Alphabet

Phase A — Understand the business

Lens 1 · Company Overview

Alphabet is the holding company for Google plus a thin sleeve of "Other Bets." Strip the structure away and it is one of the great advertising-and-distribution machines in history, now bolting a hyperscale AI-cloud business onto the side. FY2025 total revenue was $402.8B, up 15% from $350.0B. The business splits three ways:

• Google Services — $342.7B (85% of revenue), $139.4B operating income. Inside it: Google Search & other $224.5B, YouTube ads $40.4B, Google Network $29.8B (the only declining line), with the rest "subscriptions, platforms, and devices" (YouTube Premium/TV, Google One, Pixel, Play). Total Google advertising = $294.7B. This is the cash engine — a 40.7% operating margin.
• Google Cloud — $58.7B (15%), $13.9B operating income. GCP infrastructure + Workspace + the Vertex/Gemini enterprise AI platform. The growth story (see Lens 4).
• Other Bets — $1.5B revenue, −$7.5B operating loss: Waymo (the only one that matters), Verily, plus a graveyard of moonshots. A rounding error on revenue, a real drag on consolidated profit.

Customers/structure: the ad business is a two-sided auction — billions of free users on one side, millions of advertisers (no meaningful concentration — no single advertiser is material) on the other. Revenue is overwhelmingly performance-based auction spend, not recurring contracts; it flexes with the macro ad cycle. Cloud is the opposite — increasingly multi-year committed contracts, with a reported backlog of ~$155B at end-Sept 2025, +46% YoY. Suppliers: the critical input is compute — Alphabet is unusual in that it largely is its own supplier, designing its TPU accelerators (now a genuine competitive asset, not a science project) alongside large Nvidia GPU purchases. Geographic mix: US 48%, EMEA 29%, APAC 17%, Other Americas 6%. Dual-class structure (GOOGL Class A 1 vote, GOOG Class C 0 votes, Class B 10 votes held by founders) means Page & Brin retain voting control despite small economic stakes.

Lens 2 · Supply Chain

Map the chain, name the nodes:

Upstream (inputs) → Alphabet:

• Silicon: Nvidia (GPUs) and Broadcom (co-designs the TPU; the TPU is fabbed at TSMC). The TSMC dependency is the single most important external chokepoint — every frontier accelerator on earth, Alphabet's TPUs included, funnels through TSMC's leading-edge nodes and CoWoS advanced packaging. HBM memory from SK Hynix / Samsung / Micron is the second chokepoint (see kb/ai/wiki/bottlenecks.md).
• Power & data centers: the new binding constraint. The $91B capex (Lens 5) is land, shells, transformers, and electricity. Alphabet is signing nuclear (SMR) and large PPAs to secure power — the bottleneck has migrated from chips to grid interconnection and megawatts.
• Network: owns subsea cables and a private global backbone — a real structural advantage vs. cloud peers who rent more.
• Devices: Pixel hardware assembled through Asian ODMs/contract manufacturers; immaterial to the financials.

Alphabet → end customer:

• Ads side: advertisers and agencies buy through Google Ads / DV360 / Ad Manager; distribution to users runs through owned-and-operated surfaces (Search, YouTube, Gmail, Maps, Android, Chrome) plus the Network (third-party publisher inventory). The owned distribution is the moat; the Network is the commoditizing edge (and the antitrust target).
• Distribution deals: the Apple default-search agreement (~$20B/yr range, paid to be Safari's default) is the most important single contract — and the one the Search remedy just reshaped (Lens 10). Android OEMs (Samsung et al.) are the other distribution rail.
• Cloud side: sells directly and via channel partners to enterprises, governments, and — increasingly — AI labs renting TPU capacity (Anthropic is a marquee TPU customer; Alphabet is also an investor).

Single-source / chokepoint flags: TSMC leading-edge (shared industry risk), HBM supply, grid power, and on the demand side the Apple distribution relationship — losing default status outright would be a genuine traffic shock, though the Dec-2025 remedy stopped short of that.

Lens 3 · Competitive Advantages (moats)

Alphabet has arguably the widest moat in megacap tech, and the AI panic of 2024 caused the market to badly misprice it. The durable advantages, ranked:

1. Distribution at planetary scale. Six products with 2B+ users (Search, YouTube, Android, Chrome, Maps, Gmail). New capability (Gemini) ships to that installed base on day one — 750M+ Gemini users was reached largely by default-surface distribution, not standalone app downloads. No AI startup can replicate the distribution; that is the whole bear-case rebuttal in one line.
2. The full AI stack, owned end to end. Alphabet is the only company that owns all four layers: its own silicon (TPU), its own data-center/network fabric, a frontier model (Gemini 3), and the distribution surfaces to monetize it. Microsoft rents OpenAI's models; Meta lacks the cloud; Amazon lacks the frontier model and consumer surfaces. This vertical integration is the structural bull case.
3. Data + feedback loops. Two decades of query/click/maps/YouTube-engagement data — a moat for ad targeting and a non-trivial advantage for model post-training and grounding.
4. Switching costs in Cloud + Workspace, and increasingly in the TPU ecosystem (JAX/XLA lock-in for labs that build on it).
5. Brand & habit. "Google it" is a verb. Costs nothing to defend.

Bargaining power: near-total over advertisers (auction, no concentration) and over Android OEMs. Weaker against Apple (Apple controls the iOS distribution Alphabet pays to access) and against TSMC (a supplier with pricing power the whole industry depends on). Net: Alphabet needs Apple and TSMC more than the reverse — those are the two relationships where its power is genuinely limited. Source: kb/ai/wiki/positioning.md, kb/ai/wiki/bottlenecks.md.

Lens 4 · Segments

All figures, FY2024 → FY2025:

Within Services advertising (FY2024 → FY2025): Search & other $198.1B → $224.5B (+13%); YouTube ads $36.1B → $40.4B (+12%); Google Network $30.4B → $29.8B (−2%, decelerating/declining).

The story is the Cloud margin inflection. Cloud operating margin went from 14.1% (FY24) to 23.7% (FY25) as scale absorbed the data-center fixed cost. The Q1-2026 run-rate is even sharper: Cloud revenue $20.0B (+63% YoY) with operating income $6.6B, a ~33% margin. That is the single most important trend line in the company — it converts the AI capex from a cost story into a margin-expansion story. Search is accelerating modestly (not the AI-disruption death-spiral bears feared); Network is the one genuinely weak line; Other Bets losses are widening (Waymo's scaling cost).

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest print — 2026-Q1)

The most recent filing on disk is Q1 FY2026 (quarter ended 2026-03-31):

• Revenue $109.9B, +22% YoY (vs $90.2B). Acceleration vs. the FY2025 full-year +15%.
• Income from operations $39.7B (+30%); operating margin 36% vs 34% a year ago — margin still expanding even through the capex ramp.
• Net income $62.6B (+81%); diluted EPS $5.11 vs $2.81 (+82%) — but read this carefully. The headline is flattered by Other income (expense), net, dominated by gains on equity securities (mark-ups on private stakes — SpaceX, Anthropic, et al.). The operating line (+30%) is the clean read; the +81% net is not a sustainable earnings rate. (Same dynamic inflated FY2025 — see Lens 10.)
• Segment drivers: Google Cloud $20.0B (+63%) was the standout; Search & other $60.4B (+19%); YouTube ads $9.9B (+11%).
• Balance-sheet / cash-flow flag: capex was $35.7B in the single quarter, more than double the $17.2B a year prior. Annualize and you are north of $140B — the buildout is the dominant capital-allocation fact of the company now.
• Market reaction context: the comparable Q4 (Feb-2026) print is what triggered the year's worst single-day move — −8.7% on a Cloud revenue miss, a reminder that Cloud is now the swing factor the tape trades on.

Unusual vs. its own history: the net-income-to-operating-income gap (huge OI&E) and the capex-to-revenue ratio (22.7% in FY2025 vs a historical low-teens) are both at multi-year extremes. Neither is hidden, but both change the quality of the earnings.

Lens 6 · Earnings Calls (sentiment trend)

Transcripts were not on disk (transcripts/ empty), so this is reconstructed from the filings' MD&A tone and `` coverage of the last four calls. The arc over FY2025→Q1-2026:

• Tone shift: from defensive to assertive. Through 2024 the recurring management job was reassuring the Street that AI would not cannibalize Search. By the FY2025/Q1-2026 calls the framing flipped to "AI is expanding the business" — AI Overviews/AI Mode lifting query volume rather than eroding it, Gemini as a growth surface.
• What they now say constantly: "full-stack," "TPU," "Gemini," "AI Overviews," "capacity-constrained in Cloud" (demand > supply is the boast). The "capacity-constrained" phrasing is the tell that justifies the $90B+ capex.
• What they stopped saying: the apologetic "we are an AI-first company and Search is safe" hedging of 2023–24. Confidence is the dominant register.
• Capital-allocation tone: explicitly defending the capex as ROI-positive and pointing at the Cloud backlog as proof of demand.

Lens 7 · Comps

Peer table — multiples are `` (mixed sources, mid-June 2026) or n/a. Market caps approximate.

Sources: GOOGL price $370.84, cap ~$4.52T, trailing P/E 28.47; GOOGL fwd P/E ~26; MSFT 23.3/21.5, AMZN 29.4/31.2; META fwd ~17–18.

Read: Alphabet is the most expensive of the four ad/cloud megacaps on forward earnings (~26x vs Meta ~17x, Microsoft ~22x) after its +65% 2025 run. The cheap-AI-loser trade is over; you are now paying a premium for the full-stack thesis. Meta is the value alternative inside the ad complex; Microsoft is the cheaper cloud-AI exposure. EV/EBITDA and ROE multiples were not cleanly sourceable — flagged rather than fabricated.

Lens 8 · Stock-Price Catalysts (moves >5%, last ~5y, pattern)

Mostly ``:

• 2025: +~65% for the year — best-performing Magnificent-7 name. Drivers stacked: (1) Sept-2025 Search-remedy relief — stock +8% the day the judge declined to force a Chrome/Android divestiture; (2) the Cloud-margin + backlog inflection; (3) Gemini 3 closing the model-quality gap; (4) TPU credibility as an Nvidia alternative; (5) a Berkshire Hathaway stake (17.8M shares disclosed Q3-2025) as a sentiment stamp.
• Worst move: −8.7% on the Q4 (Feb-2026) print, a Cloud revenue miss.
• ATH $402.38 on 2026-05-13; trading ~$371 mid-June, ~8% off the high.

Pattern: for most of its life GOOGL traded on Search ad growth and macro. The reaction function has shifted — the market now trades it on (a) Cloud growth/margin (the −8.7% miss, the rally on backlog) and (b) antitrust headline risk. Search resilience is increasingly assumed; Cloud execution and the regulatory tape are the live swing factors.

---

Phase C — Judge people & books

Lens 9 · Management

• Sundar Pichai (CEO, Google 2015→, Alphabet 2019→). ~10 years at the top. Track record: mobile transition, the rise of Cloud from also-ran to a credible #3 with expanding margins, and — the real test — navigating the AI threat without losing Search, which by FY2025 looks like a pass. Style: deliberate, consensus-driven, criticized in 2023 for being slow to ship against OpenAI; vindicated by 2025's execution. Skin in the game: professional manager, not a founder — modest economic ownership.
• Founders Larry Page & Sergey Brin retain voting control via Class B super-voting shares despite small economic stakes. Brin in particular re-engaged hands-on with Gemini. This is a founder-controlled company run by a professional CEO — implication: strategic patience and willingness to spend through a cycle (the $90B capex is a founder-conviction bet, hard for activists to block).
• Anat Ashkenazi (CFO, from late 2024) — ex-Eli Lilly; brought a sharper capital-discipline narrative and oversaw the dividend initiation.
• Capital-allocation history: historically excellent — high-ROIC core, disciplined-ish M&A (Wiz was the big recent deal), and enormous buybacks: $61.5B / $62.2B / $45.7B across FY2023/24/25, plus the first-ever dividend initiated in 2024 ($0.60→$0.83/sh, ~$10.2B paid FY2025). The pivot in FY2025: buybacks stepped down as capex stepped up — capital is being redirected from returning cash to building AI infrastructure. Whether that earns its cost of capital is the central debate (Lens 12).
• Red flags: Other Bets is two decades of value destruction (Waymo aside); the equity-securities gains flattering reported earnings (Lens 10) are a management-presentation issue to watch, not fraud.

Lens 10 · Forensic Red Flags

Filing-grounded; every figure labeled.

Accounting-quality flags (real but not alarming):

1. Earnings flattered by equity-security marks — the biggest one. FY2025 Other income (expense), net was $29.8B, +301% YoY, of which $24.1B was gains on equity securities. These are largely unrealized mark-ups on private stakes (SpaceX, Anthropic, etc., sitting in the $68.7B non-marketable securities balance). They pushed reported net income +32% while operating income rose only +15%. The same effect inflated Q1-2026 net income to +81%. Implication: headline EPS and the trailing P/E understate the true operating multiple; these gains are volatile and can reverse violently in a down market. Use operating income, not net income, to judge the business.
2. Capex / depreciation timing. Capex exploded to $91.4B in FY2025 (+74%) and ran $35.7B in Q1-2026 alone. A large, fast-growing PP&E base ($246.6B net, up from $171.0B) means rising depreciation will be a forward margin headwind — and the useful-life assumptions on AI servers/TPUs are an estimate that, if too generous, currently flatters margins. Watch for any useful-life extension (a classic earnings-management lever).
3. Leverage stepped up: long-term debt $10.9B → $46.5B to help fund the buildout — still trivial against the cash pile ($126.8B cash + marketable securities), but the financing posture changed.
4. Receivables/inventory: no divergence flag — this isn't a working-capital-games business.

Regulatory findings (required sub-section). Read regulatory/regulatory-findings.md: 0 SEC Litigation Releases, 0 AAERs naming Alphabet 2021–2026 — clean on the accounting-fraud axis. The risk is entirely competition law, and it is large. From the 10-K's own Legal Matters note plus `` on outcomes:

• US v. Google — Search (Judge Amit Mehta, DDC). Liability against Google Aug-2024; final judgment Dec-2025. The remedy rejected the DOJ's structural demands — no Chrome or Android divestiture. Instead, behavioral: default-distribution deals (the Apple agreement) capped at one year, non-exclusive, no tying, OEM flexibility, and Google must share its search index + user-side data with "Qualified Competitors" and offer syndication licenses. Google appealed Jan-2026; DOJ cross-appealed Feb-2026. Net: the worst case was avoided (this drove the +8% Sept rally), but the data-sharing and 1-year-default remedies are a real, if survivable, drag.
• US v. Google — Ad-tech (Judge Leonie Brinkema, EDVA). April-2025 mixed liability ruling: publisher ad-server + ad-exchange tools found to unlawfully monopolize; advertiser tools and the DoubleClick/AdMeld acquisitions cleared. Remedies trial Sept-2025; DOJ is seeking divestiture of Google Ad Manager; closing arguments Nov-2025; final judgment still pending. This is now the single biggest open structural risk — a forced Ad Manager sale would carve a piece out of the Network/ad-tech stack (the weakest, slowest-growing revenue line, which limits the damage). A Texas-AG case follows after.
• EC ad-tech: Sept-2025 — €3.0B fine for self-preferencing + cease-and-desist; Google appealed Nov-2025. This drove the $3.5B legal charge in Q3-2025 that (with a separate $1.4B Q2 accrual) lifted G&A by +$6.2B in FY2025.
• EC legacy: Shopping €2.4B (paid $3.0B 2024); Android €4.3B→€4.1B (pending ECJ); AdSense €1.5B annulled Sept-2024 (EC appealing).

Bottom line on Lens 10: the books are clean of fraud (0 SEC actions) but the quality of reported earnings is reduced by the equity-mark gains, and the competition-law tail is the genuine risk — though the most feared outcome (Search structural breakup) was defused in Dec-2025, leaving the Ad-tech (Ad Manager) remedy as the live one.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (EPS, next 3 fiscal years — FY2026/27/28)

Built bottom-up from FY2025 actuals (diluted EPS $10.81 ) and the Q1-2026 run-rate. All outputs ``; arithmetic shown. Note: I project on an operating-earnings basis to strip the volatile equity-security gains — so these EPS figures are deliberately more conservative than the equity-mark-inflated reported number, and a real-world print could exceed them if marks stay positive.

Key inputs:

• Revenue growth: Q1-2026 ran +22%; assume FY2026 ~+16–18% (Cloud +40%+, Search +mid-teens, Network flat) decelerating to ~+13% then ~+11%.
• Operating margin: 32% FY2025; Cloud-mix + leverage push toward ~33–34%, partly offset by rising depreciation from the capex base — net roughly flat-to-up.
• Buybacks ~1.5–2% annual share-count shrink; tax ~16–17%.

. At ~$371 and base FY2026E ~$13.0, the forward P/E is ~28x op-EPS (consistent with the ~26 reported-basis fwd P/E, the gap being the equity-mark gains in the reported denominator). Per --watchlist rules, no Brier forecast logged.

Lens 12 · Bull vs Bear

Bull case. The only company that owns the entire AI stack — silicon (TPU), data centers + private network, a frontier model (Gemini 3), and 6 distribution surfaces with 2B+ users each — is also the one the market spent 2024 pricing as the AI loser. FY2025 broke that thesis: Cloud margin doubled to ~24% (33% in Q1-26), Cloud backlog ~$155B (+46%), Gemini at 750M+ users, Search re-accelerated, and the antitrust breakup was avoided. The capex ($90B+) is not a black hole — it is capacity for a Cloud business that is demand-constrained, and it converts into margin as it fills. Capital allocation is shareholder-friendly ($45.7B buyback + a growing dividend). Secular tailwind: the AI compute super-cycle, where Alphabet sells the picks (TPU/Cloud) and uses them.

Bear case — three things that could permanently impair value:

1. The capex never earns its cost of capital. $90B→$140B+/yr of AI infrastructure against a Cloud business that, even at $58.7B, is one-sixth of revenue. If AI demand normalizes or pricing collapses, you have a depreciation mountain and impaired data centers. This is the real bear case, and it is structural, not cyclical.
2. Search disruption is delayed, not cancelled. AI Overviews currently help, but the long-run risk — users getting answers from a chatbot (yours or a rival's) instead of a 10-blue-links page with ads — attacks the $224B Search profit pool at its economic root (fewer queries-with-ads, lower commercial intent capture). A 2024 fear that simply hasn't arrived yet.
3. Ad Manager divestiture. The one antitrust remedy still live (EDVA, judgment pending) could force a sale of the ad-tech stack.

Pre-mortem (18 months out, thesis broke): Cloud growth decelerated below 30% as hyperscale AI demand cooled and Microsoft/Amazon undercut on price; the Street re-rated the $140B capex from "growth investment" to "stranded asset"; depreciation crushed margins; and an adverse Ad Manager ruling landed simultaneously. The stock de-rated from ~26x to ~18x (Meta's multiple) on lower earnings — a 30%+ drawdown.

Are multiples too high? At ~26x forward vs Meta ~17x and Microsoft ~22x, GOOGL is no longer cheap — the re-rating did the work in 2025. You are now underwriting execution on the buildout, not a re-rating.

Contrarian view (what the market refuses to see): consensus has flipped from "AI kills Google" to "Google wins AI" — both extremes. The thing the market is under-weighting now is margin pressure from depreciation: the same capex everyone cheers as a growth signal becomes a P&L headwind in 2027–28 as those assets depreciate, even if the AI thesis is right. The risk isn't that the AI bet fails; it's that it works and still compresses near-term margins.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case:

• Revenue concentration: ~73% of revenue is advertising, and within it Search is the profit core. The entire equity is, at bottom, a bet that AI does not change how people find commercial information. That is one concentrated bet wearing a diversification costume.
• The most dangerous competitor bulls underestimate: not OpenAI — Meta. Meta has the other great ad-engagement surface, is spending comparably on AI, trades at 17x, and competes directly for the digital-ad dollar. If ad budgets are zero-sum and Meta's AI-driven engagement/targeting pulls share, Alphabet's premium multiple is the one that compresses. (OpenAI/Perplexity threaten the query; Meta threatens the budget.)
• Worst capital-allocation move: the $90B+/yr capex with no proven incremental ROIC — financed partly by stepping down buybacks and taking on $46.5B of debt. If wrong, it's the most expensive mistake in the company's history.
• Earnings quality: management is reporting +32% / +81% net income growth that is substantially equity-mark gains, not operations (+15% / +30%). A skeptic calls that an earnings-quality red flag dressed as a blowout.
• What must hold for ~$371: Cloud stays >30% growth and expands margin; Search ad pool is intact; capex earns its return; Ad Manager survives. If growth disappoints 20–30%, base EPS falls toward the ~$12 bear path and the multiple compresses to high-teens → a stock in the $230–260 range (the low end of the published analyst range, ~$220 ).
• Single permanent-impairment scenario: a genuine, sustained shift of search behavior to conversational AI (whoever owns it) that structurally shrinks the high-margin Search ad pool faster than Cloud can replace its profit. Plausibility: low-to-moderate over 3 years, rising thereafter — the slow-moving tail that never shows up in any single quarter.

Lens 14 · Management Questions (ordered by information value)

1. On the $90B+ annual capex: what incremental ROIC are you underwriting, over what horizon, and what would make you cut it?
2. As the AI server/TPU base depreciates, what is the expected gross-margin drag in FY2027–28, and are useful-life assumptions changing?
3. What share of Gemini's 750M users is incremental engagement vs. default-surface re-routing, and how does AI monetize per query vs. a classic Search result?
4. With AI Overviews/AI Mode live, what is happening to commercial-query volume and ad load — and to revenue-per-query?
5. Cloud margin hit ~33% in Q1-26 — what is the steady-state ceiling, and how much is AI-training (lumpy) vs. durable enterprise workloads?
6. If the EDVA court orders an Ad Manager divestiture, what is the revenue/EBIT at stake and your operational plan?
7. How do the Search remedy's data-sharing + 1-year-default terms change your distribution economics and the Apple relationship specifically?
8. What is the TPU strategy — does it become a merchant product (sold/rented beyond Anthropic), and what does that do to the Nvidia relationship?
9. Capital allocation: with buybacks down and capex up, what is the framework for choosing between infrastructure, buybacks, dividend, and M&A?
10. The $68.7B non-marketable securities book drove $24B of FY2025 income — how should investors think about that volatility, and would you ring-fence it from "operating" results?
11. Other Bets lost $7.5B; what is the path to Waymo standing alone, and when do you stop funding the rest?
12. Where could a competitor most plausibly leapfrog Gemini, and how is the org structured to prevent the 2023 "slow to ship" repeat?
13. What is the power/energy procurement plan (nuclear/PPAs) and is grid access a binding constraint on the buildout?
14. How exposed is the ad business to a macro/ad-cycle downturn now vs. 5 years ago, given Cloud's larger mix?
15. On founder voting control — how should public shareholders weigh that against the scale of capital now being committed?

---