Alphabet Inc. (GOOG / GOOGL)

1Business snapshot

Alphabet reports three segments plus an unallocated corporate bucket. Google Services — Search, YouTube, Android, Chrome, Play, devices and subscriptions — produced $342.7B of FY2025 revenue and $139.4B of operating income, a ~41% margin FY2025 10-K. Inside it, Google Search & other was $224.5B, YouTube ads $40.4B, Google Network $29.8B (shrinking), and subscriptions/platforms/devices $48.0B. Google Cloud did $58.7B of revenue and $13.9B of operating income. Other Bets — chiefly Waymo — did $1.5B of revenue against a $7.5B loss. A fourth line, Alphabet-level activities, absorbed $16.8B of shared frontier-AI R&D that is deliberately not charged to any segment FY2025 10-K.

The economics are simple and unusually good: an auction sells advertiser access to user intent at near-zero incremental cost, with traffic acquisition cost (20.3% of relevant revenue in FY2025) the main variable cost of distribution. Cloud is the opposite — capital-intensive, contract-based, and now partly a hardware business: in Q2 2026 Alphabet began recognising revenue from selling TPU systems into customers' own data centres, which is why inventory jumped from $2.4B to $10.0B in six months Q2 2026 release; Q2 2026 call.

2Core debate

Two questions decide the stock. First, does generative AI expand or erode the Search profit pool? Management's evidence is expansion: queries at an all-time high, AI Mode past 1 billion MAU, monetisation holding on AI Overview queries even as coverage widened to commercial searches, and Search & other up 17% in Q2 2026 Q2 2026 call. The bear reading is that answers displace the blue links, that high-value commercial queries migrate to agents, and that reported growth is flattered by pricing and vertical mix.

Second, is $195–205B of 2026 capex — rising again in 2027 — a moat purchase or a return-destroying arms race? Q2 2026 call. Alphabet is the only bidder with captive demand for its own compute as well as external demand. But the spend is now large enough that free cash flow went negative in a quarter, and the funding mix has shifted from buybacks to issuance.

3Base thesis

Alphabet enters the AI transition owning the whole stack — data centres and fibre, custom silicon, frontier models, and thirteen products with over a billion users each. That is the rare position where AI capability converts into revenue on surfaces the company already controls, rather than having to be sold into someone else's distribution. Search is expanding rather than cannibalising, Cloud is compounding a $514B backlog at a 35.6% operating margin, and the TPUs that serve Gemini are now also a merchant product. If the compute Alphabet is buying earns anything near what the last three years of compute earned, the spend is the cheapest part of the story.

Assumptions this rests on:

1. Search revenue growth stays comfortably positive as AI features scale — i.e., query growth plus improving ad relevance more than offsets any loss of monetisable link clicks.
2. Cloud converts backlog roughly on schedule (just over 50% of $514B within 24 months) without a step-down in margin once third-party bridging capacity and Wiz integration are absorbed Q2 2026 call.
3. Capex translates into revenue at a lag, so that depreciation growth (from $21.1B in FY2025) is outpaced by gross profit growth rather than the reverse.
4. Antitrust remedies remain conduct-based rather than structural — no forced divestiture of the ad-tech stack, no loss of default Search distribution economics.
5. Frontier model parity is maintained; Gemini 4 lands competitively and the acknowledged gap in agentic coding closes Q2 2026 call.

4Main bear case

The bear case is not that Alphabet loses Search quickly; it is that the shape of the business changes for the worse while the market keeps paying for the old shape. Capital intensity has gone from a rounding error to the defining feature: TTM capex of $132.4B against TTM free cash flow of $53.3B, with Q2 2026 free cash flow at negative $5.9B Q2 2026 release. Buybacks stopped in H1 2026 ($0, versus $45.4B in FY2025), replaced by a $49.6B equity raise, a $40B ATM programme, and long-term debt rising from $46.5B to $98.2B in six months Q2 2026 release; FY2025 10-K. Purchase commitments and other contractual obligations now total $811.0B, plus $43.8B of credit derivatives and $7.6B of guarantees backstopping third-party data-centre builds and a further ~$24.1B agreed Q2 2026 10-Q. That is a large fixed-cost base pointed at demand that is currently supply-constrained but not contractually guaranteed beyond the backlog.

How it breaks: AI compute demand normalises or shifts to cheaper inference; depreciation on six-year server lives lands on a decelerating revenue line; Cloud's margin proves to have been a scarcity margin; and meanwhile the Eastern District of Virginia imposes structural ad-tech remedies while the December 2025 Search judgment's data-sharing and distribution restrictions bite FY2025 10-K; Q2 2026 10-Q. In that world a fortress balance sheet has been converted into depreciating assets at the top of a cycle, and reported EPS — currently flattered by $99.0B of unrealised gains in one quarter — reverses hard.

5Business quality

High, but changing character. The durable pieces: default distribution across Search, Chrome, Android and YouTube; an auction whose density improves with scale; a data and feedback loop that no competitor replicates at consumer scale; and ten years of TPU development that gives Alphabet a cost position in inference others must buy at merchant margins ("reduced Gemini serving costs by 78%" in 2025; AI Mode response cost at its lowest since launch) Investor Presentation Jun 2026; Q2 2026 call. Google Services still earned a 41.8% operating margin in Q2 2026 while absorbing legal charges.

What has degraded is capital efficiency, not competitive position. A business that converted ~45% of FY2025 operating cash flow into free cash flow now converts under 30% on a TTM basis, and a growing share of cost — $5.8B in Q2 2026 — sits in an unallocated AI R&D bucket that no segment margin reflects. Google Network (-1% y/y) is in structural decline; Other Bets loses ~$1.8B a quarter.

6Key value drivers

Revenue is driven by (i) Search query volume × monetisable coverage × cost-per-click — paid clicks +6% and CPC +7% in FY2025, a healthy split; (ii) YouTube's shift into the living room and direct response; (iii) subscriptions (350 million paid, Q1 2026) and AI-plan mix; (iv) Cloud consumption, now led for the first time by enterprise AI solutions rather than core GCP, plus TPU system sales whose revenue lands mostly in 2027; and (v) Waymo scaling (500,000+ paid rides per week; $126B round valuation, February 2026) FY2025 10-K; Investor Presentation Jun 2026; Q2 2026 call. Margin is driven by TAC rate, content acquisition, and — increasingly the swing factor — depreciation and data-centre operating cost.

7Financial profile

FY2025: revenue $402.8B (+15%), operating income $129.0B (32% margin), net income $132.2B, operating cash flow $164.7B, capex $91.4B, free cash flow ~$73.3B, buybacks $45.4B, SBC $25.0B, depreciation $21.1B FY2025 10-K.

Trailing twelve months to Q2 2026: revenue ~$445.8B, operating income ~$147.6B (~33%), operating cash flow $185.7B, capex $132.4B, free cash flow $53.3B Q2 2026 release, author's TTM arithmetic from reported periods. Q2 2026 alone: revenue $119.8B (+24%), operating margin 34%, Cloud +82% to $24.8B at a 35.6% margin, Services +15% at 41.8%.

Balance sheet at June 30, 2026: $242.5B cash and marketable securities, $98.2B long-term debt, $131.5B of non-marketable securities (including ~$94B of SpaceX shares under sale restrictions), $321.2B of net PP&E, and $17.4B of accrued legal/regulatory fines and settlements Q2 2026 10-Q. Reported earnings are currently not a clean read: Q2 2026 EPS of $9.11 includes $99.0B of equity-securities gains, worth $6.26 of that EPS Q2 2026 release.

8Valuation

The valuation question is which earnings base to capitalise. On operating income the business is priced at roughly 27x trailing operating income for a company compounding revenue in the low-to-mid twenties. On free cash flow it is priced at ~1.3% — because free cash flow is being deliberately suppressed by an investment programme management says will grow again in 2027. Owning the stock is a judgement that current capex is deferred earnings rather than destroyed capital.

9Management & capital allocation

CEO pay is a triennial equity cycle: salary unchanged at $2.0M since 2020, no bonus, and a March 2026 award of two PSU tranches at $63M target each, vesting on total shareholder return relative to the S&P 100 8-K Mar 6, 2026 — a defensible structure, relative TSR rather than adjusted-EBITDA targets. Insiders and directors as a group hold 54.3% of total voting power through Class B shares, so public shareholders have essentially no governance lever 2026 DEF 14A. Worth noting: the Corporate Controller and Principal Accounting Officer resigned in March 2026 "to pursue another professional opportunity," with a successor named in June 8-K Apr 2, 2026; 8-K Jun 5, 2026.

Capital allocation changed regime in 2026. FY2025 was the old model: $45.4B of buybacks, $10.0B of dividends, capex funded from cash flow. H1 2026 is the new one: zero buybacks, $49.6B of equity issued (common plus 6.25% mandatory convertible preferred), a $40B ATM authorisation for SBC tax obligations, $56.2B of gross debt raised, and $33.7B spent on acquisitions — Wiz ($29.5B, into Cloud) and Intersect ($5.9B, data-centre energy) Q2 2026 release; Q2 2026 10-Q. The dividend was raised to $0.22/quarter. Management states it does not plan further equity raises beyond the ATM Q2 2026 call.

Claims to verify over time: that capex is governed by a "disciplined ROIC framework"; that third-party bridging capacity is a short-term margin cost inside a highly ROI-positive multi-year deal; and that TPU system sales expand the addressable market without diluting Cloud's margin structure.

10Risks ranked by damage

1. Capital-cycle risk — the investment does not earn its cost

Mechanism: $195–205B of 2026 capex, rising in 2027, converts into six-year-life depreciation and data-centre operating cost regardless of whether AI demand persists at current prices. Evidence today: TTM capex $132.4B vs. TTM FCF $53.3B; Q2 FCF -$5.9B; depreciation up from $15.3B (FY2024) to $21.1B (FY2025) and $13.6B in H1 2026 alone; $811B of purchase commitments Q2 2026 10-Q. What would confirm it: Cloud revenue decelerating while depreciation growth accelerates; backlog conversion slipping past the stated 24-month schedule. Impact: compresses margin and free cash flow for years; largely unfixable once spent.

2. Antitrust remedies — Search distribution and ad tech

Mechanism: The December 2025 final judgment restricts how Google distributes its services and requires sharing search data with and offering syndication to competitors; separately, the Eastern District of Virginia found Google's publisher ad tools unlawfully exclusionary and the DOJ has proposed structural remedies, with judgment awaited FY2025 10-K; Q2 2026 10-Q. Evidence today: both matters under appeal; the EC has separately fined Google over ad-tech self-preferencing; $17.4B of accrued fines and settlements on the balance sheet. What would confirm it: a divestiture order in the ad-tech remedies judgment, or an unfavourable appellate ruling on Search distribution. Impact: a structural ad-tech remedy would remove a profit stream and management attention; distribution remedies attack the highest-quality traffic Google has.

3. Answer substitution in Search

Mechanism: As AI answers and third-party agents satisfy queries directly, the share of sessions carrying a monetisable commercial click falls even if query volume rises. Evidence today: countervailing evidence dominates so far — queries at an all-time high, AI Mode past 1 billion MAU, monetisation "encouraging" on AI Overview queries — but Alphabet discloses no monetisation-per-query figure, so the claim cannot be checked independently Q2 2026 call. What would confirm it: Search & other growth decelerating while management continues to report rising query volume. Impact: hits the segment that funds everything else.

4. Cloud demand concentration and circularity

Mechanism: Backlog growth leans on a small number of very large AI-infrastructure commitments, some to counterparties Alphabet also funds — $131.5B of non-marketable securities, a $20.0B milestone-contingent commitment to a private company through 2030, and $43.8B of credit derivatives backstopping third-party data centres Q2 2026 10-Q. Evidence today: backlog $514B, up >$50B sequentially; management says the majority is typical GCP contracts but does not size the AI-lab concentration. What would confirm it: backlog growth stalling, or an impairment/writedown on a funded counterparty. Impact: revenue quality is worse than the headline growth implies.

5. Funding-regime change and dilution

Mechanism: With buybacks suspended and equity issued, share count rises rather than falls; the mandatory convertible preferred converts, and the ATM adds shares to fund SBC taxes. Evidence today: shares outstanding 12,088m → 12,230m in six months; preferred dividends now appear in the EPS bridge Q2 2026 release. What would confirm it: further equity raises beyond the stated ATM. Impact: per-share dilution; also a signal about internally-generated cash adequacy.

6. Frontier model slippage

Mechanism: If Gemini falls behind on the capabilities enterprises buy — agentic coding is the acknowledged gap — the full-stack advantage erodes to a compute-rental business. Evidence today: management explicitly concedes coding as an area needing improvement; Gemini 4 pre-training under way Q2 2026 call. What would confirm it: token growth decelerating, or Cloud AI wins shifting to third-party models on GCP. Impact: margin, not revenue, in the first instance.

11Metrics to monitor

12Thesis tripwires

13Open questions

• Cloud growth and margin excluding TPU system sales and Wiz — management says growth accelerated ex-TPU but has not quantified it.
• How concentrated is the $514B backlog among AI labs versus enterprise GCP, and how much sits with counterparties Alphabet also funds or backstops?
• What monetisation-per-query trend sits behind "queries at an all-time high"? No disclosure exists.
• Will the six-year server life hold? Any extension would flatter margins optically (see §5).
• Does Waymo stay inside Alphabet? Management deflected the structure question in Q2 2026.

14Update log

15Source map