AI Capex Supercycle: 10 Charts on Cloud Revenue, ROIC, and the Q2 2026 Earnings Reset

AI Capex Supercycle: 10 Charts on Cloud Revenue, ROIC, and the Q2 2026 Earnings Reset | Gregg Carlson
CFO Insights · Technology & Capital Markets

AI Capex Supercycle: 10 Charts on Cloud Revenue, ROIC, and the Q2 2026 Earnings Reset

AWS · Azure · Google Cloud — 2020 to 2030E · ROIC vs. WACC · Capex productivity · Q2 2026 earnings beats and estimate revisions

Disclosures and limitations. This document is general financial and educational commentary for informational purposes only. It is not investment advice, a securities recommendation, a rating, a price target, or an offer or solicitation to buy or sell any security. Author Gregg Carlson holds personal long positions in AMZN, MSFT, and GOOGL. No rated equity research (BUY/HOLD/SELL + price targets) is published on these companies. Model projections are estimates, not guarantees. Consensus figures reflect named sources at a point in time and change. Past performance is not indicative of future results. Figure 9 stock prices are approximate author reconstructions from news sources — NOT a licensed price feed; do not use for trading. Current as of August 7, 2026; not updated.

This chart file is a companion to "Is the AI Capex Supercycle Earning Its Cost of Capital? A ROIC/WACC/EVA Analysis of Amazon, Microsoft, and Alphabet" published at gregg-carlson.com. The companion report contains complete methodology, ROIC framework, segment-test limitations, scenario analysis, and full disclosures. All figures have been cross-verified against the three author financial models and named public sources cited per chart.

1 — Combined AI Capex: $116B (2022) to $645B (2027E)

Amazon, Microsoft, and Alphabet grew combined annual capex roughly 4.5× in four years. All three raised 2026 guidance during the year. 2027E bars reflect company guidance signals and the author financial model estimates.

Figure 1
Figure 1. Stacked bars by company. Totals above each bar. Dotted line = actuals vs. estimates boundary. Microsoft = fiscal year aligned to calendar label.

2 — AWS, Azure & Google Cloud: Annual Revenue 2020–2030E

Three cloud businesses on a common revenue axis. Solid bars = actuals from SEC filings. Faded bars = post-Q2 2026 Wall Street consensus (2027–2030E). Azure: only FY2025 ($75B) and FY2026 ($100B) are management-confirmed; prior years are author estimates from disclosed growth rates.

Data confidence tiers: AWS and GCP = high (10-K segment revenue). Azure 2020–2024 = author estimates from Microsoft-disclosed rates — directionally reliable, not audited. 2027–2030E = named sell-side consensus midpoints.
Figure 2
Figure 2. Navy=AWS, teal=Azure, amber=Google Cloud. Faded bars = 2027-2030E estimates. 2026 = partial actuals + model/consensus.

3 — AWS: Revenue & Operating Margin, FY2015–2030E

AWS from first segment disclosure (FY2015) through 2030E. Margin cycle: 2023 optimization trough (27.1%), 2024 AI reacceleration (37.0%), 2025 step-back (35.4%), Q2 2026 record quarterly margin of 39.4%. The 2026E annual margin (36.0%) blends four quarters, two of which are estimated.

Correction note: Two margins were corrected after model cross-verification: 2021A from 29.0% to 29.8% and 2027E from 37.0% to 37.5% — both sourced directly from the AMZN model, verified against the FY2021 10-K.
Figure 3
Figure 3. Left axis: revenue bars (navy=actual, teal=2026E, amber=2027-30E model). Right axis: margin line (solid=actual, dashed=model).

4 — ROIC vs. WACC, 2017–2030E: Small-Base Spikes, Then Convergence

All three cleared WACC in every year shown except one. Amazon's 2017 ROIC of 327% and Microsoft's 2019 figure of 154% aren't operating miracles — both companies' invested-capital bases were still tiny relative to NOPAT that early, so the ratio spikes mechanically (Amazon's 2016 invested capital was barely $1B; Microsoft's FY2016 net cash position actually exceeded its combined debt and equity, making FY2017 ROIC undefined under this methodology — its line starts at FY2018). Amazon's 2022 is the one real exception to "everyone clears WACC": ROIC of 6.2% fell below its 10.17% WACC on a 54% effective tax rate that year, compounding an operating hit from post-COVID fulfillment overcapacity (North America and International segment operating income both went negative). NOPAT recovered to 20.9% (2023) and a peak of 34.2% (2024) before all three converge into the mid-20s by 2030E as the AI-capex build-out expands each company's invested-capital base faster than NOPAT can keep pace.

Method: ROIC = NOPAT(t) ÷ IC(t−1). IC = debt + equity − cash (excess-cash; no lease cap.). WACC: 4.3% RFR, 5.0% ERP; beta 1.20 AMZN/GOOGL, 1.13 MSFT. Full 2017–2025 series independently reconstructed from 10-K/earnings-release NOPAT and invested-capital inputs for all three companies; the large majority of inputs (EBIT, tax rates, debt, equity, cash) are sourced directly from filed 10-Ks. Amazon's series is fully source-verified across all years; a small number of Microsoft and Alphabet balance-sheet inputs in 2016–2017 are reconstructed rather than directly quoted.
Figure 4
Figure 4. Log scale. Dashed horizontal lines = each company WACC. Amazon's 2017 (327%) and Microsoft's 2019 (154%) peaks reflect small early-year capital bases, not exceptional operating performance; Microsoft's FY2017 is omitted as undefined (FY2016 net cash exceeded debt+equity). Amazon dipped below WACC in 2022 on an anomalous tax year plus post-COVID fulfillment overcapacity; all series clear WACC from 2023 onward.

5 — Capex Productivity: All Three Converge to 11.6–13.0%

For every dollar of cumulative capex spent over a window, how much incremental after-tax operating profit appeared by window-end? Not a cost-of-capital test — denominator is gross capex not invested capital, and WACC is shown for orientation only. Three separately-built models starting from different historical levels converge to a 11.6–13.0% band.

Figure 5
Figure 5. Solid bars = historical actuals. Faded = 2026E-2030E model projection. Percentage labels = relative change (projected ÷ historical − 1).

6 — Q2 2026 Earnings: Actual vs. Pre-Earnings Consensus

All three beat on revenue. AWS +4.2% vs. consensus at 37% growth (fastest in 18 quarters). Azure +3pp on growth rate (43% vs. 40.2%). Google Cloud: largest absolute beat of the cycle — $24.8B vs. $18.4B consensus (+34.5%). GOOGL EPS shown excludes the $53.4B Anthropic mark-to-market gain; reported GAAP EPS was $9.23.

Figure 6
Figure 6. Faded = pre-earnings LSEG/StreetAccount consensus. Solid = actual reported. Green/red % = beat/miss vs. consensus.

7 — Cloud Growth: Consensus vs. Actual vs. Next-Quarter Guidance

Google Cloud's 37pp beat (82% vs. 45%) is the largest cloud growth surprise in this earnings cycle. Azure guided 45% constant-currency for Q1 FY2027 — above the 43% just delivered. AWS gave no separate Q3 segment guidance; ~30% implied reflects the Prime Day timing headwind on total company revenue.

Figure 7
Figure 7. Faded = consensus. Solid = actual. +pp = percentage-point beat. Diamond = next-quarter Azure guidance only.

8 — Capex Guidance Revisions: What Changed at Each Company

The core market narrative: Alphabet raised $10B with no FCF clarity — stock −7%. Microsoft reported −$15B but it was an accounting reclassification (spending unchanged) — stock +9%. Amazon raised $20B alongside record AWS margins — stock +15%. Consistent rule applied: capex rewarded when margin evidence arrives alongside.

Microsoft note: The −$15B change ($190B to $175B) is an accounting reclassification only — extending data center useful life from 15 to 25 years shifts some future leases out of reported capex. CFO Hood: "underlying investment is unchanged."
Figure 8
Figure 8. Faded = prior guidance. Solid = post-earnings. Hatched = 2027E consensus/mgmt signal. Red/green labels = change.

9 — Stock Price Reaction: Event-Window Around Q2 2026 Earnings

Alphabet's arc — initial −7%, recovery above baseline within two weeks — is the most instructive: capex without margin confirmation produced a sell-off; the recovery came once Amazon and Microsoft's prints supplied it. Limitation: approximate reconstruction from news sources — not a licensed price feed.

Data limitation: Figure 9 is NOT from a licensed market data feed. Price paths are author reconstructions from published news (CNBC, WSJ, TheStreet, TradingKey). For directional illustration only. Do not use for trading decisions.
Figure 9
Figure 9. Approximate reconstruction. Days relative to each company own report date (GOOGL Jul 22; AMZN & MSFT Jul 30). Pre-earnings close = 100.

10 — Post-Earnings Re-Rating: Price vs. Street Targets, GOOGL, AMZN & MSFT

Amazon and Microsoft: current prices track well below Street average targets, leaving room to the upside. Alphabet: current price sits closest to its Street average, reflecting the round-trip recovery after the post-earnings decline. All three names trade below their Street high targets.

Figure 10
Figure 10. Gray = pre-earnings price. Navy = current price (early Aug. 2026). Orange = Street average target. Teal = Street high target (post-earnings). Analyst opinions; not guarantees of performance.

Data Audit Summary

✓ = verified against named source    ≈ = estimate    ! = approximate only

Data SeriesSourceStatus
AWS revenue 2021-2025AAmazon 10-K SEC filings✓ Verified
AWS op. margin 2021-2025AAmazon 10-K; model cross-check✓ Verified (2021A corrected 29.8%; 2027E corrected 37.5%)
Google Cloud revenue 2020-2025AAlphabet 10-K SEC filings✓ Verified
Azure FY2025 ($75B), FY2026 ($100B)CEO Nadella earnings calls✓ Management-confirmed
Azure FY2020-FY2024Author estimates from Microsoft-disclosed growth rates≈ Author estimates — Microsoft does not disclose Azure in dollars
AMZN/MSFT/GOOGL capex (actuals)Company 10-K cash flow statements✓ Verified
ROIC all companies 2017-2025AAuthor models vs. company 10-Ks✓ Verified — independently reconstructed from 10-K/earnings-release NOPAT and invested capital; corrected AMZN 2017–2024 from originally published figures after reconciliation; MSFT FY2017 omitted as undefined (negative invested capital)
Capex productivityAuthor models vs. article✓ Verified — match companion article
AWS 2027-2030E revenue (Fig. 2)GS/KeyBanc consensus midpoints≈ Consensus — higher than author model; intentional and labelled
Q2 beats (Fig. 6) all metricsCNBC/LSEG/StreetAccount✓ Verified against named sources
Capex guidance revisions (Fig. 8)Company earnings calls July 22-30✓ Verified
Stock price paths (Fig. 9)Author reconstruction from news! Approximate — NOT a licensed price feed; directional only
Analyst price targets (Fig. 10)Named: TheStreet, TIKR, TipRanks, Simply Wall St.✓ Named and sourced individually

Key Findings

  1. MSFT and GOOGL clear WACC in every year, 2017–2030E (MSFT from 2018); AMZN dipped below WACC in 2022. Amazon's 2022 ROIC of 6.2% fell under its 10.17% WACC — an anomalous 54% effective tax rate that year, compounded by post-COVID fulfillment overcapacity, drove the shortfall. All three clear WACC from 2023 onward.
  2. The spread is compressing, but Amazon's path was the least stable. MSFT and GOOGL glide down from small-capital-base highs early in the window (MSFT as high as 154% in 2019) to mid-20s by 2030E. AMZN swung from a 327% high in 2017 (also a small-base artifact) to 6.2% (2022) to a 34.2% peak (2024) before settling into the mid-20s with the other two — resilience from here depends on projected AWS margin expansion not yet fully realised.
  3. Capex productivity converges from wide historical dispersion (13–28%) to a tight 11.6–13.0% band — three separate models, same endpoint.
  4. Q2 2026 reset the market capex pricing rule: capex + margin evidence (AMZN, MSFT) = large positive moves. Capex without it (GOOGL initial) = sell-off, then reversal on peer evidence.
  5. Cloud revenue gaps closing. Consensus puts all three at USD307–382B by 2030E vs. ~USD160B gap in 2026.

Read the Full Report

These charts are a companion to the full analytical report, which includes the complete ROIC/WACC/EVA methodology, segment-test limitation, scenario ranges, Morgan Stanley framework comparison, the depreciation assumption debate, and the AWS $1 trillion sizing note.

→ Read the full report at gregg-carlson.com
Gregg Carlson

Gregg Carlson is a CPA and CFA Institute member with 25+ years of CFO and Controller experience across public companies, multi-state operators, and family offices. He has led $700M+ in M&A and capital raise transactions across gaming, cannabis, real estate, and technology. He provides fractional CFO and Controller services at gregg-carlson.com.

https://gregg-carlson.com
Previous
Previous

Downtown Las Vegas Has Been Here Before

Next
Next

Is the AI Capex Supercycle Earning Its Cost of Capital?