Alphabet
Advertising (Search, YouTube, Network) ~70% of revenue + Google Cloud (IaaS/PaaS, enterprise AI, GCP) + subscriptions/devices + Other Bets (Waymo, etc.); monetizes a vertically-integrated AI stack (Gemini models → TPU silicon → datacenters → Cloud + ad products).
The thesis on this name
State of AI Compute
Both a buyer and a custom-silicon threat — Google Cloud + Gemini scale the demand while in-house TPUs disintermediate merchant GPUs.
State of the AI Cloud
Own the toll-collector, not the tenant: the fastest-growing large cloud (GCP +~63% YoY) with a backlog that nearly doubled to ~$462B in one quarter, and the ONLY hyperscaler renting a differentiated in-house accelerator (TPU) at scale — so it collects the capacity toll whoever wins the model war, funded by a profitable core rather than venture debt. Anthropic's up-to-1M-TPU deal (>1GW, tens of $B) proves external TPU demand.
State of the AI Cloud
The only hyperscaler renting a differentiated in-house accelerator (TPU) at scale — fastest-growing large cloud with a backlog that nearly doubled in a quarter.
State of the AI Cloud
Deepest, best-capitalized tenant book + the only scaled in-house-accelerator rental (TPU); the AI-capex factor diversified inside a profitable core.
State of Frontier AI
The only fully vertically-integrated frontier lab that is also a listed, profitable, market-multiple stock: Gemini (paid Enterprise MAU +40% QoQ, marquee logos Bosch/Mars/Merck), DeepMind world models (Genie 3 — leading the physical-AI category), in-house TPUs structurally lowering inference cost, GCP distribution, and a search cash engine funding $175-185B capex. Uniquely among the arms, the market REWARDED its capex story in 2026 while punishing MSFT/META — evidence the integrated model is being recognized.
State of Frontier AI
The only vertically-integrated frontier lab you can own outright, at a market multiple — Gemini + DeepMind world models + TPUs + GCP + search cash, and the one arm the market rewards for capex.
State of Frontier AI
The only fully vertically-integrated frontier lab you can buy: Gemini (paid Enterprise MAU +40% QoQ), DeepMind world models (Genie 3), in-house TPUs cutting inference cost, GCP distribution, and a search cash engine funding $175-185B capex. The market rewarded GOOGL's capex story when it punished MSFT/META (Fortune, Apr'26). Owns the AI stack end-to-end at a ~market multiple.
State of Physical AI
Reference-only diversified arm: Gemini Robotics / RT-class VLA research keeps Alphabet at the frontier of robot foundation models, but robotics is an immaterial fraction of a search/cloud/ads compounder. You own GOOGL for its core franchise and get embodied-AI research as a no-cost optionality, not as a thesis driver.
State of Quantum Computing
Google has demonstrated 'verifiable quantum advantage' (Willow, 2025) and runs one of the deepest error-correction programs, all funded by a dominant ads/cloud/AI franchise. Like IBM/NVDA, quantum is embedded optionality inside a cash-generative compounder — own GOOGL on its core merits with quantum as upside. Reference name (already covered); not a speculative pure-play.
State of Quantum Computing
Demonstrated 'verifiable quantum advantage' (Willow, 2025) and deep error-correction leadership, all funded by a dominant ads/cloud/AI franchise. Own on core merits; quantum is embedded optionality.
Earnings, margins, COGS & capex
Q1 FY26 was a blowout: revenue +22% to $109.9B, operating margin expanded to 36.1%, and EPS of $5.11 crushed the ~$2.66 consensus (net income $62.6B but inflated by a $36.9B unrealized equity-securities gain — core operating beat still large). Cloud hit escape velocity (+63%, $20.0B, op margin 32.9%) with a $462B backlog. The offset: capex more than doubled to $35.7B/qtr, FY26 capex guided to $180–190B, buybacks paused, and TTM FCF compressed to $64.4B.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~38¢ is cost of goods and ~28¢ operating expense, leaving ~34¢ of operating profit (~57¢ net).
Revenue trend
Margins
up +2pp YoY; FY2025 was 31.6% — structurally expanding despite capex
up sharply from ~17.8% YoY; op income tripled to ~$6.6B (segment figure per teardown)
flattered by $36.9B equity gain; underlying ~35–40%
COGS structure
Not separately disclosed; cost of revenue dominated by TAC (traffic acquisition cost paid to distribution partners incl. Apple), content/YouTube costs, and rapidly-rising datacenter depreciation as the AI capex base is placed in service.
Capex
Q1 FY26 $35.7B (more than doubled from $17.2B YoY). FY2026 guidance $180–190B (raised from $175–185B to include the Intersect acquisition). Mix ~60% servers / ~40% datacenters & networking. This is the single biggest swing factor on FCF.
Latest earnings
Revenue was the beat; EPS was not. Revenue $119.8B vs ~$116.5B consensus (+2.8%). GAAP diluted EPS $9.11 (+294% YoY vs $2.31) against ~$2.88 consensus — but a $99.0B pre-tax net gain on equity securities (the SpaceX IPO mark) added $6.26 to diluted EPS, $77.1B to net income and $21.9B to the tax provision; ex-gain diluted EPS ≈ $2.85 and ex-gain net income ≈ $35.1B, i.e. roughly IN LINE, not a 3x beat. Reported net income $112,193M ($112,107M available to common after $86M of preferred dividends — the first such deduction in Alphabet's history). Cloud +82% to $24.8B was the operating standout; the stock fell after hours on the capex hike despite the revenue beat.
FY2026 capex guidance raised to $195–205B, from $180–190B (CFO Anat Ashkenazi, Jul 22 2026 call), ~60% of technical infrastructure spend to servers and 40% to data centres and networking. On 2027: capex expected to 'increase significantly', details later; the buildout 'will continue to put pressure on the P&L in the form of higher depreciation expense.' Alphabet gives NO formal revenue or EPS guidance. Shareholder returns: buybacks $0 in Q2 FY26 and $0 YTD 2026 (vs $13.2B in Q2 FY25 and $28.3B in H1 FY25) — halted, not merely paused; dividends paid $2,689M in Q2; in July 2026 the Board declared a $0.22/share quarterly common dividend (payable Sep 14 2026) and a $12.15/share quarterly dividend on the new mandatory convertible preferred (~$0.60 per depositary share).
- Cloud revenue / growth
- $24.8B / +82% YoY (from $13,624M) — includes Wiz from its Mar 11 2026 close; Alphabet does not disclose the contribution, so the organic share of the acceleration is unknown
- Cloud operating margin
- 35.6% (operating income $8,814M)
- Cloud revenue backlog (RPO)
- $513.9B at Jun 30 2026 (total company RPO $519.5B), up >$50B sequentially from ~$462B; just over 50% expected to be recognised as revenue within 24 months
- TPU merchant line
- Q2 FY26 release states 'Google Cloud generates product revenues primarily from the sale of TPU systems'; CFO confirmed TPU system sales are reflected in the backlog, with the vast majority of that agreement's revenue landing in 2027
- Operating margin
- 34.0% (operating income $40,770M, +30% YoY)
- GAAP diluted EPS
- $9.11 (+294% YoY); ≈$2.85 excluding the $99.0B equity-securities gain
- Google Services
- revenue $94,540M (+15% YoY); operating income $39,544M
- Google Search & other
- $63,271M (+17% YoY, from $54,190M)
- YouTube ads
- $11,055M (+13% YoY, from $9,796M)
- Google Network
- $7,303M (-0.7% YoY, from $7,354M) — the only shrinking revenue line
- Google subscriptions, platforms & devices
- $12,911M (+15% YoY, from $11,203M)
- Other Bets
- revenue $382M; operating loss $(1,799)M
- Alphabet-level activities
- operating loss $(5,789)M (shared AI R&D), widened from $(3,372)M
- Q2 capex
- $44,924M (37.5% of revenue); TTM capex $132,402M
- TTM operating cash flow
- $185.7B (Q2 $39,069M)
- TTM free cash flow
- $53.3B — Q2 quarterly FCF negative $(5,855)M
- Share repurchases
- $0 in Q2 FY26 and $0 YTD 2026 (vs $13,238M in Q2 FY25, $28,306M in H1 FY25)
- Equity raised (June 2026)
- $49.6B net proceeds — $30,499M common (Class A + Class C) + $19,063M of 6.25% mandatory convertible preferred (19M shares / 385M depositary shares, converting around May 15 2029), part of a ~$80B programme announced Jun 1 2026; a $40.0B at-the-market programme was established but undrawn at Jun 30 2026; Berkshire Hathaway took $10B by private placement ($5B Class A at $351.81, $5B Class C at $348.20). $20.3B of senior unsecured notes issued in Q2 alone.
- Balance sheet (Jun 30 2026)
- total assets $921,983M; stockholders' equity $640,480M; property and equipment, net $321,212M (from $246,597M); no commercial paper outstanding
- Gemini / AI usage (company-disclosed)
- 950M Gemini App MAUs; 22B API tokens/minute; nearly 90% of the Fortune 100 using Gemini Enterprise; 'Gemini 3.5 Flash Cyber' cited in the release
- Headcount
- 198,933 employees at Jun 30 2026, from 187,103 a year earlier
- TTM P/E (do not use)
- 17.87 per stockanalysis.com — GAAP TTM, distorted downward by the ~$77.1B after-tax SpaceX gain; not a valuation anchor
Growth drivers
- Google Cloud (enterprise AI infra + AI solutions, +63%, $462B backlog) — the fastest-growing and now highly profitable segment
- Search resilience + AI Overviews / Gemini-in-Search monetizing rather than cannibalizing (+19%)
- Gemini 3 ecosystem — Gemini app >750M MAU, first-party models >10B tokens/min via API
- TPU as a merchant product — Anthropic up to 1M Ironwood TPUs (~$10B racks + ~$42B GCP RPO); external TPU demand a new revenue vector
- YouTube (ads + subscriptions) and subscriptions/devices (+19%)
- Waymo and Other Bets optionality (autonomy at scale)
Reported financials — SEC EDGAR
Audited GAAP figures pulled from SEC filings · latest filing 2026-02-05. The audited primary-source spine — not financial advice.
Revenue — annual (GAAP)
Margins & balance sheet — FY’25
Bull & bear
The operating business is compounding faster than when we last wrote, and in the direction we called: revenue +24% to $119.8B (Q2 FY26, company states the 12th consecutive double-digit quarter), Cloud +82% to $24.8B at a 35.6% operating margin with a $513.9B backlog, gross margin expanding to 61.6%, and merchant TPU now disclosed segment language rather than a thesis. GOOGL also rose to $356.13 (Jul 31 2026 close) through the month the semiconductor complex sold off — the market is not pricing it as AI-hardware beta, which is the vertical-integration argument working.
- Cloud $24,768M, +82% YoY, at a 35.6% operating margin ($8,814M operating income), with a $513.9B revenue backlog (total company RPO $519.5B, up >$50B sequentially; just over 50% expected to convert to revenue within 24 months) — big, fast and profitable at once. CORRECTION we now carry: Wiz closed Mar 11 2026 for $29.5B with goodwill booked into the Google Cloud segment, so this is the first full inorganic quarter and Alphabet does not disclose the split — the organic share of the acceleration is unknown.
- Merchant TPU stopped being optionality and became disclosure: the Q2 FY26 release states Google Cloud 'generates product revenues primarily from the sale of TPU systems', and the CFO confirmed TPU system sales are reflected in the backlog with the vast majority of that agreement's revenue landing in 2027.
- Search has still not cracked: Google Search & other $63,271M (+17% YoY, from $54,190M); Google Services $94,540M (+15%) generating $39,544M of operating income. The AI-cannibalisation bear case remains unproven in the core line.
- Margin structure improving where it is measurable: gross margin 61.6% (revenues $119,796M less cost of revenues $45,943M), up from 59.5% a year ago; consolidated operating margin 34.0% ($40,770M operating income, +30% YoY, +2pp YoY).
- Distribution and adoption at scale (company-disclosed): 950M Gemini App MAUs, 22B API tokens/minute, nearly 90% of the Fortune 100 using Gemini Enterprise — up from the >750M Gemini MAU figure we published.
- Not traded as AI-hardware beta: GOOGL went from ~$342 to $356.13 / ~$4.35T market cap (Jul 31 2026 close) across a month in which the PHLX Semiconductor Index closed at 11,311.08 after a severe drawdown — the divergence supports the vertically-integrated framing.
- Berkshire Hathaway took $10B of the June equity issue by private placement ($5B Class A at $351.81, $5B Class C at $348.20) — third-party validation on the other side of the capital-raise debate.
The funding model changed, and the self-funding bull case died with it. Q2 FY26 free cash flow was NEGATIVE $(5,855)M — the first quarter in the disclosed series where Alphabet did not self-fund its capex — buybacks are at $0 YTD rather than merely paused, long-term debt roughly doubled to $98.2B, and Alphabet raised $49.6B of net equity proceeds in June inside a ~$80B programme. FY26 capex guidance went up again to $195–205B with 2027 guided to increase significantly, and the CFO explicitly warned of higher depreciation pressure on the P&L — which is already visible in the sequential operating-margin fall from 36.1% to 34.0%.
- FALSIFIES our 'net-cash balance sheet self-funds the buildout' point: Q2 FY26 FCF NEGATIVE $(5,855)M — operating cash flow $39,069M against $44,924M of capex — the first non-self-funded quarter in the disclosed series, even with TTM operating cash flow up to $185.7B. TTM FCF $53.3B, down a further ~$11B from the $64.4B we published one quarter earlier.
- Alphabet is now raising outside capital to build: $49.6B of net equity proceeds in June 2026 ($30,499M common Class A + Class C, $19,063M of 6.25% mandatory convertible preferred — 19M shares / 385M depositary shares, converting around May 15 2029), part of a ~$80B programme announced Jun 1 2026; a $40.0B at-the-market programme established but undrawn at Jun 30 2026; $20.3B of senior unsecured notes in Q2 alone; long-term debt $98,165M, roughly doubled from $46.5B.
- Shareholder returns are halted, not slowed — we framed this as 'buybacks paused': $0 of repurchases in Q2 FY26 and $0 YTD 2026, against $13,238M in Q2 FY25 and $28,306M in H1 FY25. Alphabet now also carries a preferred dividend ($12.15/share quarterly, ~$0.60 per depositary share); $86M was deducted from net income available to common for the first time in its history (net income $112,193M; $112,107M available to common).
- Our capex figure was a full revision behind: FY2026 guidance raised to $195–205B from the $180–190B we published (Jul 22 2026 call), ~60% servers / 40% data centres and networking, with capex expected to 'increase significantly in 2027' and the buildout set to 'continue to put pressure on the P&L in the form of higher depreciation expense.' Q2 capex $44,924M = 37.5% of revenue; TTM capex $132,402M.
- The depreciation wall is no longer a forecast — consolidated operating margin fell ~2pp sequentially from Q1's 36.1% to 34.0%, while property and equipment, net rose to $321,212M from $246,597M.
- The cash cushion is less liquid than the headline: of $242,474M cash + marketable securities at Jun 30 2026, $87,063M is marketable EQUITY securities including $80.0B of SpaceX shares under short-term sale restrictions, plus a further $14,126M of SpaceX in other non-current assets restricted through Q3 2027. Ex-equity net cash is roughly $57B — and that uses long-term debt only, so it slightly overstates the position.
- The headline EPS is not earnings. GAAP diluted EPS $9.11 (+294% YoY vs $2.31) includes $6.26 from a $99.0B pre-tax gain on equity securities (the SpaceX mark: $77.1B to net income, $21.9B to the tax provision). Ex-gain diluted EPS ≈ $2.85 against ~$2.88 consensus — in line, not a beat. Revenue was the beat ($119.8B vs ~$116.5B consensus, +2.8%); the stock fell after hours anyway, on the capex hike. The release warns these marks 'could significantly contribute to the volatility of OI&E in future periods.'
- Google Network $7,303M, -0.7% YoY (from $7,354M) — the only shrinking Alphabet revenue line and the cleanest read available on third-party display disruption. Other Bets: $382M revenue against a $(1,799)M operating loss; Alphabet-level activities widened to $(5,789)M from $(3,372)M.
- Antitrust is unchanged and unargued — our timing call still stands: Google filed a 111-page appeal brief May 22 2026; DOJ filed a 144-page brief urging the court to affirm the monopoly finding and data-sharing remedies while vacating the denial of a ban on distribution payments (the Apple default payment is squarely in play); D.C. Circuit argument expected late 2026 or early 2027, no decision.
- A second regulatory front is REPORTED alongside the DOJ case: European Commission Digital Markets Act orders (reported 2026-07-16) that would require equal system-level Android integration for rival AI assistants and sharing of search query and ranking data. Secondary trade press only — the effective date and exact scope are UNCONFIRMED and we are not publishing them as fact; flagged in unresolved.
What it is worth
Ex-gain earnings power vs. price, plus reverse-DCF anchored on the disclosed backlog conversion schedule. TTM multiples are unusable this quarter and we say so rather than publishing one.
~$230–280
AI assistants erode the search-ad funnel, the Apple default payment is banned or Chrome divestiture ordered on appeal, the buildout becomes stranded depreciation, FCF stays negative and the external funding continues.
~$360–420
Note the current $356.13 close sits at the bottom edge of this range, and the range predates the $195–205B capex guide and the June equity raise.
~$450–500+
Cloud sustains high growth at expanding margins, merchant TPU revenue lands as guided in 2027, Search monetisation holds, capex ROI proves out.
(1) TTM GAAP earnings are dominated by the $99.0B pre-tax / ~$77.1B after-tax SpaceX gain, so the 17.87 TTM P/E now shown on screeners is meaningless — do not use it. (2) The only honest earnings anchor from verified figures is the disclosed ex-gain quarter: diluted EPS ≈ $2.85 in Q2 FY26 against a $356.13 close (Jul 31 2026) and a ~$4.35T common-equity market cap (12.23B shares × $356.13, excluding the ~$19B mandatory convertible preferred). Four quarters at that ex-gain level would not support a ~22–24x forward multiple, and no refreshed multiple is published here because no verified forward consensus was sourced (see unresolved). (3) The per-share base moved: 12.31B diluted shares (from 12.20B YoY) plus ~$19B of mandatory convertible preferred converting around May 15 2029 — known future dilution. (4) The reverse-DCF has one new hard anchor and one new hard cost: just over 50% of the $513.9B Cloud backlog is expected to convert within 24 months, against $195–205B of FY26 capex and a 2027 capex step-up with explicit depreciation pressure guided. The price ranges below have not been re-derived against the new capex guide, share count, preferred conversion or the restricted-SpaceX asset.
SWOT
Strengths
- Two structural cash monopolies — Search and YouTube — funding everything else: Google Services $94,540M revenue (+15%) and $39,544M operating income in Q2 FY26 alone
- Only hyperscaler selling its own AI accelerator as disclosed product revenue: the Q2 release states Google Cloud 'generates product revenues primarily from the sale of TPU systems', with those sales in the backlog and the bulk of that agreement's revenue landing in 2027
- Cloud now fast AND profitable AND contracted — $24,768M (+82%), 35.6% operating margin, $513.9B backlog with just over half expected to convert within 24 months
- Expanding measurable margin structure — gross margin 61.6% (from 59.5% YoY), consolidated operating margin 34.0% (+2pp YoY)
- Scale of AI distribution — 950M Gemini App MAUs, 22B API tokens/minute, nearly 90% of the Fortune 100 on Gemini Enterprise (company-disclosed)
Weaknesses
- No longer self-funding: Q2 FY26 FCF $(5,855)M negative; TTM FCF $53.3B against TTM capex $132,402M
- Buybacks $0 in Q2 and $0 YTD 2026 (vs $28,306M in H1 FY25), plus a new preferred dividend and known ~2029 dilution from the mandatory convertible
- Re-levered balance sheet — long-term debt $98,165M, roughly doubled from $46.5B; and $94.1B of the apparent cash cushion is restricted SpaceX equity, not spendable liquidity
- Sequential operating-margin compression (36.1% → 34.0%) as buildout depreciation begins to land, with management guiding to more of it
- Google Network shrinking (-0.7% YoY) and Other Bets still a drag ($(1,799)M operating loss on $382M of revenue); Alphabet-level activities loss widened to $(5,789)M
- Heavy reliance on the Apple Safari default deal for Search distribution — the exact remedy DOJ is asking the D.C. Circuit to reopen
Opportunities
- Merchant TPU as a second silicon franchise — now disclosed product revenue inside a $513.9B backlog with a named 2027 revenue year
- Backlog conversion as a hard anchor — just over 50% of $513.9B expected to be recognised within 24 months, which a reverse-DCF can lean on instead of guessing Cloud persistence
- Enterprise/agentic AI monetisation — nearly 90% of the Fortune 100 on Gemini Enterprise, 22B API tokens/minute
- Vertical integration into power — Intersect (renewable-energy developer) closed Mar 10 2026 for $5.9B (goodwill $2,174M, P&E $5,129M, $1,214M debt assumed) to unblock data-centre capacity — an energy-bottleneck lever peers must rent
- Security consolidation via Wiz ($29.5B, closed Mar 11 2026) inside the Cloud segment
- Embodied AI adjacency — Gemini Robotics 2 / Robotics ER 2 / On-Device 2 released around Jul 30–31 2026 (secondary sourcing); no near-term revenue relevance but real multiple-support optionality
- Portfolio pruning — GFiber agreed in March 2026 to be contributed into a new entity for $1.5B cash + a $2.0B note + a 49.99% stake, closing expected late 2026
Threats
- Stranded-depreciation risk scales with the guide — $195–205B FY26 capex, 'increase significantly' in 2027, P&E net already $321,212M
- Funding risk is now real risk — a buildout financed by equity and debt rather than cash flow is exposed to capital-market conditions
- DOJ / D.C. Circuit appeal — data-sharing remedies affirmed-or-not and the distribution-payment ban reopened; argument expected late 2026 / early 2027
- EC Digital Markets Act orders reported 2026-07-16 attacking Android assistant distribution and search-data exclusivity — nearer-dated than the appeal if confirmed, but currently secondary-sourced only
- AI assistants disintermediating the query funnel — not yet visible in Search (+17%) but visible in Google Network (-0.7%)
- OI&E volatility from the SpaceX mark itself — a $99.0B pre-tax swing in one quarter, explicitly flagged by the company as a future volatility source
- Nvidia, Microsoft/OpenAI and Amazon (Trainium) competing across Cloud and AI silicon
Moats, dependencies & bottlenecks
Moats
Medium-high (antitrust + AI-assistant erosion are the live threats) ~90% search share; the cash engine, but the most contested moat in the AI era.
Only hyperscaler with a competitive in-house accelerator; lowers AI cost basis vs Nvidia-dependent rivals and now a merchant product.
Gemini models + TPU + datacenters + Cloud + ad surface — few can replicate end-to-end.
Dominant video platform with ads + subscriptions; hard to displace.
Net-cash balance sheet funds $180B+/yr capex few competitors can match.
Dependencies
Fabricates Ironwood/TPU8 on leading-edge (2nm for TPU8); concentration + geopolitical (Taiwan) risk.
Co-designs/implements TPU ASICs (SerDes, packaging); Google-linked AI revenue ~$21B 2026 — deep silicon dependence.
distribution partner Core Search distribution channel and a central antitrust remedy target; loss would hit Search volume.
supplier + competitor Still buys Nvidia GPUs for Cloud customers who demand them even as TPU reduces internal reliance.
revenue concentration ~70% of revenue is ad-driven and macro/AI-behavior sensitive.
$180B+ capex needs energy + grid + construction capacity; a real bottleneck on the buildout.
Advantages
- Owned competitive AI accelerator (TPU) — unique among hyperscalers
- Net-cash balance sheet funding the largest capex program in tech
- Two cash-cow monopolies (Search, YouTube) subsidizing the AI build
- Full-stack control from model to silicon to ad surface
- Cloud now growing AND profitable with a $462B backlog
Weaknesses
- ~70% advertising revenue concentration
- FCF compression + paused buybacks as capex ramps
- Other Bets cumulative operating drag
- Reliance on the Apple default deal for Search distribution
- Coming depreciation wall from the capex placed in service
Bottlenecks
- Power/energy availability for datacenters constrains how fast capex converts to capacity
- TSMC leading-edge wafer allocation for TPUs
- FCF generation vs capex pace — the buildout is outrunning cash flow
- Antitrust uncertainty constraining strategic flexibility (Chrome/Apple deal)
Top signals & trends
Top signals
Q2 FY26 marks up every leg: revenue $24.8B (+82% YoY), operating margin 35.6% (operating income $8,814M), Cloud revenue backlog $513.9B (total RPO $519.5B, up >$50B sequentially), just over 50% of it expected to convert within 24 months. Still the strongest fundamental signal — but it is no longer a clean organic read: Wiz closed Mar 11 2026 for $29.5B with goodwill booked into the Google Cloud segment, and the inorganic contribution is not disclosed, so the organic share of the +63% → +82% step is unknown.
One guidance revision behind and understating the funding cost: FY2026 capex is now guided to $195–205B (Jul 22 2026 call, ~60% servers / 40% data centres and networking), with capex to 'increase significantly in 2027' and the CFO flagging that the buildout 'will continue to put pressure on the P&L in the form of higher depreciation expense.' Buybacks are not paused but halted — $0 in Q2 FY26 and $0 YTD. And the buildout is now externally funded: ~$49.6B of net equity proceeds in June 2026 (common plus 6.25% mandatory convertible preferred converting around May 15 2029), a $40.0B ATM programme established but undrawn at Jun 30 2026, $10B placed with Berkshire Hathaway, and long-term debt roughly doubled to $98.2B.
Third-party validation that TPU is a credible Nvidia alternative — opens a merchant-silicon revenue line.
Chrome divestiture unlikely but live; Apple default + data-sharing remedies are overhangs.
Not compression any more — a funding-model change. TTM FCF is $53.3B and Q2 FY26 quarterly FCF went NEGATIVE at $(5,855)M: operating cash flow $39,069M against $44,924M of capex, the first quarter in the disclosed series where Alphabet did not self-fund its capex, even with TTM operating cash flow up to $185.7B. The shortfall is being met with new debt and new equity rather than internal cash.
Trends
Alphabet a top spender; TPU integration makes it relatively advantaged but raises stranded-asset risk if demand softens.
Google leads hyperscalers; TPU8t/8i (Broadcom/MediaTek, TSMC 2nm) extends the roadmap to external customers in 2027.
Double-edged — threat to query funnel, but Gemini/AI Overviews so far monetizing search (+19%).
Drives the Cloud backlog; Gemini >10B tokens/min via API.
Search remedy appeal + ad-tech cases + EU DMA constrain distribution + practices.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Fabricates Ironwood/TPU8 on leading-edge nodes (2nm for TPU8).
Co-designs/implements TPU ASICs (SerDes, packaging); ~$21B Google-linked AI revenue 2026.
Designs TPU8i inference chip (Zebrafish) — non-US name, analysis only, not a buy framing.
Custom-silicon / interconnect supplier in Google's multi-partner chip supply chain.
GPU supplier for Cloud customers who require Nvidia despite TPU availability.
Datacenter networking gear supporting the hyperscale buildout.
Up to 1M Ironwood TPUs (~$10B racks + ~$42B GCP RPO) — flagship Cloud/TPU customer.
~70% of revenue; the demand base for Search/YouTube/Network ads.
Drive the $462B Cloud backlog across AI infra + AI solutions + core GCP.
Two-sided content/ads + subscription flywheel.
#2 hyperscaler; OpenAI/Copilot lead in some enterprise AI and a direct Search/assistant threat.
#1 hyperscaler and the other custom-silicon hyperscaler; competes in Cloud + ads (retail media).
Direct competitor for digital ad dollars (YouTube vs Reels/IG) and an AI-assistant rival.
Supplies GPUs to Google's Cloud customers but TPU competes head-on for AI training/inference.
Distribution partner (Safari default) and a platform/assistant competitor (Siri/Apple Intelligence).
AI-assistant disruptors of the search-query funnel; ChatGPT the primary consumer-AI rival.