Alphabet Inc. (NASDAQ: GOOG) — Q2-2026 Update POSITIVEScenario Fair Value$425

Highlights

• Cloud is no longer the option — it is the earnings engine, four years ahead of our model.
  Google Cloud grew 82% to $24.8B with operating income of $8.8B — a 35.6% segment operating margin against the ~18% we carried in the initiation model for FY2026. Backlog rose $54B sequentially to $514B, with management guiding that more than half converts to revenue inside 24 months. We raise FY26E Cloud revenue from $88.1B to $104.0B and FY30E from $274B to $342B. This single line is the reason we upgrade.
• The headline EPS is an accounting artifact — and the core number was a slight miss.
  GAAP diluted EPS of $9.11 against $2.95 consensus reads as a 216% beat. It is not. Alphabet's own filing attributes $6.26/share to a $99.0B pre-tax unrealized gain on non-marketable equity securities, chiefly the markup of its ~14% Anthropic stake as that company's valuation roughly tripled. Core EPS was ~$2.85 — about 3% below consensus. Our model, our estimates, and our fair value all run on the core figure.
• The market sold the capex raise. We think it sold the wrong half of the print.
  FY26 capex went to $195–205B from $180–190B, 2027 was flagged "significantly" higher, and Q2 free cash flow printed −$5.9B — the first negative quarter in Alphabet's history. The stock fell 7.4% to $317. But the incremental $15–20B of annual capex buys a business now compounding at 82% with a 35.6% margin and $514B of contracted backlog. On our numbers the Cloud upgrade is worth materially more than the capex drag costs.
• Upgrade to Positive; Scenario Fair Value to $425 — with a deliberate haircut to the math.
  The five-method weighted blend now computes to roughly $450. We publish $425. The haircut is intentional: negative free cash flow and an open-ended 2027 capex signal are genuine execution risk, and we would rather be early to the rating than early to the number. At $317, $425 implies +34% — comfortably past our +15% Positive threshold even on the haircut figure.

FQ3-26 Results

Alphabet reported Q2 2026 on July 22 with revenue of $119.8 billion, up 24% year-over-year and about 2.5% ahead of the $116.9 billion consensus — the twelfth consecutive quarter of double-digit growth and an acceleration from Q1's +22%. The operating detail was better than the headline: Google Services grew 15% to $94.5 billion, and Google Cloud grew 82% to $24.8 billion, an acceleration of nineteen percentage points from the +63% we saw one quarter earlier.

The headline EPS figure requires an immediate correction. GAAP diluted EPS printed at $9.11 against $2.95 consensus, which the wires reported as a 216% beat. It was nothing of the kind. Alphabet's own filing discloses that a $99.0 billion pre-tax unrealized gain on non-marketable equity securities increased the tax provision by $21.9 billion, net income by $77.1 billion, and diluted EPS by $6.26. Strip that out and core EPS was approximately $2.85 — roughly 3% below consensus. The largest single component of the markup was Alphabet's approximately 14% stake in Anthropic, whose valuation roughly tripled during the quarter following a $65 billion funding round; Bank of America attributes around $80 billion of the $99 billion to that position alone. A second large contributor was the SpaceX stake. Neither is cash, neither is operational, and neither belongs in a forward multiple. Every estimate and valuation figure in this note runs on the core number.

[[chart:u01]]

Search re-accelerated. Google Search & other advertising grew 17% to $63.3 billion, up from +19% in Q1 on a materially harder comparison, with management citing strength in retail and financial services. Philipp Schindler framed the mechanism directly: Gemini "supercharges our ability… to understand what people are looking for and match the right ads." AI Mode passed 1 billion monthly active users in the quarter. For a thesis whose first pillar was that Search defends in the AI-Overviews era, this is the third consecutive quarter of confirming evidence — the disintermediation narrative continues to be contradicted by the reported numbers rather than supported by them.

YouTube advertising grew 13% to $11.1 billion, a modest acceleration from Q1's +11%, with connected-TV share gains continuing to offset premium-CPM pressure from the Netflix, Amazon, and Disney ad tiers. Subscriptions, platforms, and devices grew 15%, driven by Google One and Gemini Advanced.

But the quarter's defining number was Cloud's profitability, not its growth. Google Cloud produced operating income of $8.8 billion on that $24.8 billion of revenue — a 35.6% segment operating margin, against $2.8 billion of operating income in the year-ago quarter. Our initiation model carried Cloud at roughly an 18% segment operating margin for FY2026, stepping toward the high-30s only by the end of the decade. Cloud cleared that terminal assumption in a single quarter, four years early. Backlog rose $54 billion sequentially to $514 billion, and management guided that more than half of it converts to revenue within 24 months.

The cost side is where the market focused. Q2 capital expenditures were $44.9 billion. Operating cash flow of $39.1 billion did not cover it, and free cash flow printed at −$5.9 billion — the first negative free-cash-flow quarter in Alphabet's history as a public company. Management simultaneously raised full-year capex guidance to $195–205 billion from $180–190 billion and flagged a "significant" further increase for 2027. The stock fell 7.4% to $316.68 the following session.

Guide + Capex

The capex raise was the print's market-moving disclosure. Alphabet took FY2026 capital expenditure guidance to $195–205 billion from $180–190 billion — a $15 billion increase at the midpoint, against a Visible Alpha consensus of roughly $188 billion. Anat Ashkenazi delivered it plainly: "We are updating our full-year 2026 CapEx guidance range to $195 billion–$205 billion, up from our previous estimate of $180 billion–$190 billion." She then flagged that 2027 would see a further "significant" increase, without quantifying it.

[[chart:u02]]

The framing matters as much as the number. Ashkenazi tied the raise explicitly to return, not to a land-grab: the company will keep investing "as long as we see an attractive return on that investment," and noted that demand continues to outpace capacity investment across the industry. Pichai's version was longer-horizon: "It feels like we are in very early innings of what feels like secular shift across multiple areas." That is a materially different posture from a defensive build — Alphabet is describing capex as demand-pulled, and the $514 billion backlog is the artifact that supports the claim.

Three operating disclosures corroborate the demand story. First, model API throughput reached approximately 22 billion tokens per minute, up from 16 billion a quarter earlier — a 38% sequential increase in actual inference volume, not bookings. Second, more than 9 million developers are now building monthly across Gemini APIs and developer products, which speaks to ecosystem depth rather than a handful of large committed contracts. Third, management disclosed it is expanding third-party capacity as a bridge — leasing rather than building where speed matters. That last point cuts both ways: it means reported capex understates the true capacity Alphabet is bringing online, but it also means some of the economics leak to landlords.

Our capex assumptions move up across the horizon. We take FY26E to $200 billion (the guide midpoint, from $185 billion), FY27E to $230 billion (from $200 billion, reflecting the "significant" language), FY28E to $215 billion (from $190 billion), FY29E to $190 billion (from $170 billion), and FY30E to $175 billion (from $160 billion). Depreciation follows the larger asset base with a 5–7 year life, rising from $42 billion in FY26E to $112 billion by FY30E.

The free-cash-flow path is the honest cost of the upgrade. On the new numbers, FY26E free cash flow falls to $16.5 billion from $21.3 billion and FY27E to $44.6 billion from $50.6 billion. From FY28 the larger Cloud base more than compensates: FY28E rises to $122.8 billion from $110.0 billion, and FY30E reaches $286.9 billion versus $235.6 billion previously. The trough is deeper and the recovery is steeper. Investors who need the FCF line to inflect before they own the equity will be waiting until FY2028 on our numbers — and that is precisely the constituency that sold the stock on July 22.

What would change our mind. A FY27 capex guide above roughly $250 billion at the January print, without a commensurate step-up in backlog, would break the demand-pulled framing and force us to re-anchor on the perpetuity DCF. Conversely, a FY27 guide at or below $220 billion alongside backlog clearing $600 billion would pull our modeled FCF inflection into FY2027 and argue for a fair value well above the one we publish here.

Estimates Revised

We raise revenue and earnings across the forecast horizon and cut near-term free cash flow. The Cloud line does nearly all the work on the upside; the capex guide does all of it on the downside.

[[chart:u04]]

Google Cloud is the material revision. We take FY26E Cloud revenue to $104.0 billion from $88.1 billion — our prior estimate implied a second half below the first half, which the Q2 run-rate of $24.8 billion made untenable within days of the print. The out-years step up correspondingly: FY27E to $154.0 billion (from $125.0 billion), FY28E to $212.0 billion (from $168.8 billion), FY29E to $276.0 billion (from $219.4 billion), and FY30E to $342.0 billion (from $274.3 billion). That growth path steps down from +77% in FY26E to +24% in FY30E — still a steep deceleration, and still, we think, conservative against a backlog compounding $54 billion a quarter.

The deceleration assumption deserves defending, because it is where reasonable people will disagree. Cloud is growing 82% today. We model 77% for the full year, then 48%, 38%, 30%, and 24%. The case for faster is the backlog: $514 billion with more than half converting inside 24 months implies roughly $257 billion of contracted revenue recognised by mid-2028, against our combined FY26E–FY27E estimate of $258 billion — meaning our forecast is almost exactly the contracted book with no credit for new bookings. The case for slower is that hyperscale cloud has never sustained 50%+ growth past a $150 billion run-rate, and that enterprise AI budgets are being pulled forward rather than created. We have chosen to sit close to the contracted book and let new bookings be the upside.

Consolidated revenue rises on the Cloud upgrade, partially offset in mix. FY26E goes to $496.2 billion from $475.3 billion (+23.2% growth, from +18.0%), FY27E to $590.5 billion from $549.0 billion, and FY28E to $689.7 billion from $625.8 billion. Google Services we hold roughly flat in absolute terms — Search at $263 billion FY26E and YouTube ads at $44.5 billion, both anchored on the Q2 run-rates — which means Services falls from 79% of revenue in FY26E to 61% by FY30E purely through Cloud's compounding.

Margins move up modestly. We take blended gross margin to 59.8% in FY26E (from 59.5%) rising to 61.8% by FY30E (from 61.0%), reflecting Cloud's demonstrated 35.6% segment operating margin against the ~18% we had carried. Operating expenses scale with the larger revenue base: R&D to $74 billion FY26E (from $72 billion) and SG&A to $50.5 billion (from $48.3 billion). Consolidated operating margin reaches 34.7% in FY26E and 40.4% by FY30E.

EPS rises across the horizon. FY26E diluted EPS goes to $12.27 from $11.60, FY27E to $14.98 from $13.51, and FY28E to $18.38 from $16.08. By FY30E we model $26.40, up from $21.94. These are core figures — no equity-securities gains of any kind are embedded, in this quarter or any forecast quarter.

Free cash flow falls near-term and rises later. FY26E drops to $16.5 billion from $21.3 billion and FY27E to $44.6 billion from $50.6 billion as the $30 billion of incremental FY27 capex lands. From FY28 the revision turns positive: $122.8 billion versus $110.0 billion, then $209.8 billion FY29E and $286.9 billion FY30E versus $235.6 billion prior. The FCF margin path now runs 3.3% → 7.6% → 17.8% → 26.4% → 31.9%.

One model input we did change on judgment rather than disclosure. The mid-cycle normalized free cash flow anchoring our perpetuity DCF moves from $155 billion to $190 billion, scaled to the higher terminal earnings power (FY30E FCF of $286.9 billion versus $235.6 billion previously). That lifts the perpetuity DCF from $170 to $204 per share. It remains the most conservative of our five methods and carries the smallest weight, for the same reason it did at initiation: it discounts the FY26–27 capex valley heavily and gives no credit to the backlog beyond the explicit period.

Thesis Check

Our June initiation rested on five pillars. Q2 moved two of them, in opposite directions, and left three untouched.

Pillar 1 — Search defends in the AI-Overviews era. INTACT, and stronger. We initiated arguing that conversational AI assistants had not displaced Google for high-intent commercial queries, and that the ad pie was growing faster than any share leakage. Q2 delivered Search & other advertising up 17% to $63.3 billion against a harder comparison, with strength concentrated in retail and financial services — the two verticals most exposed to exactly the comparison-shopping queries the disintermediation thesis says should be leaking to ChatGPT and Perplexity. AI Mode passed 1 billion monthly active users. Schindler's framing — that Gemini "supercharges" ad matching — describes AI as a monetization accelerant rather than a defensive patch. Morgan Stanley's $15–25 billion at-risk estimate remains the right bear framing, but three consecutive quarters have now failed to produce evidence for it.

Pillar 2 — Cloud is the second pillar of the franchise. EXCEEDED, by a wide margin. This is the pillar that carries the upgrade. We initiated modelling Cloud at +50% for FY2026 with a segment operating margin around 18%, reaching the high-30s only late in the decade. Q2 printed +82% growth and a 35.6% operating margin simultaneously — the growth rate we thought was a bull case, at the margin we thought was a terminal assumption. Backlog rose $54 billion sequentially to $514 billion. Token throughput rose 38% sequentially to 22 billion per minute, confirming the backlog is converting into actual inference rather than sitting as unfunded commitment. Cloud operating income of $8.8 billion in a single quarter now annualises above $35 billion — larger than the entire FY2025 operating income of most S&P 100 constituents.

Pillar 3 — the FY26–28 capex cycle is the central valuation tension. INTACT, and the tension worsened. We wrote at initiation that the perpetuity DCF ($168 at the time) and the exit-multiple DCF ($398) bracketed a genuine disagreement about whether the AI build pays back, and that the market was pricing closer to the optimistic end. Q2 made the near-term arithmetic worse in exactly the way we flagged: capex guidance up $15 billion, 2027 "significantly" higher, and free cash flow negative for the first time in company history. Our modeled FY26E FCF falls from $21.3 billion to $16.5 billion. Nothing about this is a surprise in kind — only in degree. It is the reason we publish below our own weighted math rather than at it.

Pillar 4 — Waymo is the largest single-name optionality. UNCHANGED. No material disclosure in the quarter. The 1 million paid rides per week target for year-end 2026 stands, and the $126 billion February valuation is unrefreshed. Our sum-of-the-parts still carries no explicit Waymo credit beyond the $3–5 per share we referenced at initiation.

Pillar 5 — antitrust is the largest tail risk. UNCHANGED. No appellate development during the quarter. Google's May 22 opening brief and the DOJ/states February cross-appeal both stand; the DC Circuit timeline still points to H1 2027. Base case remains behavioral remedies confirmed, with structural Chrome divestiture at roughly 15% probability.

Where that leaves the thesis. At initiation we said the investment debate was whether the market would pay for the FY29–30 cash-flow inflection before the company printed it. Q2 sharpened both sides of that question: the inflection got larger and more credible, and the wait to see it got slightly longer and considerably more uncomfortable. The market resolved that trade-off by selling 7.4%. We resolve it the other way — because a 35.6% Cloud operating margin at 82% growth is evidence about the size of the eventual harvest, and negative free cash flow in a single quarter of a deliberate build is evidence only about its timing. Timing risk at a 21% lower price is a better proposition than it was in June.

What would break it. If Cloud growth decelerates below roughly 45% in either of the next two quarters while capex holds at the raised level, the demand-pulled framing fails and pillar 2 stops carrying pillar 3. That is the specific combination we are watching, and it is the one that would take us back to Neutral or lower.

Valuation Updated

We raise Scenario Fair Value from $385 to $425 and upgrade the view from Neutral to Positive. At the July 23 close of $316.68, that implies +34% upside — comfortably past the +15% threshold our initiation framework sets for a Positive view, and reached through a combination of a better model and a 9.3% lower share price.

The mechanical case for the upgrade arrived before the model change. Our June initiation published $385 against a $349 price, +10.4% upside, which sat in the Neutral band. The stock is now $317. Holding the old fair value entirely unchanged, the upside is +21.6% — already Positive territory. Everything the model revision adds is on top of a rating change the price action had already earned.

[[chart:u03]]

The five-method blend now computes to roughly $449. Rolling the next-twelve-months basis to FY2027 — appropriate now that we are past mid-year — the components are:

• DCF, perpetuity growth (15% weight). Mid-cycle normalized FCF of $190 billion (up from $155 billion, scaled to the higher terminal earnings power), WACC 9.8%, terminal growth 2.5%. Produces $204 per share, up from $170. Range across the sensitivity grid: $170–$290. This remains our most conservative anchor and the one that most punishes the capex valley — it is why the weight stays low.
• DCF, exit EBITDA multiple (30% weight). FY2030E EBITDA of $475.9 billion at an 18x exit, discounted four years at 9.8%, plus the explicit-period PV and net cash. Produces $547 per share. At 16x, $489; at 20x, $605. The multiple is unchanged from initiation — the value moves entirely because FY30E EBITDA rose from $398 billion to $476 billion on the Cloud upgrade.
• Trading comps, forward P/E (30% weight). Peer median 30x against FY27E EPS of $14.98 gives $449. The 25th–75th percentile band (26x–32x) spans $389–$479. GOOG at $317 trades on 21.2x FY27E — a discount to the mega-cap peer median for the first time since our coverage began.
• Trading comps, EV/EBITDA (15% weight). Peer median 22x against FY27E EBITDA of $275 billion, bridged to equity with $72 billion of net cash, gives $518. The 18x–22x band spans $423–$518.
• Probability-weighted scenarios (10% weight). Bull $600 at 25%, base $425 at 55%, bear $210 at 20% — a weighted $449.

Weighting those five gives $449.

We publish $425 — a deliberate haircut of roughly 5%. This is the mirror image of what we did at initiation, and the asymmetry is intentional. In June the weighted math said $374 and we rounded up to $385 to credit Cloud backlog and Waymo optionality that the model did not carry. Today the math says $449 and we round down to $425, because negative free cash flow and an unquantified 2027 capex step-up are real execution risk that the model's smooth deceleration curve does not fully capture. We would rather be early to the rating than early to the number. If the January FY27 capex guide comes in at or below $220 billion, the haircut comes off.

Scenario detail.

Bull — $600 (25%, up from 20%). Cloud holds above 50% growth into FY2028 as the $514 billion backlog converts faster than the contracted schedule, capex peaks in 2027 rather than extending, and free cash flow inflects positive during FY2028. The multiple re-rates as the harvest becomes visible rather than modeled. We raise the probability because Q2 was itself evidence for this path.

Base — $425 (55%). Cloud steps down 82% → 48% → 38% → 30% → 24% through FY2030E, roughly tracking the contracted book with no credit for new bookings. Capex peaks at $230 billion in FY2027 and tapers to $175 billion by FY2030E. Free cash flow troughs at $16.5 billion in FY2026E and recovers to $286.9 billion by FY2030E. Search compounds at low-double-digits, antitrust resolves behaviorally.

Bear — $210 (20%). Enterprise AI budgets prove pulled-forward rather than incremental; Cloud growth halves by FY2028 while the raised capex has already been committed. The FY2028–29 depreciation cliff outruns revenue and the perpetuity DCF — $204 today, lower on compressed terminal cash flow — becomes the operative anchor. A structural antitrust remedy would sit inside this case.

Where the current price sits. At $317, GOOG trades below the low end of three of our five valuation methods — the exit-multiple DCF ($489 low), forward P/E comps ($389 low), and EV/EBITDA comps ($423 low). It sits inside the perpetuity DCF range and inside the scenario range. Put differently: to justify the current price you must believe either that the perpetuity DCF is the right lens (i.e. the capex never pays back) or that the bear scenario is the modal outcome. We think both are defensible positions and neither is the most likely one.

Triggers. We would move back to Neutral above roughly $420, where the upside compresses inside the +15% band. We would add aggressively below $300, where the price implies the bear case is more likely than not. A FY27 capex guide above $250 billion without commensurate backlog growth, or two consecutive quarters of Cloud growth below 45%, would force a re-rating regardless of price.