GOOGL deep dive: business lines, cost engine, 2-year price framework

2026-07-29 · spot $335.76 · companions: options report · restrategy · synthesized from 4 sourced research passes (segments, costs, Street, data centers) - analysis, not licensed financial advice

Verdict: Alphabet is executing better than the stock implies - every operating line beat in Q2 - but the market has stopped paying for revenue and started demanding capex payback. The 2-year outcome hinges on two variables, roughly in order: (1) what multiple the market pays in 2028 (the swing factor worth ~$130/share), and (2) whether Cloud's 35% margin survives the depreciation wave. The evidence-weighted 2-year range is $320-475, with a disaster floor at $185-235 (trough multiples, twice-tested) and a bull tail above $515 (TPU externalization). Base case: $405-475 by late 2028 - a grind back through the ATH, not a moonshot.

Q2 revenue

$119.8B

+24% YoY

Cloud

+82%

$24.8B · 35.6% margin

Backlog

$513.9B

10-Q verified

2026 capex

$195-205B

2027: "significantly" more

Q2 FCF

-$5.9B

first negative since IPO

Q2 buybacks

vs $13.2B a year ago

1 · The business lines (Q2 2026 actuals)

Line	Q2 rev	YoY	The story
Search & other	$63.3B	+17%	First growth deceleration after 4 accelerating qtrs. AI Overviews 2.5B users, AI Mode 1B MAU; ads "monetize at the same rate" (company claim, unaudited). TAC +10% - growing slower than revenue.
YouTube ads	$11.1B	+13%	World Cup boost; FY25 total YouTube (ads+subs) >$60B; subscriptions growing faster than ads (Premium+Music 125M subs).
Subscriptions, platforms & devices	$12.9B	+15%	Google One >150M subs (AI tiers "millions"). Play risk live: rival app stores inside Play since Jul 22 (Epic); JPM/WF est. 1-2% EPS impact.
Network	$7.3B	-7%	Structural decline, smallest ad line.
Cloud	$24.8B	+82%	Margin 20.7% → 35.6% in a year (AWS gap now ~2pts). Backlog $513.9B, Anthropic ~40% (triangulated). OpenAI now a TPU inference customer.
Other Bets	$0.4B	+2%	Op loss widened to -$1.8B. Waymo: 400k paid rides/wk, 1M target by EOY, $126B valuation (Feb raise), 20+ new cities incl. Tokyo/London.

Balance-sheet assets the multiple ignores

SpaceX stake: $94.1B fair value (10-Q). $80B becomes sellable from early Aug 2026 - a potential funding source for capex or resumed buybacks.
Anthropic-driven non-marketable securities: $124.3B (10-Q line; "overwhelming majority" Anthropic per Bloomberg). Circular with the cloud backlog - see restrategy §6b.
Cash $242.5B vs debt ~$98B: still heavily net cash, but debt is up ~9x in 18 months.
Together ~$220B+ of stakes ≈ $18/share; SOTP context: BofA has core Search+Play at ~13x once Cloud/YouTube/Waymo stripped out (vs S&P ~20x).

2 · The cost engine: where the bear case lives

Metric	FY24	FY25	Now	Trajectory
Capex	$52.5B	$91.5B	$195-205B guide	2027: MS ~$250B, D.A. Davidson ~$350B (their $350 target rests on this)
D&A	$15.3B	$21.1B	$25.2B TTM	MS: ~$549B of sector AI capex not yet through P&Ls, "accelerating through 2027-28"
FCF	$72.8B	$73.3B	$53.3B TTM	Q2: -$5.9B. Consensus qualitative: trough 2027, recovery 2028 (Wedbush: inflection 2028)
Buybacks	$62B	$45.7B	$0 in Q2	Halted to fund capex; dividends ($10B/yr) intact
LT debt	$10.9B	$46.5B	$98.2B	$30B+ Feb bond week (incl. 100-yr sterling note); Moody's sector warning

The depreciation math that decides 2028 EPS: ~$200B/yr of capex on ~6-yr server lives implies ~$33B+/yr of incremental depreciation once in service - roughly 25% of today's entire core net income, arriving 2027-2029. Morgan Stanley cut META 2027/28 EPS 3%/7% on this mechanism; no bank has published the GOOGL-specific number yet (gap).
The offset - Google's structural cost advantage: SemiAnalysis models TPU v7 at ~44% lower TCO than equivalent Nvidia Blackwell systems; the mere threat of TPUs reportedly got OpenAI ~30% off its Nvidia fleet. Google's compute cost per token is the best of any hyperscaler - the margin question is whether that advantage outruns the depreciation wave.
Energy inflation is real but secondary: PPA prices +13% (solar) / +24% (wind) YoY; nuclear (Kairos 500MW), fusion (CFS 200MW), geothermal deals land 2030s - hedges for the next cycle, not this one.
Headcount ~flat (196k), SBC ~6% of revenue and falling as a share - the efficiency program holds.

3 · The physical buildout (delivery risk on the backlog)

Accelerating, not slowing: Texas $40B (3 sites), Missouri $15B, Oklahoma $9B, Virginia $9B, Alabama $1.5B announced within months; internationally Germany €5.5B, Belgium $5.8B, India $15B (GW-scale). Anthropic's 3.5GW 2027 TPU capacity (Google/Broadcom 3-way deal) predominantly US (TX/VA/IA signals).
Only one confirmed Google cancellation: ~$1B Indianapolis campus (community opposition, Sept 2025). Industry-wide: MSFT froze ~1.5GW self-build + paused leasing (mostly non-binding LOIs); AMZN "digesting" but built 4GW in '25; META added $79B of leases in Q1 alone. Sector lease commitments: record $850B+.
The real constraint is delivery: grid queues 4-7yrs in NoVA/Phoenix/Dallas; transformers 3-5yr lead times; switchgear sold out through 2028; est. 30-50% of 2026-dated sector capacity slips to 2028+. Opposition scaling: ~$130B of US projects blocked/delayed in Q1 2026; NY enacted the first state hyperscaler moratorium.
Implication: Google's ">50% of backlog converts in 24 months" carries systemic slippage risk - slipped delivery defers revenue AND defers depreciation, compressing the P&L story in both directions.

4 · Street positioning (post-print)

Camp	Targets	Thesis
High (Citizens 515, TD Cowen 475, BMO 465, GS 450, Citi 447)	$445-515	Backlog pre-funds the capex; TPU external sales; Search is an AI winner; stakes are free options
Middle (JPM, Truist, Cantor, Evercore, Barclays, BofA)	$420-430	"Buying opportunity" but trimmed for FCF; BofA SOTP: core ads at 13x is too cheap
Low (MS 400, Piper 395, UBS 379, D.A. Davidson 350)	$350-400	Depreciation outpaces revenue (MS -5% EPS); DAD models $350B 2027 capex; UBS prefers AMZN

64 analysts: 44 Strong Buy / 14 Buy / 6 Hold / 0 Sell; avg $427.59 (+27%). Note the shape: nobody is bearish on the business - the entire dispersion is about capex ROI timing. Pre-earnings options skew was 9:4 calls; post-print IV crushed 60%→~35% with put-hedging up.

5 · Technical setup (2026-07-29)

Indicator	Value	Read
Price vs MAs	335.76	below 20d ($349.6) and 50d ($360.6); above 200d ($324.9); 50/200 still golden
RSI-14	43.6	recovered from oversold (32.5 on 7/24); neutral-weak
MACD	-8.3 (hist -1.95)	negative, below signal - downtrend momentum not yet reversed
Returns	-5.8% 20d / -12.6% 60d / +75% 1yr	correction inside a still-huge 1-yr uptrend
Bounce quality	vol ratio 0.94	rebound on below-average volume - weak conviction so far
Volatility	RV20 39.5%	realized ABOVE long-dated IV (~36%) - options not expensive vs recent movement

Levels: support = 200-day ~$325, then post-earnings low $316.7; resistance = 20d ~$350, the 50d/pre-earnings shelf ~$360, then $408.61 ATH. The technical picture is "damaged uptrend attempting to stabilize at the 200-day" - constructive above ~$317, broken below it.

6 · The 2-year framework: five bands, and what has to happen for each

Band (late 2028)	Price	~Prob*	Preconditions - what has to actually happen
Disaster floor	$185-235	15-20%	A 2008/2022-severity multiple crush lands at horizon: AI funding cycle breaks (Anthropic can't fund ~$40B/yr → backlog quality implodes + $124B stake marks down through GAAP EPS), ad recession, capex ROI verdict negative. Anchors: trough P/B 4.0-5.3, trough P/S 4.4 (hit in both '08 and '22).
Dead money	$320-365	~30%	Your glut scenario: compute oversupply 2027-28, memory cycle rolls over, Cloud margin compresses back toward mid-20s as $33B+/yr new depreciation bites, growth continues at median (~17%) but P/S reverts to median 6.69x. D.A. Davidson's $350B-capex-2027 world. Stock does nothing; theta sellers win.
Base	$405-475	~30%	Consensus EPS ~$17.3 (2028) × 24-27x. Requires: Cloud delivers backlog on schedule (grid/transformer risk manageable), margins hold high-20s-to-low-30s as TPU cost advantage partially offsets depreciation, capex plateaus in 2028 (Barclays view), Search decelerates gracefully to low-teens. Reclaims ATH during 2027-28.
Bull	$500-575	~15%	EPS $18-19 × 28-30x. Requires: cloud margin holds ≥35% through the wave (TPU advantage dominates), TPU external sales become a disclosed line (Citizens' $515 thesis), FCF inflects 2028 and buybacks resume (SpaceX stake sale is the wildcard funding source), Gemini keeps ~parity with OpenAI.
Melt-up	$600+	~5%	The 5-year median-reversion math ($614) pulled forward: AI capex supercycle extends without a glut, Alphabet is consensus AI platform winner, multiple >30x. Requires euphoria, not just execution.

*Probabilities blend your saved distribution (scaled to Dec-2028) with market-implied; the market prices the left tail fatter (P(<230) mkt 24% vs your 17%). They are judgment, not output.

The two swing variables, ranked: (1) The 2028 multiple - across these bands the multiple moves the price ~$130/share while execution moves ~$50; every historical Google drawdown was multiple-led, never revenue-led. (2) Cloud margin through the depreciation wave - 35.6% today; every 5 points of Cloud margin ≈ ~$6-7B of op income ≈ ~$0.40-0.45 of EPS ≈ ~$10-12/share at base-case multiples.

6b · Six-lens panel (2026-07-30) and EV-ranked Dec-2028 put structures

Six independent analytical personas (Stratechery aggregation, Bridgewater macro/bubble, SemiAnalysis hardware econ, forensic short-seller, quality-compounder, Damodaran DCF) each produced end-2028 quantiles; equal-weight mixture below, then real Dec-2028 GOOG quotes (sell bid / buy ask) EV-ranked against it.

Lens	P5	P25	P50	P75	P95
Stratechery	195	290	395	460	560
Bridgewater	150	245	350	440	560
SemiAnalysis	205	295	400	515	660
Short-seller	120	230	285	360	460
Compounder	250	310	407	500	600
Damodaran	170	310	375	450	630
Mixture	168	273	365	459	599

Key mixture probabilities: P(S<230)=14.2% · P(S<270)=24.2% · P(S<299)=31.7% · P(S<320)=37.3%.

Structure (Dec-28)	Credit	Max loss	EV (mixture)	EV/risk	P(profit)	EV under bear lens
340/240	$3,815	$6,185	+$874	14.1%	67.5%	-$1,219
340/250	$3,555	$5,445	+$790	14.5%	66.9%	-$1,162
310/250	$2,125	$3,875	+$521	13.4%	70.9%	-$695
310/240	$2,385	$4,615	+$604	13.1%	71.5%	-$752
320/230 (user's shape)	$2,950	$6,050	+$658	10.9%	70.4%	-$1,058
naked 320P	$5,000	$32,000	+$1,970	6.2%	75.8%	margin floats - rejected on risk-adjusted basis

All spreads are +EV under 5 of 6 lenses and -EV only under the short-seller (his P50 $285 sits below every breakeven). The entire edge = the market pricing the bear lens at higher weight than the panel's 1-in-6. Fills at bid/ask; mids improve everything ~$150-250. Management per the spread playbook.

7 · Catalyst calendar

When	Event	Why it matters
Jul 29-30 (now)	MSFT + META, then AMZN earnings	The live test of "capex scrutiny" - their 2027 language sets the sector multiple this month
Early Aug 2026	SpaceX stake becomes sellable ($80B)	Potential capex/buyback funding; also a signal about how Alphabet views the mark
Any time	Brinkema ad-tech remedies ruling (still pending)	AdX divestiture = first forced breakup; behavioral remedies = relief rally
Late Oct 2026	Q3 print	First hard 2027 capex number (MS $250B vs DAD $350B gets resolved); FCF trajectory
Late 2026 - early 2027	DC Circuit oral args (search case)	Chrome-divestiture tail risk back on the table if DOJ cross-appeal advances
Rolling	Anthropic funding rounds; memory prices; hyperscaler 2027 guides	Your crash-thesis tripwires: the backlog's counterparty, the cost-inflation cycle, and the "planning season pullback" signal

Sources: four research passes (2026-07-29) over primary filings (Q1/Q2 2026 10-Qs, 8-K exhibits), earnings-call transcripts, and press - full citations preserved in the session research outputs and archived with this report. Known gaps flagged inline: no bank has published a GOOGL-specific depreciation-EPS cut; segment-level consensus unavailable without a terminal; Brinkema ruling status should be re-checked same-day before acting; relative-P/E cross-section returned conflicting data and was omitted. Consensus estimates from aggregator scrapes (FY27 EPS ~$14.7, FY28 ~$17.3) - directionally reliable only. Probabilities are explicit judgment. Not licensed financial advice.