Alphabet (GOOGL): The 17x P/E Is Fiction. Core Earnings Say 26x — And That's Still a Buy.

1. Where the stock actually is

2. What Q2 2026 actually said (primary source)

Read from the Q2 2026 earnings release (8-K Exhibit 99.1), the 10-Q for the period ended June 30, 2026, and the Q2 2026 earnings call (CFO Anat Ashkenazi's prepared remarks and the Q&A).

3. Finding #1 — the "cheapest mega-cap at 17x" claim is arithmetically false

This is the single most repeated line about Alphabet since Berkshire's Q2 13F showed a 658% increase in the position. It is wrong, and the error is mechanical, not a matter of opinion.

Trailing twelve-month net income is ~$244.1B (TTM EPS $19.91 × ~12.26B shares). Of that, Q2 alone contributed $112.1B — and $98.0B of it was a net gain in "other income," almost entirely unrealized marks on non-marketable equity securities, primarily Anthropic and SpaceX. Taxed at roughly Alphabet's effective rate, that is ~$82B of after-tax "earnings" that did not come from selling an ad or renting a TPU.

Where I disagree with consensus: Alphabet is not the cheapest mega-cap AI name. At 26x adjusted trailing and ~30x on the core run rate, it sits in the same neighbourhood as its peers, not below them. Where I agree with consensus: at 30x on a business growing operating income ~30% with 5.2 years of contracted cloud backlog, a PEG of ~1.0 on a wide-moat monopoly is a good price. The bull case survives the correction. What does not survive is the lazy version of it.

4. Finding #2 — the reflexivity nobody is pricing: Alphabet is booking profit on its own customer

The $99B gain came primarily from marking up Anthropic and SpaceX. Anthropic is simultaneously one of Google Cloud's largest customers, is reportedly readying a $10B credit line ahead of an IPO, and its valuation is set by private rounds in which Alphabet is itself an investor and a compute supplier. Alphabet is recognising accounting income on an entity whose enterprise value is partly a function of compute commitments Alphabet finances.

Two specific downstream consequences, both tradable:

• Consequence A — the mechanical "miss." If Anthropic's IPO prices below the last private round, or if private AI marks compress at all, Alphabet books a large negative other-income line. GAAP EPS collapses, the trailing P/E mechanically doubles overnight, and headlines will read "Alphabet earnings fall 60%." Nothing operational will have changed. That is the dislocation to buy. Set an alert for the Anthropic pricing range versus its last round.
• Consequence B — the ratchet works in reverse for the narrative. Because the 17x figure is what pulled generalist and value money in (Berkshire's +658%, RenTec's Q2 buying), the same figure re-rating to 26–35x on a mark-down will trigger mechanical screening-driven selling from funds that never understood what they owned. Expect the drawdown to overshoot the fundamentals.

5. Finding #3 — the buyback machine became a net issuer, and it is paying 6.5% for it

This is the most under-covered fact in the filings. Three things happened in eight weeks:

• June 2026: per the 10-Q MD&A, Alphabet issued Class A + Class C stock and mandatory convertible preferred for net proceeds of $49.6B, explicitly "to be used for general corporate purposes, including capital expenditures to scale AI infrastructure and global compute," and entered an equity distribution agreement to sell up to $40.0B of stock via an ATM programme.
• August 10, 2026: per the 8-K, Alphabet closed $25B of USD senior notes across ten tranches — including $4.5B of 6.375% notes due 2056 and $2.5B of 6.500% notes due 2066.
• Share repurchases in the quarter: zero.

Second-order: Alphabet's $25B, on top of roughly $223B of hyperscaler IG issuance year-to-date (more than 2x all of 2025), is competing directly with the Treasury for long-duration capital. That is a non-Fed, non-Treasury structural reason the 30Y does not easily clear back below 5.00% — which is precisely the level our regime framework requires before duration-sensitive assets can rotate. Alphabet's capex plan is now a macro input, not just a company input.

6. Finding #4 — the depreciation wall is a 2027 event, and it is not in consensus

Capex grew 100% YoY in Q2. Depreciation grew 18%. That ~82-point gap is not a discrepancy — it is a deferral. Assets bought today start depreciating when they are placed in service.

Consequence: unless cloud revenue growth stays north of ~60% into 2027, core EPS growth compresses from roughly +25% to high single digits. The re-rating risk is not the Oct 26 print — it is the Q4 print in early February 2027, when 2027 capex gets a formal number. A 2027 capex guide above $300B without a matching acceleration in backlog is the bear trigger.

7. What I agree with consensus on

• Cloud is the engine and the backlog is real. $514B of remaining performance obligations against ~$99B annualised cloud revenue is the best forward-visibility number in mega-cap tech. Contracted, not aspirational.
• Search is not breaking. +17% revenue growth with AI Overviews and Gemini integrated. The Gemini app passed 950M+ monthly users. The disruption thesis has now been wrong for eight consecutive quarters.
• Vertical integration is a genuine cost moat. TPUs, and the $120B Marvell arrangement for CXL and custom NICs around TPU inference, mean Alphabet pays a lower per-token cost than any rival renting GPUs. Management explicitly flagged "very strong demand for TPUs as standalone systems from other customers."

8. The two overhangs

8.1 The AI leadership exodus (Aug 5, 2026)

Demis Hassabis moved from Google DeepMind CEO to Chair of GDM and Chief Scientist of Alphabet; Koray Kavukcuoglu stepped up as SVP of Google DeepMind; and Jeff Dean left after 27 years to found an independent public benefit corporation, alongside Oriol Vinyals and other senior researchers. The stock fell 4–5% on the day. My read: this is a real cultural risk and a fake financial risk. Gemini 3's roadmap, the TPU stack, and the $514B backlog are institutionalised, not personality-dependent. But if a second wave of departures follows into Q4, the market will start pricing a model-quality gap versus OpenAI and Anthropic, and that is worth 2–3 turns of multiple.

8.2 Antitrust — the 2027 event, not the 2026 event

Judge Mehta's behavioural remedies took effect after a stay was denied in May 2026: limits on the share of devices for which Google can pay for default status and a cap on revenue share to channel partners (ProMarket, Aug 12). Google has appealed liability and remedies; the DOJ and 35 states have cross-appealed asking for tougher relief. Oral argument is expected late 2026 / early 2027. Google still holds ~90% of search. The tail risk that matters is not a breakup — it is a strengthened remedy that severs the Apple default payment, which would hit high-margin query volume directly.

8.3 Insider activity — clean

Reviewing Form 4s over the last 30 days: Sergey Brin's August filings are conversions and gifts at $0 consideration, not open-market sales. Director/officer sales are trivial (Frances Arnold, 82 shares at $333.39; Marsida Saraci, 449 shares at $333.20). There is no insider distribution ahead of the Oct 26 print. That is a mildly positive signal, and worth stating because it is the opposite of what you would expect if management believed the paper-gain-inflated print marked a top.

9. The trade

10. Three scenarios into Oct 26

11. Macro overlay — the three levels that govern this position

• Rates: 10Y 4.734%, 30Y 5.273% at Friday's close. Above 4.75% on the 10Y, AI multiple compression re-engages and a 30x core-multiple name is the wrong place to hide. Back below 4.45% and the $410 target arrives faster than 1–3 months. Alphabet's own 6.500% 2066 coupon is now the cleanest single-security read on what long-duration AI capital actually costs.
• Credit: hyperscaler IG issuance is over 2x 2025 levels. Alphabet's $25B is a rounding error to Alphabet and a real supply event for the long end. If IG spreads widen while the 30Y holds above 5.25%, the whole AI-capex complex re-prices before any earnings data changes.
• Equities: S&P 7,674.37, Nasdaq 26,180.45 into a week with NVDA earnings plus July PCE on Wednesday and Fed Chair Warsh's first Jackson Hole keynote on Friday. Lose S&P 7,650 and the GOOGL entry fills at the low end of the $322–$350 zone — which is the outcome I would prefer.

12. Sources

• Alphabet Q2 2026 earnings release, 8-K Exhibit 99.1 (Jul 22, 2026)
• Alphabet Form 10-Q, period ended June 30, 2026 — MD&A and equity/ATM disclosure
• Alphabet Form 8-K, Aug 10, 2026 — $25B senior notes closing, tranche detail
• Alphabet Q2 2026 earnings call — CFO prepared remarks and Q&A
• Sundar Pichai, "The next chapter of our AI momentum" (Aug 5, 2026)
• DOJ appellate brief, US and Plaintiff States v. Google (Jul 28, 2026)

Not investment advice. Positions and views are the author's own and can change without notice. Reference prices are Friday Aug 21, 2026 regular-session closes; US equity markets were closed at time of publication.