Alphabet— Q2'26 Earnings Reaction (GOOGL)

Executive Summary & Action Plan

Verdict

HOLD — the quarter was operationally excellent and the stock was right to fall anyway: Cloud grew 82% on a $514B backlog, but delivering it cost a third capex raise to $195–205B, Alphabet's first negative free-cash-flow quarter, a halted repurchase program, and $70B of fresh equity and debt. The growth is real; shareholders are no longer the ones being paid first.

The print repriced the funding of AI, not the demand for it. Our fair value is essentially unchanged through the beat — everything Cloud added, capex took away.

Price $342.09 | Market cap $4.18T | Target $286.00 | Upside -16.4% IWANNAVY Fair Value (re-underwritten post-print, adversarially audited) · price as of 2026-07-22 close · after-hours $330.80 (−3.3%) · all-class share base 12.23B · Street $433.55 (Strong Buy, n=65, reference) — the Street underwrites on EPS before the $195–205B capex lands as depreciation, and credits the full backlog at today's margins

Abstract

Alphabet's Q2 2026 beat on every operating line — revenue $119.8B up 24%, Search up 17%, Cloud up 82% at a 35.6% segment margin with backlog reaching $514B — and fell 3.3% after hours because the quarter also revealed what the growth costs. Capex doubled year-over-year to $44.9B, full-year guidance rose a third time to $195–205B with 2027 guided "significantly" higher, free cash flow went negative for the first time, buybacks stopped, and the ramp is now funded by a $49.6B June equity raise, $20.3B of new notes, and a $40B ATM shelf. The headline $9.11 EPS is windfall optics: a $99.0B unrealized equity-securities gain — unattributed by the company, tied by press inference to SpaceX and Anthropic — added $6.26, leaving core EPS of $2.85 slightly below consensus. Our audited three-method value lands at $286, barely moved from $292 pre-print: Cloud's acceleration and the depreciation-and-dilution wave roughly cancel. Rating improves from the July 12 SELL to HOLD on price, not on thesis. Principal risks run both ways — backlog conversion at margin on the upside, depreciation and funding drag on the downside.

Action Plan

No position at spot: reduce into strength toward the bull band, accumulate only at fair value and below, where entry reward-to-risk reaches 1.9:1.

• Risk/Reward at fair-value entry 1.9:1 | Prob-weighted 12M return from spot −10.5% (Bull 37% × +6.7% + Base 41% × −16.4% + Bear 22% × −28.4%) | Confidence Medium

---

Last Four Quarters

Revenue reaccelerated to +24% while free cash flow collapsed from a 24% margin to negative in three quarters — the two lines of this report's whole argument. Net margin is distorted upward in the last two quarters by mark-to-market windfalls ($37.7B in Q1, $99.0B in Q2).

*Windfall-inflated; core Q2'26 EPS $2.85 (+23% YoY) after backing out the $6.26 equity-gain contribution.

---

Revenue & Profit Mix

Search is still the engine at $63.3B; Cloud is the growth story at $24.8B and now carries a real margin. The other-and-tax line in the bridge is the $98B windfall — read it as a balance-sheet event, not earnings power.

---

Business Lines

• Search & other (~53%): +17% YoY with AI features driving query growth. Must know — the AI-disruption bear case keeps failing on the ads line; this is the cash that funds everything else.
• Google Cloud (~21%): +82%, 35.6% operating margin (from 20.7% a year ago), $514B backlog — more than five years of current run-rate. Must know — management is supply-constrained and will rent third-party capacity in Q3 as a bridge; growth is gated by delivery, not demand.
• TPU merchant sales (new): Cloud now includes "product revenue from the sale of TPU systems"; inventory jumped $2.4B → $10.0B and the "vast majority" of contracted revenue lands in 2027. Must know — Google just entered the merchant AI-silicon market its own suppliers sell into.
• Subscriptions, devices, YouTube (~20%): subscriptions +15% on Google One AI tiers; YouTube ads +13%. Steady compounders, not the story this quarter.

---

IWANNAVY Fair Value

Log-median of three methods, adversarially audited: the audit charged the SOTP with ~$23B/year of unallocated AI R&D the analyst had left out (the raw sum-of-parts priced 114% of Alphabet's actual operating profit) and moved the share base to the post-raise 12.23B. Result $286 — a 34% discount to the Street, because we charge the $195–205B buildout to cash flow and depreciation now, while consensus banks the backlog at today's margins first.

• Thesis breaker: two consecutive quarters of positive FCF alongside a maintained capex guide — proof the buildout pays for itself in-period — would force the mark toward the $330s.

---

Read-Through: What This Print Does to Related Companies

The after-hours tape told the story cleanly — the payer fell 3.3% while its suppliers rallied (Micron +2.5%, Nebius +3.9%, CoreWeave +2.9% after hours). Capex is cost to the spender and revenue to the chain. The chart shows our coverage stance on the affected names at their July 22 closes; the print's pressure on each mark is in the notes below.

• Broadcom (AVGO) — the print's cleanest beneficiary: the raise is TPU-heavy, Gemini's 950M users run inference on Broadcom-designed silicon, and merchant TPU sales extend the TAM beyond Google's own budget. Our $320 mark is now the most upward-pressured in coverage.
• Micron (MU), TSMC (TSM), Vertiv (VRT), Arista (ANET), Astera (ALAB), Credo (CRDO) — demand-side support across the board: the implied H2 run-rate of $57–62B/quarter is racks, wafers, HBM, power and fabric. It postpones — but does not refute — the cycle-normalization logic in our below-spot marks; the Korea memory prints (7/29–30) walk into a tailwind confirmed hours earlier.
• Neoclouds (NBIS, CRWV) — the year's best sentence for the group: Google will "expand the use of third-party capacity in Q3 as a bridging strategy." A supply-constrained hyperscaler is now a customer. Near-term utilization validated; the same call's "significantly higher" 2027 self-build starts the clock on the glut our low marks anticipate.
• Hyperscaler peers into 7/29–30 — the reaction template is set: a beat plus a capex raise still gets sold unless FCF and buybacks survive. Microsoft fits the template best (FCF-positive, buyback intact), Meta worst (all-internal spend, no backlog line to show), Amazon in between, and Oracle is the paradoxical winner — Google just normalized backlog-led, debt-funded, FCF-negative building.
• Cerebras (CBRS) — supports our caution: the world's largest inference workload (Gemini, ~22B tokens/min) runs entirely on in-house TPUs at a 35.6% cloud margin, and Google is now selling inference hardware into the merchant market Cerebras needs.
• Software (NOW, CRM, ADBE) — ServiceNow printed the same night: revenue and EPS beats, cRPO +21.5%, AI ACV past $1B with agentic deployments up 9x, guidance raised, +4.7% after hours. Cheap tokens are an input for the workflow layer, not yet its replacement — our $124 mark is confirmed and the 7/22 selloff to $95.46 was an accumulation window. CRM's HOLD stands (Gemini Enterprise attacks its pricing exactly where it is weakest); ADBE's BUY is untouched by this print.

---

Catalysts & Risks

---

References

• Q2 2026 press release, 8-K EX-99.1 (filed 2026-07-22) · Q2 2026 earnings call transcript · Quote, consensus and news feed (fetched 2026-07-23)
• Google justifies its massive AI spending with a booming cloud business (TechCrunch, 2026-07-22) · WSJ on the negative FCF quarter (2026-07-22) · Business Insider on the $98B windfall (2026-07-23)
• ServiceNow Q2 print and transcript (2026-07-22) · Prior-quarter releases: Q1'26 · Q4'25 · Q3'25
• Companion house reports: AI Complex cross-layer outlook (2026-07-22) · AI Compute Silicon sector (2026-07-21)