AI Investment & Strategic Diligence · Forecasting Case · May 2026
A two-engine analysis of Search and Cloud, the 2026 CapEx pulse, and why the free cash flow trough is timing rather than terminal.
Bottom Line
Methodology & Sources
The findings below are not management commentary repackaged. The forecast is built from public-disclosure sources and structured so every assumption can be re-checked against the document it came from. Four methodology pillars hold the case together.
Alphabet's most recent 10-K, four sequential 10-Q filings, the Q4 2025 earnings release, the Q4 2025 earnings call transcript, and segment-level operating disclosures. No private data, no proprietary surveys.
Alphabet is modeled as two distinct businesses: Search/Services and Google Cloud, with Other Bets as optionality. Modeling them separately is the key analytical move, because AI affects them in opposite directions on margin and revenue.
Services and Search revenue growth, Services and Cloud operating margins, Cloud backlog, consolidated CapEx, D&A, and free cash flow. Every other input rolls up into one of these nine, which keeps the forecast from drifting on assumptions that don't actually drive the answer.
Every business claim used in the forecast (query growth, monetization rate, Cloud backlog conversion, CapEx pulse) maps to one specific KPI and one specific public-disclosure source, so any reader can trace an assumption back to where it came from.
The Situation
The investor question on Alphabet has shifted from "is Search durable?" to two finer questions running in parallel. Does AI raise or lower Search's terminal value? AI Overviews and Gemini change the cost per query and reshape the monetizable surface. Does Cloud's AI-driven growth cover the associated CapEx and depreciation? A $180B 2026 CapEx number is the single hardest valuation fact in the case, and it does not land cleanly in either segment.
The case approaches both questions through a two-engine business architecture: Search / Services and Google Cloud, with Other Bets (primarily Waymo) treated as optionality. A blended forecast would net an AI-driven margin headwind inside Services against an AI-driven revenue tailwind inside Cloud, hiding both directional calls. Keeping them separate also lets the forecast reflect that roughly half of 2026 ML compute is allocated to Cloud, matching CapEx burden to the revenue engine it supports.
The case study walks through the two-engine architecture, the driver logic that moves each KPI, the mechanism findings, the 2026 FCF trough, and the continuing-value framework that nets the AI tailwinds against the AI headwinds. Every forecast assumption is traced to a specific public-disclosure source: a 10-K filing, a quarterly earnings release, or management commentary on the call.
Business Architecture
Services is a high-margin, auction-based, incumbency-protected engine where AI changes unit economics. Cloud is a capacity-constrained infrastructure engine where AI simultaneously raises demand and raises CapEx and depreciation. Modeling them separately is the single most important methodological decision in the case.
Google Search · YouTube · Gmail · Network · YouTube Premium · YouTube TV · Google One · Pixel
Sells ad impressions and clicks produced by query-driven surfaces, plus a subscription and device tail. Advertisers pay CPC/CPM at auction; consumers pay subscription and device prices. Unit economics rest on query volume × ad load × CPC, net of Traffic Acquisition Costs.
Margins are driven by auction pricing, TAC share, and serving cost per query, which is where AI inference enters the cost function. Reinvestment has historically been modest but is now rising materially with AI inference compute.
GCP infrastructure · Workspace · Vertex AI · Gemini APIs · Agents
Sells consumption-based infrastructure, per-seat productivity, and AI solutions. Customers are enterprises and developers. Unit economics rest on consumption × unit price and backlog conversion, with margins driven by TPU/GPU utilization, pricing discipline, and mix shift to higher-margin Workspace and AI-solutions.
Reinvestment is the largest line in Alphabet's capital budget: data centers, accelerators, and sales capacity. The 2024 to 2025 margin step (14.1% to 30.7%) shows what scale absorption looks like once a CapEx wave is in place.
Primarily Waymo. 2025 segment revenue of $1.9B and operating loss of $8.6B are too small to move the consolidated story but absorb capital and are worth watching for commercialization milestones. Modeled as optionality, not as a steady-state contributor.
Driver Tree & KPI Spine
The forecast is built around a KPI spine: nine financial variables that, taken together, determine Alphabet's reported numbers. They are Services revenue growth, Search & other growth, Services operating margin, Cloud revenue growth, Cloud operating margin, Cloud backlog, consolidated CapEx, D&A, and free cash flow. Above the KPI spine sits a driver tree: each driver below maps one specific business claim to one of those variables and one source of disclosure, so the assumption can be re-checked against a 10-K, 10-Q, or earnings call quote. AI Overviews monetization rate and AI Search usage uplift are included as proxies because per-query economics are not directly disclosed.
Two independent data points from management: AI Overviews is "driving over 10% increase in usage" in the U.S. and India, and AI Mode "is driving incremental total query growth for Search." This driver moves the Services revenue growth line before any monetization call is required.
Two claims, both from management. AI Overviews monetize at "approximately the same rate" as traditional Search, eliminating the near-term cannibalization haircut in the base case. Gemini "increased our ability to deliver ads on longer, more complex searches that were previously challenging to monetize," widening the monetizable surface at the margin.
AI inference raises cost per query. This is an inference from segment margin disclosures rather than a direct disclosure, but it creates a structural floor slightly below the pre-AI Services margin, independent of the near-term D&A drag. Variable moved: Services operating margin, steady state at ~38% vs. the 40.7% 2025A peak.
The 2024 to 2026 CapEx pulse pins 2026 CapEx at the $180B guidance midpoint. The D&A resulting from that pulse creates near-term margin and FCF drag, with management explicitly telling investors that 2026 D&A "meaningfully increase[s]." Together these drive the 2026 FCF trough that anchors the valuation discussion.
Cloud demand is supply-limited, not demand-limited, by management's own characterization. Backlog of $240B at Q4 2025 covers ~3.4× 2025 Cloud revenue. This underwrites 40% Cloud growth in 2026 without an independent demand forecast. Cloud operating margin recovers from a 26% 2026 dip to 30% by 2028 as utilization absorbs new capacity.
Roughly half of 2026 ML compute is tagged to Cloud, per management commentary. This matters for segment margin optics: Cloud revenue partly funds the D&A burden that would otherwise land entirely in Services. Effective tax modeled at 17%, consistent with 16.4% in 2024 and 16.8% in 2025.
Key Findings · Mechanism
Five mechanism findings, each tied to an empirical cross-check from public disclosures. The empirical cross-checks are the load-bearing parts of the case: management commentary describes intent, the cross-checks confirm whether the intent is actually showing up in the numbers.
AI Overviews is "driving over 10% increase in usage" in the U.S. and India. AI Mode is "driving incremental total query growth for Search." Gemini "increased our ability to deliver ads on longer, more complex searches that were previously challenging to monetize." Most importantly, AI Overviews monetize at "approximately the same rate" as traditional Search per management. The Q4 2025 print rebuts the cannibalization thesis empirically: Search & other grew 17% YoY to $63.1B.
Cloud is a supply-limited growth engine. Backlog grew from $155B in Q3 to $240B at Q4 2025. Cloud revenue grew 48% YoY to $17.7B. Management "still expect[s] to remain in a tight demand/supply environment in Q4 and 2026." Backlog coverage of approximately 3.4× 2025 Cloud revenue underwrites the 40% 2026 growth assumption without a separate demand call. Operating margin stepped to 30.7% in 2025 from 14.1% in 2024, and should dip to roughly 26% in 2026 on new-capacity D&A before re-expanding.
AI inference raises cost per query, derived from segment margin disclosures. This creates a structural floor slightly below the pre-AI Services margin, independent of the near-term D&A drag. The knob it moves is the Services operating margin: steady state at approximately 38%, below the 40.7% 2025A peak. The gap between the peak and the steady-state assumption is the entire question of whether AI leaves Alphabet structurally worse off, and the answer is "modestly worse, not cliff."
Half of 2026 ML compute is allocated to Cloud, per management. This matters for segment margin optics because Cloud revenue partly funds the D&A burden that would otherwise land entirely in Services. The forecast allocates the CapEx pulse to the revenue engine it supports, rather than treating it as a Services-side overhead. Without this allocation, Services margins look worse and Cloud margins look better than the underlying business mechanics suggest.
Android, Chrome, and Search default agreements continue to anchor query share at the entry points where AI matters most. AI is changing what happens after a query, not whether the query reaches Google. The distribution advantage is the reason the cannibalization thesis is structurally weaker than it is for pure-search competitors that lack default-placement economics.
Capital Cycle
The hardest valuation fact in the case is the 2026 FCF trough. With 2026 CapEx at the $180B guidance midpoint and D&A "meaningfully increasing," the proxy FCF drops from $73B in 2025 toward zero in 2026. CapEx plateaus near $190B through 2027 and 2028, revenue outgrows D&A, and the FCF proxy recovers toward $50B by 2028. The capital-cycle compression is real, but it is timing.
What Would Change Our View
If any of the following move materially, the recommendation gets revisited. Each is a specific, observable disclosure event tied to a particular line in a quarterly filing, not a qualitative shift in narrative.
Sets near-term FCF and the 2026 to 2028 D&A path. A $20B upward revision would push the 2026 FCF proxy meaningfully negative and delay recovery by a year. The most directly observable variable in the case.
The Services margin floor is an inference from segment disclosures, not a directly disclosed number. The gap between the 40.7% 2025 peak and the steady-state assumption is the entire question of whether AI leaves Alphabet structurally worse off. Watch for explicit segment disclosure of inference compute cost.
The 40 / 35 / 30% 2026 to 2028 path is defensible on today's backlog but assumes backlog rebuilds as current contracts convert. A broader AI-capex pullback in 2027 (driven by hyperscaler peer behavior) would mis-size this line.
Continuing Value
| Continuing-Value Assumption | Mature State | Direction vs. 2025A |
|---|---|---|
| Revenue growth (g) | 3% | Below 2025 consolidated growth; reflects mix and law-of-large-numbers |
| Blended operating margin | ~33% | Below 2025 consolidated margin; reflects Services floor + Cloud equilibrium |
| CapEx / revenue | ~15% | Down from 2026 peak intensity; up from pre-AI baseline |
| Effective tax rate | 17% | Consistent with 16.4% (2024) and 16.8% (2025) |
Recommendation
The defensive-innovation reading dominates the self-cannibalization reading. Search monetization is holding up, and Cloud's AI-driven growth arrives before most of the new CapEx turns into depreciation. The 2026 FCF trough will suppress optics in real time, which, if the monetization thesis holds, is the opportunity rather than the risk.
What anchors the call: the empirical Q4 2025 Search & other print of +17% YoY to $63.1B; AI Overviews monetizing "at approximately the same rate" as traditional Search; $240B Cloud backlog covering ~3.4× 2025 revenue; the 14.1% to 30.7% Cloud margin step that demonstrates utilization absorption mechanics.
What would reset the call: $20B+ upward revision to 2026 CapEx, explicit disclosure of compressed Services segment margins driven by inference cost, or a broader AI-capex pullback in 2027 that mis-sizes the Cloud growth assumption.
What FutureInSites Brings
Investment-grade analysis of how AI changes the economics of incumbent platforms. The Alphabet case is the proof point for forecasting work where the valuation question is concentrated in continuing-value assumptions and the analytical move is keeping two engines apart instead of blending them.
Where AI changes unit economics on one side of the business and raises CapEx on the other, blending segments hides both calls. The Alphabet case rebuilds the architecture from segment disclosures and assigns CapEx burden to the engine it supports.
Cost per query, ad density within the SERP, monetization-rate benchmarking against management commentary, and segment-margin path modeling. The cost-side analysis that "AI hype" coverage tends to skip.
Long-query monetization through Gemini, Overviews monetization parity, and AI Mode usage uplift modeled as a forecast knob, not a press-release narrative. Quantified into the Services revenue growth line.
The 2026 FCF trough is the case's hardest valuation fact. Capital-cycle compression vs. structural deterioration is a distinction many AI forecasts collapse. We model the trough explicitly and let the recovery path speak for itself.
Every forecast assumption ties to a specific public-disclosure source and one of nine KPI variables that determine reported performance. AI Overviews monetization rate and AI Search usage uplift sit alongside as proxies. Auditable, not discretionary.
Where AI's net effect on terminal value is the actual question, the case explicitly enumerates which AI mechanisms raise CV and which lower it, then makes the net call. Volume and Cloud tailwinds vs. inference cost floor.