AWS Decelerates Into the AI Capacity Wall — Capex Steps Up, Q2 Op-Inc Guide Lands Below Street, and the Market Asks Who Pays for the Buildout
Key Takeaways
- Top-line beat ($155.7B revenue +10% ex-FX, EPS $1.59 vs ~$1.36 Street) was driven by retail efficiency (re-architected inbound network, +8% units, NA op margin 6.3% even after one-time charges) and ad strength ($13.9B, +19% YoY) — the headline numbers were unambiguously good.
- AWS printed +17% to $29.3B — a 200bps decel from Q4's +19% — and missed the $29.4B consensus by a hair; Jassy explicitly said the business is capacity-constrained ("as fast as we put capacity in, it's being consumed") with GPU/motherboard supply still bottlenecked, framing the shortfall as supply-side rather than demand-side. Backlog $189B (+20% YoY) supports that read; 4.1-year duration.
- Q2 operating-income guide of $13.0B–$17.5B sits well below the ~$17.6–$17.8B Street midpoint, with Brian Olsavsky citing tariff cost absorption, Kuiper launch expense recognition, the seasonal SBC step-up, and a deliberately wider range "for the general uncertainty." The midpoint implies ~$15.25B — a ~14% YoY decline from Q2 2024's $14.7B is technically positive growth, but the wide band itself is the message.
- Q1 cash capex stepped to $24.3B, predominantly AWS infrastructure plus Trainium2 silicon — the company is doubling down into the AI buildout exactly as the AWS growth print suggests demand is outrunning supply. The $3.3B Anthropic-conversion gain in non-operating income flatters net income but is one-time and below-the-line.
- Rating: Initiating at Hold. The retail/ads engine is firing, AWS demand looks intact, and the $189B backlog argues the AI capacity story is real — but the soft Q2 op-income guide, an AWS-decel optic, and the still-rising capex curve into a tariff-uncertain consumer environment leave the risk/reward balanced rather than asymmetric. We want to see Q2 prove that the AWS reacceleration is on the way, not just being promised.
Results vs. Consensus
| Metric | Actual | Consensus | Beat/Miss | Magnitude |
|---|---|---|---|---|
| Revenue | $155.7B | ~$155.0B | Beat | +0.4% |
| EPS (GAAP, diluted) | $1.59 | ~$1.36 | Beat | +17% |
| Operating income | $18.4B | ~$17.5B | Beat | +5% |
| AWS revenue | $29.27B | ~$29.42B | Miss | −0.5% |
| AWS operating income | $11.5B | ~$11.0B | Beat | +5% |
| Advertising revenue | $13.92B | ~$13.74B | Beat | +1.3% |
| NA operating margin | 6.3% | ~6.2% | In line | (7.2% ex-charges) |
| Intl operating margin | 3.0% | ~3.4% | Miss | (3.7% ex-charges) |
Quality of Beat/Miss
- Revenue beat is narrow but real: ~$700M of upside on $155B is mostly retail unit strength (+8% units worldwide) and ad acceleration. Ex-FX growth was +10%; reported was +9%.
- EPS beat is amplified by a one-time non-operating gain: the $3.3B pre-tax Anthropic convertible-conversion mark-up sits in non-operating income. Underlying operating EPS quality is strong but the $1.59 headline overstates the run-rate.
- AWS revenue miss is supply-constrained, not demand-constrained — that is what management told us, and the $189B backlog (+20% YoY) corroborates. But the optic of a decel print into the most-watched line of the company is what matters into Q2.
- Retail margin headline was muddied by one-time charges (legacy customer-returns accruals + costs to receive tariff-pre-buy inventory). Excluding both, NA would have been ~7.2% and Intl ~3.7% — fundamentally a beat. Street noticed; the reaction reflects this.
Segment Performance
| Segment | Revenue | YoY (ex-FX) | Op Income | Op Margin | Notable |
|---|---|---|---|---|---|
| North America | $92.9B | +8% | $5.84B | 6.3% (7.2% ex-charges) | Inbound network productivity drove cost-to-serve down; +8% units |
| International | $33.5B | +8% (+5% reported) | $1.02B | 3.0% (3.7% ex-charges) | Margin expansion despite charges; FX a ~$1.4B WW headwind |
| AWS | $29.27B | +17% | $11.5B | ~39.5% | $117B run rate, $189B backlog +20% YoY, capacity-constrained |
AWS — The line everyone is watching
Amazon Web Services revenue of $29.27B was up 17% YoY — a 200-bps deceleration from Q4's +19% and a small miss vs. the ~$29.4B Street consensus. On the surface that is a problem; the company's most-scrutinized line decelerating in the same quarter Microsoft Azure printed +33% AI-driven and Google Cloud accelerated to +28% feeds directly into the "AWS is losing relative share in the AI cycle" bear narrative. Operating income was $11.5B — a 39.5% operating margin, the highest in several quarters — driven by Graviton/custom-silicon savings, network-design wins, and continued data-center power-usage optimization. Backlog of $189B is up 20% YoY with a 4.1-year weighted average duration, which is the leading indicator that the supply-constraint framing is the right one.
"Our AI business right now is a multi-billion dollar annual run rate business that's growing triple-digit percentages year-over-year. And as fast as we actually put the capacity in, it's being consumed. So I think we could be driving more revenue for the business if we had more capacity. We have a lot more Trainium 2 instances, and the next generation of NVIDIA's instances landing in the coming months. I expect that there are other parts of the supply chain that are a little bit jammed up as well — motherboards and some other componentry." — Andy Jassy, CEO
Jassy's framing is consistent: demand for AI workloads is outrunning the ability to land Trainium2 and NVIDIA P5/P6 capacity, and the company is bringing on "an increasing amount of capacity in the back half of the year." The key incremental data points are (a) Anthropic is training on Trainium2 — a marquee external proof point, (b) Trainium2 is claimed to deliver 30-40% better price/performance than NVIDIA GPU instances, and (c) Bedrock added Claude 3.7 Sonnet, Llama 4, DeepSeek R1, Mistral Pixtral Large, and Amazon Nova premier in the quarter. Customer wins listed in management's prepared remarks span Adobe, Uber, Nasdaq, Ericsson, Fujitsu, Cargill, Mitsubishi Electric, GDIT, GE Vernova, Booz Allen Hamilton, NextEra Energy, Publicis Sapient — a strong enterprise-direct mix that matters more than hyperscaler-peer comparisons in any single quarter.
Assessment: If the H2 capacity story lands as advertised, this print is a setup for a reacceleration. If it doesn't, AWS becomes the locus of a multi-quarter narrative reset. Backlog supports the bull view; the optic of an earnings-day miss does not.
North America — Inbound network compounding into margin
NA revenue of $92.9B was +8% YoY with operating income of $5.84B at a 6.3% margin (~7.2% excluding the historical-returns and tariff-pull-forward inbound charges). The story here is that the regionalized fulfillment network is now in its second major iteration: management redesigned the inbound network to feed regional hubs more efficiently, which raises the share of products available in each fulfillment center, lowers cost-to-serve, and shortens delivery distances. CFO Olsavsky said Q1 set "new delivery speed records with our fastest delivery ever for Prime members around the world" and "delivered more items same day or next day in Q1 than any other quarter in our history." Worldwide third-party unit mix held steady at 61% — the high-margin marketplace mix is intact.
Advertising revenue of $13.9B (+19% YoY) is a key driver of NA segment economics. Reach metrics — 275M+ U.S. ad-supported users — and the full-funnel framing (Prime Video, Twitch, IMDB, NFL/NBA/NASCAR, Amazon DSP) suggest the advertising line is increasingly competitive with the largest digital ad platforms in scale and product breadth. At ~9% of revenue but with significantly higher incremental margins, ad growth is doing real work on segment profitability.
Assessment: The retail thesis is operationally strong. Ex-charges, NA margin would have been a record. The risk is that Q2's tariff-cost absorption and seasonal SBC step-up obscure this trajectory in next quarter's print.
International — Margin progress, FX headwind
International revenue of $33.5B was +8% ex-FX (+5% reported). Operating income of $1.02B at a 3.0% margin (~3.7% ex-charges) continues the multi-year pattern of structural margin progress. The segment carries the bulk of the $1.4B WW FX headwind as USD strength persists.
Assessment: The structural margin trajectory remains intact. Currency is a translation drag, not an operational issue.
Key KPIs
| KPI | Q1 2025 | Q4 2024 | YoY | Trend | Read |
|---|---|---|---|---|---|
| AWS revenue growth | +17% | +19% | — | Decel | Capacity-constrained per mgmt |
| AWS backlog (RPO) | $189B | ~$157B | +20% | Accel | Demand visibility intact |
| AWS operating margin | ~39.5% | ~36.9% | +260bps | Up | Custom silicon + network savings |
| Advertising revenue | $13.92B | $17.3B | +19% | Strong | Full-funnel + Prime Video |
| Worldwide paid units | +8% | +8% | — | Steady | Demand resilient pre-tariff |
| 3P unit mix | 61% | ~61% | flat | Stable | High-margin mix preserved |
| Cash capex | $24.3B | $27.8B | — | Elevated | ~$100B+ implied for FY25 |
| FCF (TTM) | $25.9B | $38.2B | −32% | Compressing | Capex absorbing OCF |
Key Topics & Management Commentary
Overall Management Tone: Confident on long-cycle strategy, deliberately defensive on near-term tariff and capacity uncertainty, and unusually granular about which Q2 cost lines are stepping up. Jassy's voice on AI ("not even at the second strike of the first batter in the first inning") is the strongest he has been about the size of the AWS opportunity in any recent quarter. Olsavsky's Q2 commentary widened the operating-income range explicitly because of "the general uncertainty that we're seeing" — a tonal shift from the more confident Q4 print.
The AWS Capacity Bottleneck
The single most important sentence on the call was Jassy's "as fast as we actually put the capacity in, it's being consumed." Management is saying, on the record and twice, that AWS could be growing faster if the supply chain weren't bottlenecked at GPUs and motherboards. This is the framing that turns the +17% print from a deceleration story into a coiled-spring story — but only if the capacity actually lands in H2 as promised. Trainium2 is the marquee bet: claimed 30–40% better price/performance vs. NVIDIA GPU instances, Anthropic training on it, "starting to lay in capacity in larger quantities with significant appeal and demand." If the chip ramps and AWS reaccelerates in Q3/Q4, the supply-side framing was correct. If not, the bear case (AWS losing relative AI share) gets meaningful air.
"Before this generation of AI, we thought AWS had the chance to ultimately be a multi-$100 billion revenue run rate business. We now think it could be even larger." — Andy Jassy, CEO
Assessment: This is the central debate of the rest of 2025 for AMZN. Backlog growth (+20%) and the breadth of the Trainium2 + Bedrock + Anthropic relationship support management's framing. We are willing to give the supply-side thesis 1–2 quarters of evidence before re-rating — that is what makes Q1 a Hold rather than an Outperform.
Tariffs — The Defense, Not the Offense
Jassy spent more airtime on tariffs than on any single topic except AWS. The framing was defensive (we haven't seen demand softness yet, prices haven't materially moved, sellers haven't repriced) but with a strategic-positioning lens: Amazon's broad selection, 2M+ global seller base, and progressive component-sourcing diversification position it relatively better than concentrated-China retailers. Olsavsky added that Q1 included real costs to receive forward-bought tariff-anticipation inventory and that Q2 will include "tariffs that we'll be paying on retail purchases based on current tariffs … not large in Q2."
"We have not seen the average selling price of retail items appreciably go up yet. Some of this reflects some forward buying we did in our first-party selling and some of that reflects some advanced inbounding our third-party sellers have done, but a fair amount of this is that most sellers just haven't changed pricing yet. Again, this could change depending on where tariffs settle." — Andy Jassy, CEO
Assessment: Jassy is right that Amazon's structure (broad selection, marketplace diversity, component diversification) is relatively advantaged. But "we haven't seen demand attenuate yet" is a present-tense statement, and tariff settlement remains the single largest non-AI exogenous risk to the consumer business. Q2's "wider range" guide is the company saying so without saying so.
Capex Stepped Up, Free Cash Flow Compressed
Q1 cash capex of $24.3B is the highest first-quarter capex in Amazon history, and TTM FCF compressed to $25.9B from $38.2B at year-end. The bulk is AWS infrastructure (data centers, custom silicon Trainium ramp), with fulfillment/transportation network investment running alongside (rural delivery expansion: $4B through 2026). This is a deliberate choice. Management is accelerating into the AWS demand signal; the capacity-constrained narrative is, in part, an admission that 2024 capex wasn't enough.
Assessment: This is the right strategic posture if the AI demand curve is real. The risk is that capex compounds faster than AWS revenue, compressing FCF for multiple years before the curve catches up. We will be watching the AWS revenue / AWS capex ratio every quarter from here.
Alexa+ and Project Kuiper — Optionality, Not Yet Material
Alexa+ rolled out to 100K+ users in early access — Jassy was animated about the product (multi-step agentic actions, conversation flow without wake-word repetition, dinner-reservation booking). Free with Prime, $19.99/mo otherwise. Half-billion-device installed base provides immediate distribution. Project Kuiper had its first production-satellite launch this past week; commercial service planned later in 2025. Both are real; neither is material to the FY25 model.
Assessment: Optionality. Worth tracking as Q2/Q3 launch cadence builds. Kuiper launch costs will be expensed in Q2 (a known headwind to margin).
Anthropic Partnership Deepening
The $3.3B Q1 Anthropic-related non-operating gain came from converting a portion of Amazon's convertible Anthropic notes to non-voting preferred stock. More important than the gain itself: Anthropic is training its next generation of models on Trainium2 on AWS — the most prominent external validation of the chip stack. The Bedrock + Anthropic + Trainium2 stack is increasingly the central bull pillar for AWS, alongside customer breadth.
Assessment: The Anthropic relationship is now structurally important to AWS's AI positioning, not just a financial investment. The mark-to-market gain is non-cash and one-time, but the strategic implication is durable.
Guidance & Outlook
| Metric | Q2 2025 Guide | Implied YoY | Street pre-print | Read |
|---|---|---|---|---|
| Net sales | $159.0B–$164.0B | +7% to +11% | ~$161B | Midpoint in line |
| Operating income | $13.0B–$17.5B | −12% to +19% | ~$17.6–$17.8B | Midpoint well below |
| FX impact | ~10bps headwind | — | — | Marginal |
The Q2 op-income guide is the headline of the print. A $4.5B-wide range is unusual even for Amazon, and the midpoint of $15.25B is ~$2.4B below the Street's pre-print expectation. Olsavsky explained the components: (i) seasonal SBC step-up (carries through next 4 quarters), (ii) Project Kuiper launch costs (expensed until commercialization), (iii) tariff-related inventory cost absorption, and (iv) "general uncertainty" about consumer demand and the macro. He added that "some strength in April based on what could end up being some pre-buys" was an offset. The width of the range is the company saying it has lower visibility than usual, even with eight weeks of the quarter already in the book.
Implied Q2 op-income decline: at the low end, ~$13B is below Q2 2024's $14.7B (a ~12% YoY decline). At the high end, ~$17.5B is +19% YoY. The midpoint ~$15.25B is +4% YoY — meaningfully decelerating from Q1's +20%.
Street at: ~$17.6–$17.8B pre-print; will recalibrate to roughly $15.5–$16.0B post-guide.
Guidance style: Conservative-cautious. Amazon historically beats the high end of its op-income range; the wide band suggests the company is buying flexibility rather than telegraphing weakness. But the market doesn't always parse the difference.
Analyst Q&A Highlights
AWS Capacity and AI Reacceleration Timing
- Ross Sandler, Barclays: Asked whether AWS would be in a position to "capture enough AI revenue to drive acceleration" this year given capacity constraints. Jassy said the AI business is "multi-billion dollar annual run rate growing triple-digit YoY," that capacity is being consumed as fast as it lands, and that "supply chain issues and the capacity issues will continue to get better as the year proceeds." Notably did not commit to an in-year reacceleration.
Assessment: The non-commitment is the data point. Management would commit if they had visibility; they don't. - Justin Post, Bank of America: Asked about AWS revenue lumpiness and growth-rate gap vs. competitors. Jassy gave a structural answer (sales-cycle lumpiness in enterprise migrations + AI demand spikes that are hard to forecast) and pivoted to base-rate effect: "17% on a $117B run rate is still pretty significant growth." Assessment: Polite reframing of the relative-growth question. Doesn't address it.
- Brent Thill, Jefferies: Asked for the backlog number and whether enterprise migration is springing back or stalling. Dave Fildes confirmed $189B backlog (+20%, 4.1-year duration). Jassy's answer was that "enterprises realize they need to do both" — AI pilot work AND infrastructure modernization. Migration deals are "a multiyear process" but conversation conversion is "meaningful." Assessment: Backlog is the highest-quality data point on the call. Migration commentary is consistent with a re-acceleration thesis but doesn't provide near-term timing.
Q2 Operating-Income Guide Composition
- Eric Sheridan, Goldman Sachs / Brian Nowak, Morgan Stanley: Both pressed on the components of the soft Q2 op-income guide. Olsavsky cited (a) Q2 SBC step-up (sets the rate for the next 4 quarters), (b) Kuiper launch expenses, (c) tariff-pay-through, and (d) widened range "for the general uncertainty." Notably did NOT call out incremental AWS investment as a major driver of the wider range.
Assessment: The carve-up suggests this isn't a demand miss being telegraphed; it's a deliberately conservative guide. But the Street's number was ~$17.7B; even the high end matches that. Midpoint disappointment is real.
AWS Margins at ~40% — Sustainable?
- Doug Anmuth, JPMorgan: Asked about the drivers of AWS's ~40% margin. Olsavsky: software/process improvements optimizing server capacity, custom networking gear, data-center power efficiency, Graviton custom silicon. Said margins will "continue to evolve" as Gen-AI services mix grows and infrastructure investment lands in H2. Assessment: ~40% is the cyclical high. H2 capex landings will compress this number — that is a known dynamic, not a surprise.
Tariff Strategy
- Brian Nowak, Morgan Stanley: Asked about operational priorities to make sure Prime Day, Thanksgiving, and the holidays go smoothly amid tariff uncertainty. Jassy: support 3P sellers, manage inbound inventory carefully (don't over-stock fulfillment nodes), and lean on broad selection and seller diversity. Olsavsky confirmed Q2 will include some tariff cost absorption "not large in Q2."
Assessment: Operational answers, not strategic surprises. The 2M+ seller base is a real durable advantage in a high-tariff regime.
Alexa+ Adoption
- Doug Anmuth, JPMorgan: Asked how Amazon shifts user behavior on Alexa+. Jassy emphasized agentic action capability (booking restaurant reservations, controlling smart-home devices in sequence), distribution via 500M+ device installed base, and the "rewiring" effect of conversational interaction without wake-word repetition. Assessment: Anecdotal but enthusiastic. Inflection point if usage scales; immaterial if it doesn't.
What They're NOT Saying
- FY25 capex framework: Management has not formally re-stated a full-year capex number this print. Q1 came in at $24.3B; if H2 capacity ramp is real, full-year could be $105–115B, materially higher than the ~$100B baseline most models carried into print. The omission lets the Street triangulate but is conspicuous.
- AWS Q2/Q3 growth path: When Sandler explicitly asked when AWS could reaccelerate, Jassy declined to give a quarter. That's the single most important commitment management could have made and didn't.
- Trainium2 unit economics: Strong qualitative claims ("30–40% better price/performance") but no revenue-attribution disclosure. We don't know what fraction of AWS AI run-rate is Trainium2 vs. NVIDIA today.
- FTC antitrust trial timeline: No mention. Trial status remains a tail risk that doesn't surface in earnings commentary; that is consistent with prior quarters but worth flagging.
- Anthropic gain decomposition: The $3.3B is disclosed as conversion-of-convertible-notes-to-non-voting-preferred. The implied valuation step-up is not broken out — investors have to back-fit.
- Holiday positioning under tariffs: Jassy answered Brian Nowak's question with general principles but did not commit to a specific tariff-scenario operating-income preservation strategy. The Q4 print will tell us whether the answer was complete.
Market Reaction
- After-hours move (May 1): AMZN traded down ~2% in extended hours immediately after the release. Initial pop on the EPS beat was reversed by the Q2 op-income guide.
- Next-day (May 2) early indications: Stock moderately weaker pre-market; expected to give back ~1–3% on session open as Street recalibrates Q2 op-income to ~$15.5–$16.0B.
- Volume: Elevated — earnings always print high-volume sessions for AMZN.
- Analyst reactions (within first 12 hours):
- Generally maintained ratings; price targets nudged down by ~2–5% on most desks reflecting the Q2 op-income reset
- Bull camp emphasized $189B backlog, Trainium2 ramp, and the supply-side framing of the AWS print
- Bear camp emphasized the AWS decel optic vs. Azure/GCP, the wider Q2 op-income range as a confidence signal, and the rising capex curve compressing FCF
The price reaction is a clean read of the print: the EPS beat and retail strength are not enough to overcome the AWS-decel optic and the soft Q2 op-income guide in a market that is pricing AI infrastructure spend with rising scrutiny. The $3.3B Anthropic gain is being correctly discounted as one-time.
Street Perspective
Debate: AWS — Supply-side miss or demand-side decel?
Bull view: $189B backlog (+20%), capacity-constrained per management, Trainium2 ramping, Anthropic training on it, H2 capacity landings will reaccelerate growth. The AWS print is the bottom of a supply-driven trough.
Bear view: Azure printed +33%, GCP +28% — AWS is structurally underperforming AI hyperscaler peers in a moment where AI is the only growth that matters. Management has been promising capacity relief for two quarters; the chip-supply-chain answer is starting to feel like a deflection.
Our take: Backlog growth and Trainium2's Anthropic anchor strongly support the supply-side framing. But "trust me, capacity is coming" needs to convert to a Q2 or Q3 reacceleration print to retire the bear case. Probabilistically, we lean bull — but until evidence lands, the market is right to keep the multiple in check.
Debate: Capex — Discipline or build-and-they-will-come?
Bull view: Q1 capex of $24.3B funds backlog conversion. AWS operating margin at ~40% means each marginal dollar of AI capacity is highly accretive once filled. Capex peaks are a feature of cycle leadership, not a bug.
Bear view: TTM FCF is now $25.9B, down from $38.2B — capex is outrunning OCF growth. If AWS revenue doesn't reaccelerate, FCF compression continues, and at some point the market starts demanding visible-return-on-capex disclosure.
Our take: Right framework is AWS-revenue-per-AWS-capex over the next 4 quarters. Q1 dollar capex was high but the AWS RPO grew $30B+ YoY — that ratio still works. We'll re-evaluate quarterly.
Debate: Tariffs — Structurally advantaged or exposed?
Bull view: Amazon's 2M+ seller base, broad selection, multi-year component-diversification, and 1P forward-buy capability mean it weathers tariff regimes better than most retailers. CEO is on offense.
Bear view: 60%+ of GMV is via 3P sellers, many of which are China-direct. If tariffs settle at high levels, ASP rises broadly and demand attenuates, NA segment revenue and margin both compress.
Our take: Amazon's structural position is genuinely advantaged versus single-source retailers. But "advantaged" doesn't mean "immune." Q2 will start to show the first meaningful tariff cost-absorption signal. We need to see one quarter of post-tariff data before re-rating.
Model Update Needed
| Item | Pre-print model | Suggested change | Reason |
|---|---|---|---|
| FY25 revenue growth | +10% | +10% (unchanged) | Q1 in line; Q2 guide implies similar pace |
| FY25 op income | ~$80B | ~$74–78B | Q2 guide reset; H2 SBC + Kuiper expense |
| AWS FY25 growth | +18% | +17–18% | Q1 print + capacity-bound H1, H2 reaccel |
| AWS FY25 op margin | 37% | 36–37% | H2 capex landings compress; Q1 was peak |
| FY25 capex | $100B | $105–115B | H2 capacity push; AI capex tailwind |
| FY25 FCF | $45B | $30–40B | Capex absorbing OCF growth |
| Advertising FY25 | +18% | +19–20% | Q1 +19% + full-funnel breadth |
Valuation impact: ~$2–4 lower fair value on FCF compression; offset by ~$1–3 from raised AWS LTV reflecting backlog growth and Trainium2 optionality. Net target moves modestly lower; rating-justifying repricing requires Q2 evidence.
Thesis Scorecard Post-Earnings
| Thesis Point | Status | Notes |
|---|---|---|
| Bull #1: AWS reaccelerates in 2025 on AI demand | Neutral | Q1 decel, but $189B backlog and capacity-bound framing keep thesis alive |
| Bull #2: Retail margin expansion compounds via inbound/regionalization | Confirmed | Ex-charges NA margin ~7.2% — a record |
| Bull #3: Advertising compounds at 18%+ | Confirmed | +19% YoY, full-funnel breadth |
| Bull #4: Trainium2 + Anthropic anchor durable AWS AI position | Confirmed | Anthropic training on it; capacity ramping |
| Bear #1: AWS losing relative AI share to Azure/GCP | Challenged | Optic supports it; backlog refutes it |
| Bear #2: Capex compresses FCF for multiple years | Confirmed | TTM FCF down $12B YoY; trend continues |
| Bear #3: Tariffs compress NA margin and demand | Neutral | Q1 minimal impact; Q2 will be the first real read |
| Bear #4: FTC antitrust overhang | Neutral | No update; tail risk persists |
Overall: Thesis unchanged on net. Retail and ads outperformed, AWS is in a wait-and-prove-it window, capex curve is the cost of capturing the AI cycle.
Action: Hold. Initiate at Hold pending (a) Q2 AWS reacceleration signal and (b) Q2 NA margin holding ex-tariff costs.