Story

The Full Story

For a quarter century the story was "we organize the world's information, and advertising accidentally pays for everything." That story survived the Google-to-Alphabet reorg (2015), the "Year of Efficiency" cost reset (2023), and the generative-AI scare of late 2022. What finally changed it was compute: from FY2023 to FY2025 Alphabet tripled capital expenditure to $91.4B, re-segmented the narrative around a "full-stack AI" company, and saw Cloud operating income cross from a $1.9B loss to $14B+ profit. Management credibility, battered by the Bard demo in 2023 and the August 2024 DOJ search-monopoly ruling, has since rebuilt: operating margin is back to 32%, Gemini 3 shipped to 15 half-billion-user products, and an adverse remedies ruling in September 2025 turned out far lighter than feared. The question now is not whether Alphabet can participate in the AI era but whether $91B/yr of capex produces durable economics before Search monetization is re-papered for an AI-answer world.

1. The Narrative Arc

Alphabet's IR story has moved through five clearly distinguishable chapters. The text below the timeline is deliberately short — the visuals carry the arc.

The shape of this curve is the story. Two separate accelerations — ads through ~2011, then mobile + YouTube + Cloud stack from ~2015 — give way to an AI-era capex inflection starting 2023. Every inflection maps to a narrative pivot management eventually had to explain.

2. What Management Emphasized — and Then Stopped Emphasizing

The 10-K Business section is where management's talking points are durable enough to count. Reading FY2020 → FY2025 side by side, the shift is unmistakable: "machine learning" and "moonshots" yield to "full-stack AI," "Gemini," and "TPUs." COVID-era references disappear without comment. "Other Bets" prose shrinks year over year.

Three patterns that matter:

• COVID and DEI language did not decline gradually — both essentially disappeared between FY2022 and FY2024. DEI survived a full-page treatment in FY2020–FY2022 and was gone by FY2025. Sustainability prose shrunk similarly. Management never explained this; it is a quiet pivot.
• "Moonshots" language dipped in 2023–2024 as AI moved from Other Bets to the core. The FY2025 10-K actually says moonshots "are now incorporated into our core products" — reframing AI as the hit rather than the bet.
• "Capital expenditures" is the one word whose frequency and dollar-count both compounded every year. FY2020: $22.3B. FY2024: $52.5B. FY2025: $91.4B. There is no equivalent Meta-style "Year of Efficiency" at Alphabet after 2023 — spending is only back-loaded.

3. Risk Evolution

Risk factors drift slowly, which makes the drift informative. Below, I map eight recurring risk clusters against the six 10-Ks in the dataset.

The story the heatmap tells:

• AI risk went from a paragraph to the top of the deck in two years. In FY2022 the 10-K treated AI as one flavor of "new products" risk. FY2024 added a dedicated risk factor — "Issues in the development and use of AI may result in reputational harm and increased liability exposure" — and FY2025 expands it further.
• Antitrust language escalated from rote to specific. FY2022: general regulatory-scrutiny prose. FY2024: named exposure to the DOJ Search case (the Mehta ruling landed in August 2024). FY2025: explicit discussion of remedies, Digital Markets Act (EU), and Japan's smartphone competition act. This is the only risk that materially transformed the 10-K.
• Capital intensity moved from implicit to explicit. FY2020 mentioned capex was material. FY2025 has an entire MD&A paragraph framing the $91.4B number, plus new language about US tax changes (immediate R&E expensing + accelerated depreciation) that soften the cash impact. Management started pre-selling the capex case rather than defending it post hoc.
• Disappeared quietly: the word "COVID" is gone from risk factors; there is no "pandemic-related uncertainty" language after FY2022. Supply-chain risk stayed elevated but shifted in framing from pandemic to tariffs/trade.

4. How They Handled Bad News

Alphabet has had three genuinely adverse events since the data begins — a rare sample size for a company this large. The pattern is consistent: minimize in the moment, re-frame within a quarter, commit more money to the problem.

One short quote earns its place here. From the FY2023 10-K Business section:

"Since 2016, we have been an AI-first company."

Why it matters: that sentence, appearing for the first time in the FY2023 filing as a standalone claim, is a retroactive re-anchoring. The 2016 I/O keynote used the phrase, but in ten years of 10-Ks before FY2023 the word "AI-first" does not appear as a corporate identity claim. Management used the ChatGPT-shock moment to rewrite the priors. It worked — by FY2025 nobody questions that framing.

5. Guidance Track Record

Alphabet deliberately does not give quantitative forward guidance on revenue or EPS. What management does commit to — explicitly or implicitly — are strategic promises: capex envelopes, margin trajectories, Cloud profitability, AI product shipping timelines. Those are judgable.

Credibility score: 8.5 / 10. Most strategic commitments delivered, often ahead of bearish expectations (Cloud profitability, Gemini ramp, operating margin rebuild, Waymo). The one clean miss is the FY2025 capex guide — $75B was announced with conviction in January and the actual number was $91.4B by December. The miss cuts both ways: it shows demand is real but also that management's visibility into its own infrastructure cycle is worse than it sounds on calls. The score is not a 9.5 because Alphabet rarely gives hard numbers, so there are fewer promises to judge — and because the "never apologize" pattern means that on the rare occasion management is wrong, the reader has to reconstruct the miss themselves.

6. What the Story Is Now

The current story is simpler than it has been in fifteen years: Alphabet is a full-stack AI company whose legacy ad monopoly is funding a multi-year build of AI compute, models, and agent surfaces — and the DOJ has declined to break any of it.

What the reader should believe, and what to discount:

• Believe: Cloud is a real, profitable, capacity-short business and will be a $100B+ revenue segment in the next cycle. Gemini is competitive at the frontier, not two years behind as the 2023 narrative feared. Operating margin 30%+ is sustainable at current mix, even with $91B capex, because Services revenue still grows ~12%/yr.
• Discount: the implicit claim that AI Overviews preserves the Search economics. Management has not published the comparative monetization data and nobody should assume identical yield per query. Discount the "capex is self-funding from operating cash flow" framing at face value — it is true today but depreciation peaks in 2027-2028.
• Watch: Q1 2026 commentary on Search click volumes in AI-Overview-heavy queries. That single disclosure, or its absence, will tell you whether the story is simpler or more stretched than it looks.