- The Cloud Cover
- Posts
- Oracle’s $455B AI Shockwave Hits the Cloud
Oracle’s $455B AI Shockwave Hits the Cloud
This Week in Cloud — September 10, 2025
Welcome back to The Cloud Cover, your weekly guide to the business and technology shifts shaping cloud computing. This week, Oracle’s $455B AI contracts set a new benchmark for the scale of cloud commitments, Microsoft flexes pricing power on enterprise renewals, and AWS strengthens resilience for global inference. The common thread? The cloud is entering a new era defined by multi-year commitment and a sharpened focus on trust and scale. Let’s unpack it.
💸 Oracle’s $455 Billion Bombshell and the True Price of AI
This week, the most shocking number in cloud didn't come from a performance benchmark, but from an accounting line item. Oracle reported its Q1 FY2026 earnings, and while current revenue missed estimates, it was completely overshadowed by a single metric: Total Remaining Performance Obligations (RPO) exploded by 359% year-over-year to reach $455 billion. For context, that’s the total value of contracted future business that has not yet been recognized as revenue.
According to CEO Safra Catz, this unprecedented growth was driven primarily by the signing of just four multi-billion-dollar contracts in the quarter, all for training large language models. The company projects this backlog will exceed half a trillion dollars soon. This disconnect between today’s revenue and tomorrow’s contracted tsunami of cash is the clearest signal yet of OCI’s success in the hyper-competitive market for large-scale AI training infrastructure.
I know that we have been highlighting Oracle quite a lot these last few months, perhaps more than you would expect compared to larger established providers like AWS. However, the past two quarters of financial results and major partnerships have been indicators to me that Oracle is a genuine player in the space, one that is gaining ground and likely here to stay. Hopefully, we will hear some big announcements and launches from other providers through the fall, but it has been Oracle that stole the show this summer.
Lastly, this particular news is a reminder (as though we even needed one) of the sheer scale of capital being committed to AI. These aren't pay-as-you-go workloads; they are massive, multi-year contracts designed to secure vast amounts of GPU capacity. The AI land grab is happening now, and it's being fought with long-term, multi-billion-dollar lock-ins.
🔍 The Rundown
Global Bedrock Inference: The new cross-region inference for Claude Sonnet 4 improves resilience for global AI applications. This feature mitigates risks tied to single-region capacity constraints or service disruptions, a critical need for production-grade generative AI.
Neptune Public Endpoints: Developers can now connect to the Neptune graph database from outside a VPC without complex networking configurations. While this simplifies workflows, it creates a new attack surface that security teams must now explicitly govern and monitor to prevent unauthorized access.
ECS Console Shell: A new ECS Exec feature was launched in the AWS Console. It allows developers to securely shell into running containers directly from CloudShell, removing the need to configure inbound ports or manage SSH keys.
AMD Turin VMs:New general purpose (Dasv7), memory-optimized (Easv7), and compute-optimized (Fasv7) virtual machines are now in public preview. Based on AMD's 5th Gen 'Turin' processors, they promise significant performance boosts, including up to a 130% gain for web server workloads over the prior generation.
M365 Volume Discount Changes: In a significant licensing update, Microsoft will eliminate tiered volume discounts for online services like Microsoft 365 within its Enterprise Agreements, effective November 1, 2025. This policy shift will result in effective price increases ranging from 6% to 12% for many of Microsoft's largest customers when they renew their agreements.
AI Agent Orchestration: Microsoft is pushing a strategic vision to make Azure Logic Apps a central hub for orchestrating sophisticated AI agents. The company declared it is "Ushering in the Era of Multi-Agentic Business Process Automation," signaling a move beyond simple integration to automated workflows driven by collaborating AIs.
Gemini IDE Integration:Gemini Code Assist received several iterative updates, including a /deploy command for Cloud Run and the ability to prioritize specific repositories for more relevant context. Separately, a new IAM Role Picker in preview embeds Gemini in the console, allowing admins to use natural language to find the most appropriate, least-permissive IAM role for a task.
EU Egress Fee Elimination: Ahead of the EU Data Act's enforcement on September 12, Google has eliminated cloud-to-cloud data transfer fees in the EU and UK. This move goes beyond the act's requirement to charge "at cost," a more aggressive step than competitors who have stated they will only charge at cost.
Gartner Magic Quadrant Leader: For the third consecutive year, Oracle was named a Leader in the Gartner® Magic Quadrant™ for Distributed Hybrid Infrastructure. The report positioned Oracle furthest for "completeness of vision," validating its differentiated strategy of delivering its full portfolio of cloud services across the public cloud, dedicated customer regions, and multicloud environments.
Open Source AI Models: The OCI Generative AI service now provides beta access to OpenAI's pretrained open-source gpt-oss models. The move expands model choice on the platform, catering to enterprise demand for powerful, open-weight models suitable for complex reasoning tasks.
📈 Trending Now: To Lock-In or Not to Lock-In
The big story this week isn’t about a single feature; it's about the money. The financial maneuvers from Oracle and Microsoft signal a fundamental shift in the economics of cloud consumption, driven entirely by AI. The era of flexible, consumption-based pricing for cutting-edge services is being replaced by a calculated strategy to secure long-term, high-value commitments.
Oracle's $455 billion in RPO, fueled by a handful of massive AI training deals, is Exhibit A. These aren't customers swiping a credit card; they are enterprises signing multi-year, multi-billion-dollar contracts that effectively lock them into OCI for the foreseeable future of these workloads. Concurrently, Microsoft's decision to eliminate tiered volume discounts for its largest Enterprise Agreement customers is a classic exercise of pricing power. As Copilot and other AI services become indispensable, Microsoft is confident that its platform is too embedded in customer operations to lose, even with an effective price hike of 6-12% for its biggest clients.
For cloud consumers, the message is clear: the market is moving toward less flexible and more expensive long-term contracts. This new reality demands more sophisticated upfront negotiation and meticulous long-term capacity planning than ever before.
📅 Event Radar
15-18
Few tickets remaining!
17
Registration still open
8-10
Speakers list now available
28-29
Register today!
👋 Until Next Week
The industry-wide focus has continues to shift from benchmarking foundational models to building the operational frameworks to let them act autonomously. Google is betting that security will be the primary enabler for agentic AI, while Microsoft is building the orchestration layer, and AWS is hardening the underlying inference engine for global scale. The next battle won't just be about who has the smartest model, but who provides the safest and most efficient way to let it run the business.
Stay tuned.
Do you enjoy these emails? Your friends and colleagues might, too! Help us grow the cloud community by sharing the newsletter with others.