This Week in Cloud — May 28, 2026
Welcome back to The Cloud Cover, your essential guide to navigating the dynamic world of cloud for Solutions Architects, engineers, and IT leaders. This week, Snowflake makes a $6B bet on AWS, Google turns AI prototypes into live cloud apps, and the hyperscalers keep tightening the path from governed data to production AI. Let’s dive in.
⚡ Snowflake's $6B Signal
For the last year, cloud AI stories have mostly sounded like a contest over models, chips, and how many agents a vendor can cram into a keynote. This week we saw a slightly cleaner signal. Snowflake signed a five-year, $6 billion strategic collaboration agreement with AWS, aimed squarely at enterprise agentic AI workloads.
Snowflake is committing real spend to AWS Graviton compute and AI infrastructure at the same time its own business is accelerating. The company reported $1.33 billion in Q1 2027 product revenue, up 34% year over year, and raised its full-year outlook. In other words, the data platforms selling AI to enterprises are turning around and buying cloud infrastructure at a massive scale to make those promises real.
In some sense, this shouldn’t be surprising. Enterprise AI does not live in a vacuum. It lives next to the data, identity systems, procurement rules, and cost controls that exist inside a company. Snowflake and AWS are trying to make the data layer and compute layer feel like one system, reducing the need to move sensitive data around just to run models or agents against it.
For cloud buyers, this is the AI flywheel made visible. Data gravity pulls workloads toward the platform where the data already sits. Then agentic AI raises compute consumption around that data. Then cloud marketplaces and committed-spend agreements make the whole thing easier to buy, but harder to casually unwind. The winning move may be owning the shortest path between enterprise data and AI execution.
🔍 The Rundown
Snowflake's $6B AWS Bet: Snowflake expanded its AWS collaboration with a $6 billion, five-year commitment tied to Graviton compute and AI infrastructure. This is one of the clearest signs yet that enterprise agentic AI demand is converting into hard cloud capacity commitments, not just experimentation budgets.
OpenAI-Compatible SageMaker: Amazon SageMaker AI now supports OpenAI-compatible APIs for inference endpoints, letting teams use the OpenAI SDK, LangChain, and Strands Agents by changing an endpoint URL. It is a pragmatic portability move for customers that want OpenAI-style tooling without giving up VPC control, instance choice, or SageMaker operations.
Database Reliability Push: Amazon RDS added ENA Express for Multi-AZ replication, improving throughput and reducing latency variability for write-heavy HA databases. Aurora MySQL also moved to the MySQL 8.4 LTS line, giving database teams a more current long-term-support target.
Cloud-Native File Identity: Azure Files now has GA support for Entra-only identities for SMB access, removing the need for Active Directory, hybrid sync, or managed domain controllers in supported scenarios. For AVD, FSLogix, and cloud-native Windows estates, that is a meaningful simplification of the old identity tax around file shares.
Fleet Networking Preview: Azure Kubernetes Fleet Manager introduced Cilium-based cross-cluster networking in public preview. The goal is direct service-to-service communication and policy enforcement across AKS clusters, which matters as platform teams split production workloads across regions, environments, and business units.
Signed Malware Takedown: Microsoft disrupted Fox Tempest's malware-signing-as-a-service operation, which abused Microsoft Artifact Signing and short-lived certificates to make malicious binaries look legitimate. It is a reminder that trust services become attacker targets the moment they become deployment shortcuts.
AI Studio to Production: Google AI Studio now lets new users deploy up to two full-stack applications to the Google Cloud Starter Tier, with Cloud Run, Firebase Auth, Firestore, and Cloud SQL support. It is a clever on-ramp, but also a classic cloud funnel. Prototypes become real infrastructure before teams have fully thought through ownership, security, and cost.
Feature Flags for AI Apps: AppLifecycle Manager added public preview feature flags built on OpenFeature and flagd. The AI-era angle is useful: teams can change prompts, rollout percentages, and risky behavior through configuration rather than redeploying code every time an agent needs a guardrail adjustment.
Account Suspension Warning: Railway published an incident report after an erroneous Google Cloud account suspension disrupted its production infrastructure. This looks customer-specific rather than a broad GCP outage, but it is still an uncomfortable example of automated trust and abuse systems becoming part of the availability model.
Project Jupiter Ground Game: Oracle launched a public campaign around its Project Jupiter data center campus in Dona Ana County, New Mexico, highlighting local jobs, water-system investments, fuel-cell power, and community funding. It is not a product launch, but it is a useful look at the new reality of AI infrastructure. Securing land, power, water, and local trust is now part of cloud execution.
📈 Trending Now: Control Plane Portability
While there weren’t any “game changing” announcements this week, there was one interesting theme underneath several of the smaller releases. Portability is becoming a product feature vendors use to get workloads in the door.
AWS is making SageMaker speak the OpenAI API dialect. Google is turning AI Studio prompts into Cloud Run, Firebase, Firestore, and Cloud SQL deployments with almost no procurement friction. Azure is removing old Active Directory dependencies from file shares while extending Kubernetes networking across fleets. Everyone is working to reduce friction at the boundary between "this is an app idea" and "this is now running on our platform."
That is useful, but it also changes the job for architects. In addition to all of the other platform concerns, there is a new one around “accidental production.” When a developer can go from prompt to deployed app, or from OpenAI-style SDK call to a private SageMaker endpoint, governance needs to arrive earlier in the lifecycle. Identity, cost attribution, kill switches, service ownership, and exit paths all have to be designed early. The winners will be the platforms that make the easy path also the governable one.
📅 Event Radar
28
Even more AI sessions coming to a city near you.
31
Join for the latest AWS news and announcements.
2-3
Join for Microsoft's main dev oriented conference.
4
Latest Snowflake updates you should know.
👋 Until Next Week
Despite a slightly quieter week, we did see some interesting announcements. Fewer abstract AI promises, more spend commitments, identity plumbing, database updates, networking controls, and production guardrails. That is probably healthy. The next phase of enterprise AI will be decided more by who can make AI workloads boring enough to run every day.
Do you enjoy these emails? Your friends and colleagues might, too! Help us grow the cloud community by sharing the newsletter with others.