basics: What Buying GOOGL Now Is a Bet On

Simply put, buying GOOGL today is a bet on 3 things holding true simultaneously:

1. AI Overviews / AI Mode won't erode Search monetization density more aggressively than expected——meaning the "AI as a Search accelerant" narrative continues to be validated over the next 4–6 quarters;
2. Google Cloud's 60%+ hypergrowth is "capacity-constrained" rather than "peak demand"——meaning post-Q2/Q3 capacity ramp, growth stabilizes above +50% and the 24-month backlog-to-revenue conversion cadence holds;
3. Antitrust doesn't produce a structural breakup before 2027——meaning Chrome and Android remain inside Alphabet, and renegotiating Apple TAC raises the Search revenue-share payout rather than flipping the default search arrangement.

If all 3 hold, a 28x forward P/E is defensible——because it reprices GOOGL from "Search cash cow" to a compound of "full-stack AI + third Cloud pillar + embedded Waymo optionality."

If one breaks, the 30–35% valuation premium gets roughly half recaptured (forward P/E reverts to 22–24x ≈ stock price $315–340).

If two or more break, valuation falls back below the 5-year historical mean of 21x (forward P/E 18–20x ≈ stock price $260–290).

The outcome hinges on the pace at which all 3 materialize — not on any single quarter's earnings print.

About Data Classification

All specific figures cited in this article are tiered across the following 4 levels of reliability:

• Tier A: Audited / Official Disclosures——Alphabet Q1 2026 8-K (SEC filing) / Q1 2026 earnings release / historical 10-Ks / court judgment documents. Examples: Q1 revenue $109.9B, operating margin 36.1%, Google Services $89.6B, Search $60.4B, YouTube ads $9.88B, Google Cloud $20.0B, Cloud backlog over $460B, capex Q1 $35.7B, FCF TTM $64.4B.
• Tier B: Management Commentary——Statements by Pichai / Anat Ashkenazi on the Q1 2026 earnings call and in the earnings release. Examples: Cloud is capacity-constrained, AI experiences driving query volume to all-time highs, FY 2027 capex will increase significantly, Cloud backlog "slightly over 50% converts within 24 months," Waymo surpassing 500K rides/week.
• Tier C: Mainstream Media / Industry Trackers——CNBC, Reuters, TechCrunch, Search Engine Journal, Search Engine Land, NPR, CRS Report, etc. Examples: AI Overviews reaching 2B MAU by mid-2025, AI Mode reported at the 100M-user scale in 2025, Anthropic valuation ~$350B / Anthropic deal up to $40B, reportedly up to 1M Ironwood TPUs, DOJ + multi-state appeal timelines, SEO-estimated CTR decline data.
• Tier D: Scenario Assumptions / Estimates——OurAlpha's $80–150B Waymo valuation range (not transaction-based), forward P/E scenarios implying specific price ranges, the "break level 1 / break level 2" bear-case framework implying -15%–-30% price moves, the "slight unit price decline × large volume increase" analytical framework. These represent editorial judgment and scenario analysis, not investment advice, not statements of fact.

Key caveats:

• The 2B MAU figure for AI Overviews and the 100M user figure for AI Mode are primarily from 2025 reporting——Alphabet's Q1 2026 earnings did not break out these products separately. Any reference to "new Q1 2026 evidence" should point to the +19% Search revenue growth figure itself, not user counts.
• Google has never separately disclosed CTR, ARPU, or ad load data for AI Overviews / AI Mode——All third-party SEO estimates (CTR decline of 34.5–64.4%, AI Mode ads in ~25% of results, Seer Interactive organic CTR dropping from 1.76% to 0.61%, etc.) primarily reflect pressure at the publisher / organic traffic level and should not be directly equated with monetization pressure on Google Search advertising revenue, nor treated as official Google data.
• The $460B+ Cloud backlog includes TPU hardware sales contracts——this is a scope difference vs. AWS / Azure; TPU contracts should be called out separately when making cross-company backlog comparisons.
• Of the up to $40B Anthropic deal, $30B is contingent on milestones——this is not cash already spent; the 5GW TPU / 1M Ironwood chip figures are media-sourced and do not appear as line items in Alphabet's Q1 release, nor should they be read as completed deliveries.
• Circular financing risk——The investments Big Tech is making in AI startups (Google → Anthropic, MSFT → OpenAI, AMZN → Anthropic), cloud revenue, and long-term compute commitments are mutually reinforcing. Markets may eventually demand clearer separation between genuine external demand and intra-ecosystem demand, which could affect the valuation premium assigned to backlog and capex figures.
• All paragraphs labeled "OurAlpha perspective / view / scenario estimate" = editorial judgment, not investment advice.
• Valuation data reflects closing prices around 2026/5/15; market prices are subject to change.

This content is for informational purposes only and does not constitute investment advice, trading advice, or any guarantee of returns.