With the megacaps rolling out prints, $Amazon(AMZN.US) posted its Q2 2026 results after the U.S. close on Jul 31 Beijing time. The numbers vindicate Dolphin Research’s take last quarter: it is AMZN’s turn to shine.

1) AWS is taking off? YoY growth accelerated from 28% to 37%, reaching $42.2bn. ARR from Anthropic, which spun out of AWS’s ecosystem, is surging and delivering visible tailwinds to AWS.

More importantly, unlike most cloud peers, AWS is accelerating both revenue and margin: OPM rose 170bps QoQ to 39.4%, near a record high.

2) Remaining performance obligations: RPO reached $496bn, up roughly $100bn QoQ. Dolphin’s back-of-the-envelope suggests $150bn in new bookings, well ahead of Microsoft.

The stock of backlog keeps building, and new bookings far exceed revenue recognized. As capacity comes online, AWS revenue growth should mechanically re-accelerate.

3) Capex is surging: Up nearly $10bn QoQ in a single quarter, the absolute level is the highest among the hyperscalers. Given AMZN’s retail is also asset-heavy, the portion actually allocated to cloud is likely similar to Google’s ~$45bn run-rate.

The full-year target was raised to $220bn from $200bn on storage price hikes. That implies ~$60bn per quarter in the next two quarters, signaling AI infra spend can keep dancing even after the pullback.

4) Cash flow has cracked: With only ~$45bn of operating inflow in the quarter and capex of $53bn this quarter heading toward $60bn, FCF will remain deeply negative.

Because cash flow has turned, management laid out a systematic ROI framework. This aligns with Dolphin’s prior read-through on Google’s spend in ‘The Trial of Google: Is the Market Fickle, or Is Heavy Investment a Sin?’

a) The company sizes build-out against ‘committed’ orders in hand, converting contracts into revenue as capacity goes live. b) Long-cycle assets like sites lift upfront outlays but depreciate over longer periods and remain usable, while short-cycle servers and network gear can be pre-bought based on demand; the current bulk orders reflect strong committed pipeline.

c) Payback for short-cycle assets like servers is about three years, with useful lives of 5–6 years, so FCF should emerge after year three.

That said, Dolphin flags risks around penalties on customer cancellations, supplier contract breaches, and duration mismatches between upstream POs and downstream contract balances. The logic is sound and improves transparency, but it does not erase all doubts.

At heart, the market still worries about the end-customers’ ability to pay. Budget reshuffles from AI-driven cost takeout are not enough; AI must drive incremental revenue for customers’ customers, and that evidence remains limited.

5) Will capex peak? Given AMZN’s self-run retail, Dolphin’s proxy of capex-to-gross profit is already ~50%. Many wonder if capex has topped out, as pushing further would mean deploying the bulk of revenue, not just profits.

Based on AMZN’s explanation, Dolphin believes capex planning is tied to hyperscalers’ orders in hand. As long as RPO keeps piling up, the capex story remains intact, but ultimately it depends on model advances to sustain bookings.

6) Will depreciation bite margins? Possibly. D&A was ~$20bn this quarter, ~10% of revenue, holding high for two quarters. This still trails the $50bn+ quarterly capex cash outflow, and with a higher mix of short-cycle assets like servers, depreciation pressure will rise.

Margin absorption will rely on faster revenue growth to dilute that burden.

7) Retail looked solid: With Prime Day pulled forward, retail beat Dolphin’s expectations, accelerating to +16% YoY. North America retail margin held at 7.9%, Intl retail at ~4.1%, a decent but mixed showing.

Cost inflation is not materially squeezing retail. Mix continues to improve as high-margin ads gain share, providing a persistent tailwind to retail GPM.

8) Strong quarter, middling guide: Next-quarter revenue guidance implies up to +12% YoY, with management citing the Prime Day pull-forward from Q3 to Q2. OP guidance of $22.5–26.5bn also does not top the Street.

On Dolphin’s math, this implies limited acceleration in cloud revenue and profit elasticity. However, AMZN typically beats guidance; the question is by how much, so treat the guide as a baseline.

Dolphin’s overall view

In short: retail is fine, cloud is powerful, guidance is so-so. But with cloud inflecting and AMZN’s conservative guides, the market is unlikely to fret the outlook.

The crux remains cloud ROI and the cash burn. Management’s framework speaks for the broader CSP cohort and does help reassure the market.

Management also highlighted two KPIs: annualized AI revenue and chip sales, each already at ~$25bn, with AI revenue growing triple digits. Last quarter, chip sales were annualizing at ~$20bn.

Dolphin’s rough split suggests AI contributed roughly half of AWS’s 37% YoY growth this quarter, with legacy workloads the other half. As noted last quarter, AMZN wants to tell a story in the inference era where in-house silicon cuts internal cost and drives external revenue, and while models lag, the hardware line is tracking well.

Bottom line, as inference, agents, and coding land, AMZN’s 2026 cloud strategy (Anthropic-linked cloud consumption + in-house silicon) is bearing fruit. AWS is entering a more certain upcycle, with the slope of acceleration driven by order accumulation and capacity ramp, while retail simply needs to avoid becoming a drag.

On valuation, given the spread between retail and cloud, Dolphin values them separately. Retail at 25x PE on 2026 profit growth of ~20% implies ~$700bn.

For cloud, assuming this year’s profit grows 45–50% on both revenue and margin gains, a 40x PE implies ~$2.3tn. Sum-of-the-parts is ~$3.2tn, or ~15% upside.

That equates to a blended ~25x PE vs. the current ~21x. With cloud margins converging and growth catching up to Microsoft and Google, there is room for multiple expansion.

The upside may not be massive, but visibility is high. Keep watching Anthropic’s ARR momentum as a corroborating signal.

Related charts:

1. AWS: clean revenue and profit acceleration

2. Capex surging vs. slower-rising D&A

(Note: capex excludes inflows from asset disposals)

3. Retail skin, cloud soul: profits rest on AWS

4. Retail: Prime Day pull-forward; ads shining; 1P solid; 3P and ads both strong

5. Opex: disciplined across the board except R&D

<End of text>

Prior Dolphin Research on AMZN:

Earnings takeaways

Oct 30, 2025 call ‘AMZN (Trans): Compute capacity to double again by 2027’

Oct 30, 2025 earnings take ‘AWS reversal: has AMZN finally turned the corner?’

Aug 1, 2025 earnings take ‘AWS stalling, retail risks rising — does AMZN need to crouch before the next leap?’

Aug 1, 2025 call ‘AMZN (Trans): 2Q capex run-rate a good guide for 2H’

May 2, 2025 call ‘AMZN (Trans): AI compute still bottlenecked’

May 2, 2025 earnings take ‘Heavy AI spend and tariffs: is AMZN back in a dip?’

Feb 7, 2025 earnings take ‘AMZN: leaning in on cloud capex — will margin release pause again?’

Feb 7, 2025 call ‘AMZN (Trans): GPU-cloud depreciation cycle near; DS a clear positive’

Risk disclosure and disclaimer: Dolphin Research disclaimer and general disclosure