Bernstein on AWS & Datadog: 'Just How 'Lumpy' Could 1Q25 Be?'
Bernstein analyst Mark Shmulik discusses AWS and Datadog following weaker-than-expected Google Cloud and Azure prints.
The analyst commented: "While Q4 AWS revenue growth was inline, the expectations bar had come in a touch post weaker than expected Google Cloud and Azure prints. Management commentary around a “lumpy” growth trajectory for 2025 has investors once again questioning the shape and durability of AWS growth. Like peers, AWS is AI capacity constrained with constraints only expect to ease in 2H25, but just how bad could it be?
Our web metrics pointed to AWS workload acceleration Q/Q in Q4, supported by increasing AI contribution and in-line with the ~19% Y/Y growth delivered. We held back this note, waiting to see the print and foreward looking commentary because… … the 1Q25 setup looks more difficult. In our prior notes we lay out our web metric SSO methodology along with channel checks to help us call the +1 quarter. We worried the SSO data might be off this quarter, as it points to a weaker Q1 vs. 2024 — 4Q24 Q/Q engagement growth went negative, a setup we haven’t seen since 1H23. If the data holds, it would imply 1Q25 ex-AI revenue growth of ~10.5-11% Y/Y. Q1 is also a tough comp lapping Leap Year and only a half quarter of IPv4 contribution, adding another ~140bps of headwind or a base growth rate of <10% Y/Y.
Could AWS decelerate in Q1? Yes, a floor of 15.5% Y/Y is possible. Based on backward correlation with Cloud Native linked names, like DDOG, we think AI has contributed ~$0.6B, ~$1-1.2B, ~$1.4-1.5B, and ~$1.7-1.89B in AWS revenues by quarter through 2024. This growth cadence correlates strongly to Azure’s AI growth, which faces similar capacity constraints. If correct, AWS is adding ~$400-500M of Y/Y growth in AI capacity each quarter. Based on our Azure regression, Q1 could see $2-2.2B incremental Y/Y. Adding AI contribution to our base model points to ~15.5% Y/Y.
WAIT! AI P*Q remains the wildcard, and may offer some upside. There may be a sneaky “get out of jail free” card for AWS: capacity constraints across all hyperscalers + scaling demand = spot prices skyrocketing. We note several H100 consumption prices have more than doubled by early February vs. end of December. If this pricing power continues through 1Q, incremental price-led AI contribution could be closer to $3B. Something we will watch throughout the quarter.
A tough read across for DDOG + other consumption infrastructure? YES! The AWS to Datadog correlation requires more assumptions, but we think it still has some directional signal. The AWS SSO weakness was part of the reason we reduced our Q1 model a couple of weeks ago. If the Price x Quantity issue does emerge, it may further depress demand for more applications going into production, and slow the hoped for tailwinds across the AI and Cloud Native linked consumption infrastructure stocks (e.g., databases, middleware like Confluent, CDN/internet security like Cloudflare, and consumption operational software like Datadog and Okta’s CIAM business). We note Cloudflare’s management told us in their callback Workers AI is not being used as fall-back capacity for customers frustrated by hyperscaler cost.
Amazon: After a streak of relatively “clean” quarters, Amazon’s Q4 results were a touch more complicated against a high bar — the quarter itself was strong, another OI beat, including a massive surprise with NA operating margins shooting up to 8%, and AWS margins still strong in the high-30s (hitting 40% excluding server useful life changes). However, AWS just met a recently lowered expectations bar and a tepid Q1 guidance with management’s “lumpy” commentary on AWS growth trajectory raises some new questions for investors to solve. We do note that AWS returned to the leadership position in terms of incremental new cloud dollar share capture in 4Q, thus we enter 2025 from a position of strength. We expect AWS growth will decrease sequentially in Q1, and remain weak in Q2 as supply constraints persist with limits on pricing power. As supply constrains ease in 2H25, we model AWS growth reaccelerating to ~19% Y/Y levels. Margins should remain strong throughout, even accounting for updated changes to the useful life of servers and network equipments. We believe that Amazon may trade sideways for a bit as investors get back into the rhythm of trading around third-party AWS data quantifying the “lumpiness” and rebuild out depreciation and ROIC tables. We rate Amazon Outperform, PT $275.
Datadog: we updated our model a couple of weeks ago, applying these early signals (here). And provided a deeper discussion on potential guidance disappointment vs. sell-side consensus last week (here). The Cloudflare earnings last week made us think our recently updated model may be about right — Cloudflare also correlates strongly to new application demand and Datadog. Our Q1 model for Datadog follows a similar expectation we have for Cloudflare in Q1 based on their guidance (note one difference: Cloudflare benefited from the Leap Year last year, while Datadog largely did not due to differences in license models). We are looking forward to hearing management discussion of expected demand in Q1, and are watching the behavior through Q1 by customers, driven by the higher inference price, to see if there is further downside to our previously updated model. For instance, our published model assumes Q2 is a bit more normal and has growth QoQ, but if the increased prices dampen the rate of AI projects going into production, then the quarter may be weaker. On the other hand, if model prices start falling fast enough, we could get an accelerated pace of applications into production, and perhaps even by Q2 see a more rapid rollout behavior. This note offers no additional changes to our model, $151 price target, or Outperform rating."