Alphabet Inc. (NASDAQ: GOOGL / GOOG)

1. What changed: H1 2026 turned Alphabet into a different company

Between January and June 2026 four things happened at once, and all four are visible in the filings.

The four structural moves

• Wiz closed 11 March 2026 for $29.5bn. $22.7bn booked as goodwill in Google Cloud, $8.3bn of intangibles (7 to 10 year lives). Goodwill rose to $57.8bn; not tax deductible.
• Intersect (renewable energy) acquired for $5.9bn. Vertical integration into power generation to ease data centre build out bottlenecks; $5.1bn of it was property & equipment.
• $49.6bn of equity issued in June 2026, part of an $80bn programme: 58m Class A/C shares at about $353, a $10.0bn private placement to a Berkshire Hathaway affiliate, and $19.25bn of 6.25% mandatory convertible preferred. A $40bn at the market programme, undrawn at 30 June, follows in Q3. Alphabet had never issued equity for cash before.
• $51.8bn of new senior notes across six currencies, taking face value of long term debt from $49.1bn to $101.1bn, and to $126.1bn after August. Net cash was $142bn.

2. Twelve months of current reports: every 8-K and its exhibits

Alphabet filed 18 Form 8-K reports between 27 August 2025 and 27 August 2026. It has never filed a Form 6-K, 20-F or 40-F: as a US domestic issuer it reports only on 10-K, 10-Q and 8-K, so those forms contribute nothing here. The full log is below; the three items that change the arithmetic follow.

The capital expenditure guidance was raised at every one of the last five calls

The three that matter

• The equity programme is $80bn, not the $49.6bn on the balance sheet. The 4 June press release (EX-99.1) sets out $30bn of underwritten offerings, $10bn to a Berkshire Hathaway affiliate, and a $40bn at the market programme that had not started at 30 June and was expected to begin in Q3 2026. Roughly $30bn of the ATM is earmarked for calendar 2026 employee equity tax obligations: Alphabet is switching from delivering shares net of tax to settling those taxes in cash and reissuing the stock. That removes the share shrinkage net settlement used to provide. Diluted shares already rose from 12,198m in Q2 2025 to 12,309m in Q2 2026.
• Debt is $126.1bn of face value, not $101.1bn. The 10 August 8-K reports a $25.0bn US dollar offering across ten tranches, from floating rate notes due 2028 to 6.500% notes due 2066, at a weighted average fixed coupon of 5.61% and a weighted average maturity of 15.5 years. Together with the May euro, Canadian dollar and yen offerings, Alphabet has funded in six currencies in four months. Q2 interest expense was already $1,278m, up 4.9 times year on year; the August tranche adds roughly $1.4bn a year on top.
• Free cash flow turned negative in the quarter, and management said so. The Q2 release reconciles quarterly free cash flow at $24.5bn, $24.6bn, $10.1bn and minus $5.9bn across the last four quarters. The same release quantifies the earnings distortion directly: of the $9.11 of Q2 diluted EPS, $6.26 came from the equity securities gain, which also carried $21.9bn of tax. That is the company's own number, not an analyst adjustment. LTM free cash flow of $53.3bn is below the $67.0bn of FY2021.

Secondary but relevant

• Capped calls limit conversion dilution to about $530. Alongside the mandatory convertible, Alphabet bought capped calls struck with cap prices of $532.67 on Class A and $527.80 on Class C. Below those levels the hedge absorbs the dilution from the 2029 conversion; above them it does not, so the bull case carries dilution the base case does not.
• The stock plan reserve grew by 200m Class C shares at the 5 June annual meeting, equal to 1.6% of shares outstanding. Say on pay passed with 19% of votes against, and a shareholder proposal for equal voting rights drew 31% support against the founders' Class B block.
• Q1 2026 gives the missing middle of the curve. Revenue of $109.9bn grew 22%, Cloud grew 63%, paid subscriptions reached 350m, and backlog nearly doubled to over $460bn. Cloud accelerated to 82% in Q2, while backlog additions fell from $220bn in Q1 to $54bn, the sharpest change in the disclosed series.
• Google Cloud product revenue is now defined as the sale of TPU systems. The segment description was rewritten this quarter, and inventory rose from $2.4bn to $10.0bn in six months, $6.7bn of it in Q2 alone. Alphabet is building hardware to sell, not only to run, and management expects most of the revenue to land in 2027.

4. Twenty two quarters, from every 10-Q

The annual view hides the shape of the story. Pulled from all seventeen 10-Q filings since Q1 2021, with the fourth quarter of each year derived from the 10-K, the quarterly series shows a company that bottomed in Q4 2022 at 1% growth and a 24.5% operating margin, then compounded for thirteen straight quarters. It also shows two breaks in that trend: the $3.5bn European Commission ad tech fine in Q3 2025, and the sequential margin decline in Q2 2026 as depreciation rose.

What the quarterly series shows that the annual one does not

• Growth ran from 1.0% in Q4 2022 to 24.2% in Q2 2026, with two sequential decelerations. Q2 2026 was the twelfth consecutive quarter of double digit growth.
• The margin peak was Q1 2026, not Q2. Operating margin reached 36.1% in Q1 2026 and fell to 34.0% in Q2. Depreciation reached the margin one quarter earlier than the annual figures show: operating income still rose, to $40.8bn, but it grew 2.7% sequentially while revenue grew 9.0%.
• Free cash flow inflected in Q1 2026, not Q2. Quarterly FCF ran $24bn to $25bn through 2025, halved to $10.1bn in Q1 and went negative in Q2, two quarters of deterioration.
• Cloud margin has held while scaling. The segment went from a 51.7% loss margin in Q1 2021 to 35.6% positive in Q2 2026, and has not given a quarter back since Q1 2023.
• Google Network has declined for sixteen consecutive quarters, from a $9.3bn peak in Q4 2021 to $7.3bn, while Alphabet's total revenue doubled.

What the ratios say

• Margins expanded as growth accelerated. Operating margin went from 26.5% in 2022 to 32.0% in FY2025 and 33.1% LTM, as growth rose from 7.0% to 20.1%.
• Cost discipline held even through the AI build. S&M fell from 9.4% of revenue (2022) to 7.1% (FY2025). R&D rose to 15.2%.
• Capital intensity is the break in the series. Capex went from 8.7% of revenue (FY2021) to 29.7% LTM; capex/D&A is 5.2x, meaning the asset base is growing more than five times faster than it is being written off. Depreciation has not yet caught up; that is a future P&L event, not a past one.
• FCF conversion collapsed. FCF margin fell from 26.0% (FY2021) to 11.9% LTM. Payout topped 100% of FCF from FY2023 to FY2025, funded from cash and then by issuing stock.
• Leverage is rising from a near zero base. Debt/equity 0.11 at FY2025 and 0.15 at 6/26, and debt has doubled twice in eighteen months.

6. Segment analysis: two companies inside one

Revenue concentration by line of business

Revenue by geography, ten quarters

Alphabet reports revenue for four regions, and only at the consolidated level: there is no geographic split of Google Services against Google Cloud in any filing, so the regional mix below cannot be attributed to a segment. What it does show is that the H1 2026 acceleration was almost entirely American. Of the $23.4bn of year on year revenue growth in Q2 2026, 63% came from the United States alone, and the US crossed 50% of group revenue for the first time.

7. Quality of earnings: the $149bn that is not earnings

Alphabet's nonmarketable equity portfolio was remarked upward by $135.9bn in H1 2026 and $24.1bn in FY2025, a total of $149.0bn over the LTM, or 49.8% of pretax income. The carrying value of nonmarketable equity securities rose from $64.1bn at 31 Dec 2025 to $124.3bn at 30 Jun 2026, of which $87.9bn is now classified Level 2 (observable price changes in orderly transactions, i.e. third party funding rounds, not quoted prices). These are unrealised, noncash, and reversible: FY2022 saw a $5.5bn loss on the same line.

8. Sector analysis: Alphabet against the mega cap platform cohort

Alphabet competes in three overlapping sectors: digital advertising (against Meta and Amazon), cloud infrastructure (against AWS and Microsoft Azure), and the AI model layer (against OpenAI, Anthropic and Meta). Peer comparison must therefore be done twice: once at the company level, once at the cloud segment level.

Cloud head to head

9. Ownership and insider activity

Two facts dominate the ownership record, and they point in opposite directions. Warren Buffett has built Alphabet into 12.6% of Berkshire Hathaway's equity portfolio in three quarters. Alphabet's own executives have not bought a single share on the open market in twelve months.

Beneficial owners above 5%

Insider transactions, 1 August 2025 to 27 August 2026

10. What management has said, and what it has not

All five earnings calls from Q2 2025 to Q2 2026 were read from Alphabet's own transcripts. The most important finding is a negative one.

11. Scenario analysis, FY2026E to FY2028E

Core assumptions common to all three cases

• FY2026 is largely locked. H1 2026 revenue of $229.7bn is actual; H1 has run at 46% to 47% of the full year for three years, bracketing FY2026 at $488 to $502bn.
• The useful life assumption is ours, not the company's. Alphabet did not discuss server useful lives on any of the last five earnings calls, and the Q2 2026 10-Q discloses no change to the six year server life adopted in 2023. The 6, 8 and 10 year assumptions below are analyst judgement and are the largest single source of error in this model.
• Capex is modelled from guidance ($195 to $205bn for 2026) and then diverges. Depreciation is built from a vintage model: capex incurred before 2026 depreciated over a 12 year blended life (which reproduces FY2025 actual depreciation of $21.1bn to within 1.6%), new AI vintages over 6/8/10 years in bear/base/bull.
• Margins are modelled before depreciation, which is then subtracted. This separates the operating story from the accounting consequence.
• Equity security marks are excluded from all forecasts; they are unforecastable and noncash.
• Share count now rises in every case. The $40bn ATM starts in Q3 2026 and the RSU tax change removes the shrinkage net settlement provided. Diluted shares were 12,309m in Q2 2026 against 12,198m a year earlier; only the bull case, which assumes buybacks resume in 2027, gets back below 12,200m by 2028.
• Preferred dividends of $1,203m p.a. deducted; mandatory conversion falls in May 2029, outside the window, and capped calls absorb it up to $530.
• Tax rate 16.5% to 18.0%; other income turns negative as interest on $126bn of debt, already $1,278m in Q2 alone, outruns interest income on a shrinking cash pile.

Valuation

12. Critical factors driving the bull case

The bull case is not "AI works". It is four specific, measurable things happening together. Each has a disclosed metric you can track quarterly.

13. Sensitivity analysis

A. FY2028 core EPS vs revenue CAGR and margin before depreciation. Base case 19% / 47.5% → $14.95.

B. Implied 2028 price vs core EPS and exit P/E. $341 today ≈ $14 EPS at 24x, or $11 at 32x.

C. The decisive grid: FY2028 core EPS against assumed useful life of new AI infrastructure and FY2028 capex, holding revenue growth (19% CAGR) and margin before depreciation (47.5%) at base case.

Reading the grids

• Depreciation alone moves EPS by a fifth. Grid C freezes revenue and operating performance at base case, yet FY2028 EPS ranges from $13.08 to $16.80. Cutting the assumed useful life of new AI infrastructure from 12 to 6 years costs $3.07 of EPS (19%), a pure accounting judgement management can revisit at any time.
• Growth and margin matter equally. Grid A: lifting revenue CAGR from 16% to 22% at base margin adds $3.25 of EPS; lifting margin before depreciation from 44.5% to 50.0% at base growth adds $2.49. Alphabet has delivered 20% growth but never a 50% margin before depreciation, so the margin axis carries more uncertainty.
• The multiple carries as much risk as the earnings. Grid B: at $15 of 2028 EPS, the difference between a 16x and a 32x exit multiple is $240 to $480 per share. Alphabet has traded between 17x and 30x forward earnings over the last five years.
• Breakeven. Holding today's price flat to end 2028 requires roughly $14.20 of core EPS at a 24x multiple, modestly below our base case of $14.95. The stock is therefore discounting something close to, but slightly above, the base case.

14. Share price, monthly, five years

GOOGL month end close, Aug 2021 to Aug 2026, from consolidated monthly OHLCV bars, split adjusted. ¹ 2021 measured from the 30 Aug 2021 close; 2026 is year to date.

15. Risks, and the case against this analysis

• A large outside investor has taken the opposite view. Berkshire Hathaway went from a $4.3bn starter position in Q3 2025 to $37.8bn by June 2026, 12.6% of its equity book, paying open market prices for roughly $7bn of it while the stock traded between $290 and $385. That position implies the cash flow trough is temporary.
• The bear case may overstate depreciation. If AI infrastructure remains productive for 8 to 10 years, and Alphabet has already extended server lives once, from 4 to 6 years, then the 6 year bear assumption overstates the charge. There is no way to resolve this before the assets age.
• The equity marks may prove conservative, not aggressive. Treating $149bn of gains as outside earnings isolates the operating business, but the underlying stakes are assets. If they can be monetised, the balance sheet is stronger than the core earnings framing implies. Level 2 marks based on funding rounds are, however, procyclical.
• Regulatory overhang is unresolved and not modelled. The US v. Google remedies, EU adtech proceedings and the $1.5bn PriceRunner award are ongoing. Structural remedies affecting Chrome or default search agreements would invalidate the Services revenue path in all three scenarios.
• Concentration. Search & other is 56% of revenue and most of the profit. A step change in how people search would outweigh everything else here; the metrics do not show it.
• The ATM is open ended dilution. $40bn is authorised and undrawn at 30 June, and the stock plan reserve grew 200m Class C shares. Both dilute while the buyback stays off.
• Buyback suspension is a signal, not just a cash flow item. A company that returned $50bn to $70bn a year from 2021 to 2025 stopping entirely, and at the same time selling equity below its own recent highs, is placing internal reinvestment ahead of returning capital. Repurchases fell to zero in H1 2026 from $17.4bn over the prior twelve months.
• Model limitations. The depreciation engine simplifies the asset base; forecasts exclude GFiber; exit multiples are judgement. Figures are from filings; forward cases are opinion.