U.S. Internet & U.S. SMID-Cap Software cloud-consumption trends Report Interpretation

Overview

This data update examines whether AWS single-sign-on web activity is signaling a change in cloud-consumption demand. Bernstein argues that Q3 results should still benefit from strong earlier-quarter activity and favorable comparisons, but that persistent late-quarter softness could create a more meaningful Q4 drag in non-AI AWS revenue and related consumption software.

Core views

Bernstein's AWS SSO web metrics had signaled unusually strong cloud demand from Q3 2025 through Q1 2026, followed by reported AWS ex-AI revenue growth rising from an estimated roughly 14% year on year in Q2 2025 to an estimated 21.5% in Q2 2026. That acceleration supported cloud-consumption companies across databases, communications infrastructure, cloud operations, and application distribution. The institution now says that the slowing signal it first identified in May has persisted into early September. AWS web-metric growth has trailed the quarter-on-quarter levels seen in Q3 2025 throughout July and into early September, with the gap widening in the first two weeks of September, partly because Labor Day occurred later this year. Bernstein does not see this as a Q3 reporting problem. The web data were strong at the start of Q2, with weakness emerging only in May and June. Because the measure operates with about a one-quarter lead to reported financials and captures something closer to incremental annual recurring revenue added during a quarter, the institution believes quarter-on-quarter growth remains reasonably strong for Q3. A weak year-ago comparison, attributed to tariffs and DOGE, also helps the year-on-year comparison. Bernstein therefore expects AWS ex-AI revenue growth to accelerate by about 50 bps in Q3. The concern shifts to Q4. Although metrics improved after the seasonally slow U.S. July 4 period, Bernstein says the rebound was insufficient to offset the weak starting point. It needs another two weeks of September data before drawing firm conclusions, but estimates that, if current trends persist, the resulting year-on-year growth headwind could be up to 100 bps worse than currently embedded in its Q4 AWS ex-AI estimates. Its existing model assumes roughly a 150-bps Q4 headwind to AWS ex-AI year-on-year growth versus the 22% it expects in Q3. AI demand is the important offset. Bernstein believes a meaningful share of recent acceleration at AWS and Datadog came from AI revenue. It identifies OpenAI and Anthropic as Datadog's two largest AI-native customers and as useful indicators of broader AI-infrastructure demand. OpenAI's revenue grew 18% quarter on quarter in Q2, while its ARR reportedly reaccelerated by more than 20% month on month in July. Anthropic's ARR exceeded $65 billion in July, versus $47 billion in May and $9 billion at the end of 2025. Bernstein interprets this as evidence of accelerating token consumption across the AI-native ecosystem, which could support AWS and Datadog enough to reduce the effect of weaker traditional workloads on total revenue. In contrast, it sees a more pronounced Q4 revenue-growth headwind for consumption-linked names with limited AI exposure, including Twilio and Cloudflare. The report emphasizes that this is alternative data rather than a definitive forecast. Its historical fit has been good, but the indicator can disconnect from results because of shifts in pricing, credits, consumption per worker, sales pull-forwards, or changes in the vendor's collection methodology. Bernstein says a fundamental explanation remains necessary and is gathering additional evidence that could confirm or refute the current read-through. It made no changes to its models, price targets, or recommendations.

Analysis framework

Bernstein uses daily AWS SSO-page engagement data as a leading indicator of AWS ex-AI revenue, testing time shifts and finding the strongest historical correlation at roughly one quarter. It adjusts the relationship for factors such as rate-card changes, credits, rationalization, consumption per person, and sales timing, then compares current quarter-on-quarter patterns with prior-year levels and combines the signal with AI customer-demand indicators.

Methodology notes

• Industry AnalysisVolume-price decomposition

  Separating AI-driven and non-AI cloud-consumption revenue trends

  The report isolates AWS ex-AI revenue to assess whether traditional workload consumption is weakening, while separately considering AI demand as an offset to total growth.

• Other

  Lagged regression of AWS SSO engagement against revenue growth

  Bernstein tests daily SSO engagement at different time shifts and finds that prior-quarter engagement has the strongest relationship with current-quarter AWS revenue growth, after adjustments for commercial and operational factors.

• Valuation methodsDCF (Discounted Cash Flow)

  DCF valuation used in the covered-company target-price methodologies

  The report's disclosure appendix uses discounted cash flow alongside other valuation methods for the maintained price targets of Amazon, Datadog, Twilio, and Cloudflare.

• Valuation methodsSOTP (Sum-of-the-Parts) Valuation

  Amazon sum-of-the-parts valuation

  Amazon's $320 target uses a 50/50 blend of SOTP—valuing Retail and AWS separately—and DCF.

• Valuation methodsPS valuation

  Peer SaaS P/Sales valuation for Datadog, Twilio, and Cloudflare

  The price targets use P/Sales multiples for the next five to eight quarters, selected from peer SaaS regression work and operating-margin cohorts.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Amazon (AMZN)

  AWS is exposed to weaker non-AI workload consumption but may receive AI-revenue support.

  Strengths

  AI infrastructure demand may offset softness in traditional AWS workloads.

  Weaknesses

  AWS web metrics imply a potential Q4 ex-AI revenue-growth deceleration.

  Risks

  Regulatory actions, higher capital intensity from new investments, and share loss in e-commerce or AWS could pressure financials, margins, free cash flow, or valuation.

• Datadog (DDOG)

  Cloud-operations demand is linked to AWS consumption, with AI-native customers providing an offset.

  Strengths

  OpenAI and Anthropic momentum may support AI revenue and broader cloud-infrastructure demand.

  Weaknesses

  Traditional cloud-consumption weakness remains a risk.

  Comparison

  More AI exposure than Twilio and Cloudflare in Bernstein's discussion.

  Risks

  Competition, slower-than-expected margin expansion, execution issues, missed earnings, and a high valuation could create downside.

• Twilio (TWLO)

  A consumption-linked software company exposed to slower AWS-related traditional workloads.

  Strengths

  Potential upside from better unit economics through scale leverage or competitors failing to match its offering.

  Weaknesses

  Limited AI exposure may leave it more exposed to the Q4 consumption slowdown.

  Comparison

  Bernstein sees a more pronounced potential headwind than for AI-supported AWS and Datadog.

  Risks

  Slower synergistic M&A, weaker CPaaS growth among retail-exposed customers, competition, execution missteps, or missed earnings.

• Cloudflare (NET)

  An application-distribution and consumption-linked company exposed to slower AWS-related traditional workloads.

  Strengths

  AI-driven edge PaaS demand or broader cybersecurity-platform success could exceed revenue expectations.

  Weaknesses

  Limited AI exposure may make the non-AI cloud-consumption slowdown more consequential.

  Comparison

  Bernstein sees a more pronounced potential headwind than for AI-supported AWS and Datadog.

  Risks

  Cloud-growth dependence, sales-model transformation risk, limited consumption visibility, AI-product execution, hyperscaler competition, and high valuation.

Key data

• AWS ex-AI revenue growth~14% YoY in Q2'25 to ~21.5% YoY in Q2'26Estimated acceleration following earlier strong AWS web-metric signals.
• Q3 AWS ex-AI growth outlook~50 bps accelerationBernstein's expected year-on-year acceleration, supported by strong early-Q2 activity and a weak comparison base.
• Q4 AWS ex-AI estimate riskUp to -100 bps additional headwindPotential deterioration versus Bernstein's current Q4 estimate if late-Q3 web trends persist.
• Current Q4 AWS ex-AI growth headwind assumption~150 bpsVersus 22% year-on-year growth expected in Q3.
• OpenAI revenue growth18% QoQ in Q2; ARR grew 20%+ month-over-month in JulyUsed as an indicator of renewed AI-infrastructure demand.
• Anthropic ARROver $65B in July; $47B in May; $9B at end-2025Cited as evidence of rapid AI-native ecosystem growth.
• Amazon target price$320Outperform rating maintained.
• Datadog target price$237Market-Perform rating maintained.
• Twilio target price$199Market-Perform rating maintained.
• Cloudflare target price$164Market-Perform rating maintained.

Impact & implications

Bernstein's central implication is that a non-AI AWS-consumption slowdown may become more visible in Q4, while accelerating AI workload demand can cushion AWS and Datadog. Twilio and Cloudflare are portrayed as more exposed to the traditional cloud-consumption weakness because of their more limited AI exposure.

Risks

• The AWS SSO metric is alternative data and may cease to be predictive despite its historical fit.
• Changes in AWS pricing, credits, customer rationalization, consumption per worker, sales timing, or the data vendor's methodology could distort the relationship between web activity and revenue.
• Amazon faces regulatory, capital-intensity, competitive, and valuation risks.
• Datadog, Twilio, and Cloudflare face competition, execution, growth, margin, and valuation risks as specified in the report.

What to watch

• The remaining two weeks of September AWS web-metric data before drawing a firmer Q4 conclusion.
• Whether the Q4 web-metric gap versus 2025 persists or narrows after the Labor Day timing effect.
• OpenAI and Anthropic growth as indicators of AI-native workload and token-consumption demand.
• Whether AI revenue offsets weaker traditional consumption at AWS and Datadog.
• Whether slower AWS-related web metrics translate into Q4 revenue pressure for Twilio and Cloudflare.