UBS Field Checks at Oracle AI World: Stable Cloud Spending Fuels Hyperscalers

UBS’s field checks at Oracle’s AI World suggest a broadly positive but measured story: “stable, healthy” cloud spending is underpinning demand, and the largest hyperscalers — Amazon (AWS), Google Cloud, and Microsoft (Azure) — are positioned to capture the bulk of that growth, albeit with different upside and risk profiles.

Background​

Cloud spending has moved from optional IT modernization to a near‑universal requirement for enterprises building AI, analytics, and modern SaaS. Over the past year the narrative has shifted from “will customers spend?” to “how and where will they spend?” UBS’s outreach to customers and partners at Oracle’s AI World — cited in recent market coverage — captured that transition: enterprise budgets are more committed, purchases are increasingly tied to AI and data modernization projects, and deals are being written with longer horizons and more reserved capacity than before.
At the same time, hyperscaler earnings through mid‑2025 show the same pattern UBS describes: Google Cloud accelerating, Microsoft leveraging integrated software+cloud monetization, and AWS remaining the scale leader but facing more modest growth comparatives. Those public results confirm the qualitative color UBS gathered in conversations with customers and partners.

---

Market snapshot: where the numbers stand today​

Short, verifiable picture of hyperscaler performance in mid‑2025:

• Google Cloud: reported ~32% year‑over‑year revenue growth to roughly $13.6 billion in Q2 2025, with a marked improvement in operating margins as the business scales. Management has openly increased capex guidance to support rapid AI demand.
• Microsoft (Intelligent Cloud/Azure): Azure and cloud‑adjacent services have been growing at high‑teens to low‑30s percentages depending on metric; Microsoft’s Intelligent Cloud segment reported revenues near the $29–30 billion quarterly mark, with Azure calling out materially faster AI workload growth inside the segment. Microsoft’s product integration (M365/Copilot/Teams/Dynamics) gives it a distinctive pathway to monetize AI at the app and seat level.
• Amazon Web Services (AWS): AWS remains the largest single cloud revenue generator, with reported quarterly sales roughly in the low‑$30 billion range (about 17–18% year‑over‑year growth in Q2 2025), though its growth has trailed Google Cloud and Azure in recent quarters. AWS is the foundational profit engine for Amazon and continues to invest heavily in data center capacity and custom silicon.

Independent market trackers place global cloud infrastructure market shares at roughly AWS ~30% / Microsoft ~20% / Google Cloud ~13% (Q2 2025 snapshot), underscoring the concentrated nature of the market but also the rising share gains of Microsoft and Google.

---

What UBS heard at Oracle AI World — field intelligence that matters​

UBS’s conversations with customers and partners at Oracle’s AI World (reports indicate about 11 direct interviews in that event context) produced several repeatable themes that matter to investors and enterprise buyers:

• Budgets are committed, not speculative. Many organizations are moving from pilot projects to contracted capacity for AI training and inference, which favors cloud vendors who can deliver committed GPU/accelerator supply and contractual SLAs.
• Longer contracts and reservations are becoming the norm. Customers prefer predictability for model training schedules and are willing to sign multi‑year capacity commitments — which in turn creates backlog and RPO (remaining performance obligations) dynamics different from short‑term consumption.
• Technical fit and integration often trump pure price. Buyers repeatedly prioritized end‑to‑end AI tooling, data services, and integration with enterprise applications (security, identity, compliance) when selecting cloud partners, which helps explain Microsoft’s advantage when AI features are bundled with Office/M365 and Google’s traction with data/ML engineering teams.

UBS’s conclusion — that Amazon, Google, and Microsoft are poised to benefit from “stable, healthy” cloud spending — is therefore grounded in direct customer signals rather than just top‑down surveys. That gives the thesis practical weight for near‑term revenue visibility, but it also implies divergent outcomes depending on which vendor can convert bookings into recognized revenue efficiently.

---

Why the big three are likely beneficiaries — strengths and differentiators​

1) Scale and capacity (AWS)​

• Why it matters: Large AI training and inference projects need racks, power, networking and a global footprint. AWS’s absolute scale and ecosystem give it an enduring advantage for customers prioritizing global reach and elasticity.
• Concrete strengths: the largest service catalogue, custom silicon (Trainium/Inferentia/Graviton), and a broad partner ecosystem that can assemble specialized stacks. Those assets help AWS compete on TCO and verticalized solutions.

2) Integrated monetization (Microsoft)​

• Why it matters: Microsoft can capture higher‑value monetization by embedding AI into widely deployed productivity suites (Copilot in Microsoft 365), enterprise applications (Dynamics, Power Platform), and developer tools (GitHub).
• Concrete strengths: distribution into seats and enterprise processes means AI features translate quickly into recurring, measurable revenue. Microsoft’s scale of commercial bookings and its ability to upsell Copilot/AI features gives Microsoft unique monetization leverage.

3) AI tooling and model engineering (Google Cloud)​

• Why it matters: Customers building large ML pipelines and custom models often prioritize tooling, TPUs/GPUs, and deep data analytics. Google’s Vertex AI, BigQuery, Gemini lineage, and TPU-based infrastructure are purpose-built for those workloads.
• Concrete strengths: Google has demonstrated both growth and improving profitability in cloud, and its data/ML stack is frequently the choice for large training jobs and analytics‑heavy workloads. The company’s mid‑2025 results show Google Cloud’s revenue and margins improving markedly.

---

The AI tailwind — why cloud spending is different this cycle​

AI workloads change cloud economics in three structural ways:

• Scale of compute demand: Training LLMs and large models consumes orders of magnitude more accelerators (GPUs/TPUs) than typical enterprise workloads, increasing both capex and contracted demand.
• Contracting behavior: Organizations and AI labs prefer reserved, long‑dated capacity to avoid spot shortages; that creates large backlogs and RPOs for vendors but delays recognition and requires careful conversion execution.
• New monetization vectors: Managed model hosting, inference credits, domain‑specific AI services, and data‑to‑model integrations carry higher gross margins than raw compute and can lift cloud profitability as they scale.

These dynamics explain why UBS’s field checks find “stable, healthy” spending: budgets are allocated to AI projects where cloud is the simplest scaling vehicle. For hyperscalers, that creates both revenue opportunity and execution risk tied to capacity and capital intensity.

---

Capex, supply constraints, and margin dynamics — the tradeoffs​

Hyperscalers are expanding capacity aggressively to meet AI demand, but building data center capacity at hyperscale is capital‑intensive and subject to supply‑chain and energy constraints.

• Capex increases: Alphabet raised 2025 capex guidance materially as it doubled down on AI infrastructure; Microsoft and Amazon have similarly disclosed multi‑year, large capital spends to secure capacity. These moves signal confidence in demand but push near‑term depreciation and gross‑margin pressure.
• GPU/accelerator supply: Dependence on a narrow set of accelerator providers (notably NVIDIA) and geopolitical export controls can create chokepoints. That’s a practical constraint on how fast clouds can convert backlog into revenue.
• Margin tension: Higher‑margin AI services can offset infrastructure depreciation, but only after scale. Early phases of capacity buildouts typically compress gross margins until utilization and monetization catch up. UBS and other analysts highlight this two‑stage dynamic consistently.

---

Risks and red flags investors and CIOs should watch​

• RPO vs. revenue recognition: Large backlogs and multi‑year reservations are encouraging, but they are not immediate revenue — conversion timelines and recognition schedules matter. Treat huge RPO numbers cautiously until conversion cadence is visible.
• Execution on capacity buildouts: Power, real estate, environmental approvals, and local supply chains are real constraints. Missed delivery timelines weaken backlog credibility.
• GPU/chip supply and pricing: Supply shocks, pricing volatility, or export restrictions can materially slow deployments and raise costs, impacting margins.
• Concentration risk in mega‑deals: Some vendors report very large reported commitments (or media‑circulated figures about single customers). Until counterparties or contract details are disclosed, those headlines require discipline and skepticism. UBS flagged examples where press narratives outpaced verifiable contract details.
• Regulatory and geopolitical exposures: Data sovereignty rules, procurement restrictions, antitrust scrutiny, and national security reviews can complicate large cloud contracts, especially for public sector or cross‑border AI deployments.

---

What this means for each company — a concise, comparative assessment​

Amazon (AWS)​

• Strengths: unmatched scale, broad service catalogue, custom silicon and deep partner ecosystem.
• Near‑term view: benefits from stable cloud spending but faces tougher growth comps and margin pressure from higher infrastructure depreciation. Execution of cost efficiencies and TCO improvement will be key.

Microsoft (Azure)​

• Strengths: integrated distribution through Microsoft 365 and enterprise apps, clear path to monetize AI at the seat level; large commercial bookings and diversified cloud revenues.
• Near‑term view: Microsoft is well placed to convert AI adoption into recurring revenue; capex timing and partner governance (notably OpenAI relationships) remain watchpoints.

Google Cloud​

• Strengths: fastest‑growing major cloud business in recent quarters with strong margin improvement; deep ML tooling and TPU/GPU investments.
• Near‑term view: Google Cloud can surprise on the upside if its backlog and enterprise sales motion continue to improve; large capex increases are required to sustain the AI runway.

---

Actionable investor signals — what to watch over the next 12 months​

• Quarterly cloud revenue and operating‑income trends from the hyperscalers (are AI services lifting margins?).
• Capex cadence and depreciation schedules (do companies show utilization gains that justify spending?).
• RPO/backlog conversion rates and any named contract confirmations that substantiate headline backlog figures.
• Supply‑chain disclosures and supplier purchase commitments for accelerators (NVIDIA, AMD, custom silicon).
• Price and contract term changes in the market (e.g., new reserved pricing, burst‑capacity credits, or managed model hosting tiers).

---

Practical guidance for enterprise buyers and CIOs​

• Map workloads to cloud‑grade tiers: reserve capacity for model training, use managed inference for latency‑sensitive production, and keep less critical workloads on flexible consumption plans.
• Negotiate exit and portability terms: containerization, model formats, and clear data‑egress rules reduce vendor lock‑in risk.
• Ask for capacity roadmaps and SLA commitments tied to accelerator availability. These operational details determine whether a vendor can meet production timetables for model training and inference.
• Consider multicloud strategies for risk diversification, but negotiate discounts and technical support for cross‑vendor deployments. UBS’s field checks show many enterprises opting for multi‑vendor approaches to ensure capacity and regulatory compliance.

---

Critical analysis — strengths of the UBS read and where it may understate risks​

UBS’s field work is valuable because it captures customer intent rather than relying solely on booking numbers or press reports. That makes their conclusion — hyperscalers stand to benefit from stable, healthy cloud spending — credible.
Strengths of the UBS read:

• Grounded in direct customer conversations at a focused event (Oracle AI World).
• Reinforces verifiable earnings trends showing the major clouds winning AI‑related commitments.

Where UBS’s read could understate risk:

• Backlog and RPO headlines can create optimism that’s premature if conversion is delayed by power, chip or construction bottlenecks. UBS’s conclusions rely on intent; execution remains the gating factor.
• Competitive dynamics are fluid: smaller, specialized vendors and regional providers can win vertical or regulated workloads that hyperscalers can’t economically serve, which fragments some of the expected upside.

Those caveats matter because the path from signed commitments to recognized revenue is operational and often protracted in this cycle.

---

Bottom line — sensible positioning for investors and CIOs​

• For investors: exposure to the hyperscalers remains a pragmatic way to play the secular AI and cloud transition, but portfolio sizing should reflect differing risk/reward profiles: Microsoft offers diversified monetization and cash‑flow resilience; Google offers upside via fastest growth and margin recovery; AWS offers scale and profit durability but faces tougher growth comps. UBS’s field checks add confidence to that allocation framework while reminding investors to watch conversion and capex metrics closely.
• For CIOs and procurement leaders: treat vendor commitments as operational projects — secure contractual SLAs on accelerator availability, negotiate multi‑year reserved capacity with exit/egress protections, and design multicloud portability into model pipelines. UBS’s customer interviews echo this pragmatic buyer posture.

---

The UBS message is neither bullish in the sense of a frothy surge nor bearish: it is measured. Cloud spending is stabilizing and turning healthy precisely because AI projects have moved from curiosity to committed infrastructure programs. That structural shift favors the hyperscalers — Amazon, Google, and Microsoft — but the path from backlog to sustainable profits is conditional on execution: capacity delivery, accelerator supply, disciplined capex, and the ability to monetize AI services at scale. Investors and enterprise buyers who anchor decisions on those convertibility signals — not just headline bookings — will be best positioned for the next phase of the cloud‑AI transition.

---

Source: Seeking Alpha Amazon, Google, Microsoft likely to benefit from 'stable, healthy' cloud spending: UBS (AMZN:NASDAQ)