The hyperscaler scoreboard that landed after the latest round of earnings is blunt: Amazon Web Services (AWS) delivered a respectable rebound — $33.0 billion in quarterly revenue and roughly 20% year‑over‑year growth — but the momentum picture is dominated by competitors posting faster sequential and percentage gains and much larger forward backlogs, leaving AWS playing catch‑up at the very moment the cloud market is being redefined by AI.
Background / Overview
The past few quarters have reshaped how enterprise buyers and investors evaluate cloud vendors. Where scale and absolute revenue once decided the narrative, the AI era has shifted attention to three additional metrics: (1) growth in AI‑related revenue, (2) the composition and convertibility of contract backlog or RPO (remaining performance obligations), and (3) the ability to productize AI into turnkey, enterprise‑grade services that reduce time‑to‑value for customers. These shifts favor vendors that can combine model stacks, managed hosting, purpose‑built accelerators and integrated software monetization — and the latest results show Microsoft, Google and Oracle pulling ahead on some of those axes.This feature analyzes the headline numbers, the forward‑looking signals embedded in RPO/backlog figures, and the strategic product and operational differences that explain why AWS’s Q3 acceleration looks good in isolation but troubling in context.
The quarter in numbers: revenue, growth and backlog
Headline revenue and growth
- AWS: $33.0 billion in quarterly revenue, roughly +20% year over year — a notable acceleration from recent quarters and the figure that produced the market relief rally.
- Microsoft Cloud (Intelligent Cloud and related cloud revenue): reported at roughly $49.1 billion for the comparable quarter, with strong percentage expansion that translated into larger sequential dollar adds than AWS.
- Google Cloud: $15.2 billion in quarterly revenue, up about mid‑30s percent year over year — the smallest absolute base among the three but the fastest percentage growth in recent quarters.
Backlog / RPO: the forward view that matters
Quarterly revenue is backward‑looking; RPO/backlog shows contracted future business and therefore signals convertibility of demand into revenue. The latest round produced stark contrast across the hyperscalers:- Oracle reported an RPO/backlog that surged to about $455 billion — an extraordinary jump driven by a handful of multibillion‑dollar AI infrastructure contracts and reserved capacity commitments. That number dwarfs Oracle’s prior backlog and grabbed investor headlines.
- Microsoft reported commercial RPO/backlog of roughly $392 billion (recent updates put that figure near $392B), a very large balance that management tied to Azure commitments and commercial software deals including OpenAI‑related commitments.
- Google Cloud’s backlog rose to about $155 billion, reflecting very strong sequential and year‑over‑year increases driven by enterprise AI deals and reserved infrastructure commitments.
- AWS’s backlog was reported in the neighborhood of $195–200 billion depending on the disclosure and press reconciliation — growth for AWS’s backlog is real but it is smaller both in percentage uplift and relative convertibility compared with Microsoft and Oracle.
Why backlog (RPO) trumps a single quarter of growth
Remaining performance obligations and backlog figures are not perfect apples‑to‑apples across companies: definitions vary, contract durations differ, and Microsoft’s RPO includes subscription and commercial software commitments that are not pure infrastructure. Still, there are three critical reasons RPO matters now more than ever:- Predictability: RPO represents contracted future revenue. When large, multi‑year RPO is concentrated in AI infrastructure (GPUs, reserved capacity) it signals near‑term capacity demand that will convert into recognized revenue and justify material capital expenditure.
- Capital allocation: large, AI‑driven backlog gives a vendor the business case to accelerate region builds, order accelerators, and price economies of scale. That in turn improves price‑performance for customers and can become a self‑reinforcing advantage.
- Market narrative & customer momentum: enterprise buyers prefer vendors who can demonstrate long‑dated commitments and the ability to deliver reserved capacity; sales teams use backlog to close enterprise deals. A big backlog is an argument for both sales leverage and investor confidence.
Product, go‑to‑market and engineering: why the gaps exist
AWS: breadth, modularity — and slower productization
AWS’s historical advantage is the deepest and broadest service catalog in the industry: IaaS, PaaS, databases, edge services, developer tooling, custom silicon (Graviton/Trainium/Inferentia) and a global datacenter footprint. Those advantages create lock‑in and make AWS the natural home for massive legacy and regulated workloads.But the AI buying behavior has favored vendors that productize models into managed, outcome‑oriented services — plug‑and‑play Copilots, embedded AI features in applications, and turnkey managed model hosting with predictable SLAs. AWS has been moving in that direction (Bedrock, SageMaker advances, managed third‑party models), yet the company’s long history as a provider of modular building blocks means customers still often need to assemble their own AI stacks. That longer time‑to‑value disadvantages AWS in sales cycles where quicker payback and packaged features win selection decisions.
Microsoft: integration + monetization
Microsoft’s advantage is enterprise distribution and monetization through software seats and workflows. Azure is paired with Microsoft 365, Dynamics, Power Platform and Copilot integrations — which means Microsoft can monetize AI both via infrastructure consumption (Azure) and by upselling AI features as seat‑based products inside widely used application suites. That multiplies revenue channels and makes it easier to translate AI interest into recurring, higher‑margin revenue. The RPO/backlog figures reflect substantial seat and cloud commitments, including OpenAI‑related deals.Google Cloud: product‑led AI momentum
Google’s strength lies in data and ML tooling — BigQuery, Vertex AI, Gemini models and custom accelerators (TPUs) — which resonate with data teams and model builders. Google’s recent results show that this product‑led approach is turning into large, reserved multi‑year deals and a fast‑growing backlog. For analytics‑ and model‑centric workloads, Google’s integrated data‑to‑model story offers a lower friction path to production than a parts‑based approach. That helps explain why Google is capturing a disproportionate share of the new AI wallet relative to its base.Oracle: database proximity and the AI infrastructure play
Oracle’s RPO surge to the mid‑hundreds of billions is a dramatic outlier and reflects large, concentrated contracts for AI infrastructure, including announced deals tied to major model providers. Oracle’s pitch is database and data‑plane proximity for performance‑sensitive AI, plus aggressive pricing to capture reserved infrastructure deals. While the size of Oracle’s RPO invites skepticism and verification (some contracts appear concentrated with a small number of big customers), the headline is important because it shows that even legacy enterprise vendors can move aggressively into the AI infrastructure market.Strengths AWS still owns — and why those matter
It’s important to be precise: AWS is not “out of the race.” The company retains structural advantages that are durable and meaningful for many enterprise workloads.- Absolute scale and global footprint: AWS’s global presence and mature operational practices are unmatched for latency, compliance and regulated workloads. That matters for large enterprises and governments.
- Breadth of services: The largest service catalog reduces lift for complex, heterogeneous applications that require many managed primitives.
- Custom silicon and vertical engineering: Graviton, Trainium and Inferentia (and continued investment in accelerator strategy) give AWS levers to optimize price‑performance for a wide range of workloads.
- Financial firepower: Amazon’s ability to fund capex at scale means AWS can still build the capacity needed for AI workloads — the backlog and reservation patterns will dictate where that capex flows.
Risks and structural challenges for AWS
- Narrative and speed of productization. AWS’s engineering culture historically prioritizes building primitives; the market now rewards packaged AI experiences. If competitors continue converting AI demand into feature‑rich, preintegrated products faster than AWS, that gap can translate into sustained share erosion for certain classes of AI workloads.
- Backlog and conversion rate. AWS’s backlog growth is real but smaller than Microsoft’s and Oracle’s headline numbers. Conversion timing (how quickly RPO turns into recognized revenue) depends on accelerator supply and data‑center build schedules — two constraints that can slow recognition even when contracts exist. That gives rivals time to lock in customers with managed solutions.
- Perception vs. reality. Perception matters in enterprise procurement cycles and investor sentiment. The optics of being the slowest major grower in percentage terms — particularly in an AI‑fuelled environment — can affect sales morale, partner narratives and enterprise RFP outcomes. The quarterly number matters less than the sustained narrative across several quarters.
- Margin and pricing pressure. To defend share, AWS may face periods of aggressive pricing or concessioning on reserved capacity — actions that compress margins and could reduce the financial headroom to invest as aggressively as rivals in productization or capex.
What to watch in the next two quarters (practical signals)
- Adoption metrics for managed AI services (e.g., Bedrock, SageMaker GA uptake, named enterprise Copilot‑style deployments). Look for customer case studies and percent of revenue attributable to managed AI features rather than raw IaaS consumption.
- RPO/backlog convertibility: what percentage of reported RPO is expected to convert in 12 months vs. beyond 24 months. Shorter weighted durations imply quicker revenue recognition and tighter correlation with quarter‑over‑quarter revenue growth.
- CapEx cadence and accelerator supply: how quickly vendors can bring GPUs/accelerators online to satisfy reserved contracts. Delays here can slow conversion despite large backlogs.
- Sales motion evidence: Are hyperscalers signing multi‑year, non‑cancelable reserved contracts with top model builders and enterprises? Or are recent large backlogs concentrated in a handful of customers whose execution remains uncertain? The breadth of customers matters.
- Pricing & margin signals: Will AWS defend revenue with discounts/credits? Will rivals sacrifice margin to capture share? Watch gross margin and operating income trends in cloud segments.
Pragmatic advice for WindowsForum readers and IT buyers
- Favor portability and modularity where lock‑in risk is material, but exploit vendor competition for reserved capacity pricing when hardware timelines are predictable. Use standardized model artifact formats (ONNX, TorchScript) and containerized runtimes to keep destinations interchangeable.
- Negotiate contract terms that tie vendor commitments to adoption or performance milestones rather than credits alone. This helps ensure reserved capacity is matched to usable outcomes.
- Pilot managed, productized AI features alongside raw IaaS experiments. Measure time‑to‑value and total cost per unit of business outcome, not just $/GPU‑hour. That will reveal whether a vendor’s packaged offering actually reduces engineering overhead.
- Build telemetry and observability into models and inference runtimes from day one. Chargeback and cost‑awareness are essential when costs move from marginal CPU cycles to GPU‑hour economics.
Critical assessment: strengths, overstated claims and unverifiable items
- Strengths acknowledged: AWS’s size, operational maturity and silicon investments are real advantages that protect many enterprise workloads from short‑term disruption. The company’s recent quarter is a legitimate sign of improvement from depressed growth rates earlier in the year.
- Credible competitor momentum: Microsoft’s and Google’s revenue and backlog figures are corroborated across multiple earnings transcripts and analyst writeups; their momentum in converting AI interest into contractual commitments is verifiable and material.
- Oracle’s RPO requires cautious interpretation. The headline $455 billion number is real as reported by Oracle, but the figure is concentrated and driven by a small set of very large contracts. That raises legitimate questions about diversification and convertibility; independent reporting flagged both the size of the commitments and the potential concentration around a handful of counterparties. Treat Oracle’s figure as a major signal, but also as an item requiring continued verification as the contracts convert into recognized revenue.
- AWS backlog rounding: public reporting and analyst reconstructions place AWS backlog in the ~$195–200 billion range; that’s materially large but noticeably smaller than Microsoft’s and Oracle’s reported RPO balances. The exact comparability across companies is limited because of definitional differences (what each company includes in “RPO” vs. “backlog”). Flag this as an apples‑to‑apples caveat — the direction of the signal stands even if absolute comparability is imperfect.
Final thought
A single quarter of 20% growth and $33 billion in revenue is an achievement any other company would celebrate. But AWS is not “any other company.” It built the modern cloud and set baseline expectations for what leadership looks like. The AI era has changed the battleground: it is not just about raw infrastructure anymore but about productized intelligence, embedded application revenue, and large multi‑year reserved commitments that skew the market’s future revenue profile.The current earnings cycle shows a new reality: momentum matters as much as base. Microsoft and Google — and in headline cases Oracle — are sprinting into AI‑first deals and reserved capacity that promise to convert into revenue and strategic customer lock‑in. AWS’s recent acceleration is necessary and important, but it is not yet sufficient to shift the narrative back in its favor. The decisive chapters of this Cloud War will be written in backlogs converting to revenue, capex and accelerator fulfillment, and which vendor can turn AI interest into durable, monetizable enterprise outcomes fastest.
Source: Cloud Wars AWS, Despite Q3 Surge, Still Plodding Far Behind Microsoft, Google, Oracle
