Google Cloud Adds Sandboxed Serverless Execution as Infrastructure Vendors Push Control

Google Cloud Targets LLM-Generated Code Risk

Google Cloud introduced Cloud Run Sandboxes in public preview, a new environment for safely executing untrusted, LLM-generated code. The feature addresses the emerging risk of production systems running AI-generated workloads without adequate isolation. Google also upgraded Looker instances from FIPS 140-2 to FIPS 140-3 Level 1 compliance automatically, tightening the security posture for regulated customers. Bigtable now supports direct connectivity that bypasses Google's front end under specific criteria, reducing latency for high-throughput data workloads.

The competitive pressure is on AWS Lambda, App Runner, Azure Container Apps, and traditional database stacks. Google is differentiating on execution safety and data-path control rather than raw capacity. For buyers evaluating platform modernization, Cloud Run now handles security-sensitive workloads that previously required custom sandboxing or off-platform execution. The FIPS upgrade reduces audit friction for government and financial services buyers. Bigtable's direct connectivity can shift disaster recovery and hybrid architecture decisions for teams running latency-sensitive transaction processing.

Looker 26.12 rolls out from July 12 to July 26 over a two-week window. Google also previewed disaster recovery for VMware Engine using Jetstream and added Apache Solr integration to AlloyDB for external search workloads. The migration path from legacy App Engine Images to Cloud Run reduces the technical debt burden for teams stuck on deprecated infrastructure.

IBM Expands z17 and LinuxONE 5 Packaging to Rack Mounts

IBM added rack-mount configurations across the full z17 and LinuxONE 5 portfolio, the first time IBM has offered rack-mount packaging alongside single-frame systems for its mainframe and enterprise Linux platforms. The packaging change preserves flagship performance and security specifications while reducing physical deployment constraints.

This matters for banking, insurance, government, and large-scale transaction buyers because rack-mount packaging simplifies data-center fit and procurement planning. IBM is competing against HPE, Dell, Oracle, and public-cloud mainframe-replacement strategies by lowering the friction of adopting or refreshing Z and LinuxONE environments. Buyers no longer face a forced facility redesign to deploy or expand IBM infrastructure.

The shift is toward hybrid deployment flexibility for transaction-heavy and highly regulated enterprises that still require mainframe-class uptime and security. For buyers running core banking, claims processing, or government workloads, the rack-mount option can accelerate refresh cycles and reduce capital planning complexity.

DXC Launches Private Cloud+ with Public-Cloud Pricing Model

DXC Technology released DXC Private Cloud+, positioned as delivering public-cloud-like flexibility and pricing while maintaining full control over sensitive data and workloads. The announcement centers on a consumption-based pricing structure rather than disclosed unit pricing. DXC is competing with VMware Cloud Foundation, HPE GreenLake, Dell APEX, Azure Stack, AWS Outposts, and other managed private cloud providers.

The value proposition targets workloads too sensitive or too expensive to keep in hyperscale public cloud. For buyers, this can shift budget conversations by preserving governance while negotiating for elastic, subscription-style pricing. The immediate buying impact is for workloads subject to data sovereignty requirements, high egress costs, or compliance mandates that prohibit public cloud deployment.

What This Means for Cloud Infrastructure Buyers

The pattern across these announcements is vendors selling control, compliance, and workload-specific infrastructure rather than generic hyperscale expansion. Google is addressing LLM execution risk and regulated workload requirements. IBM is reducing the deployment barrier for mainframe-class infrastructure. DXC is offering private cloud with public-cloud pricing.

IDC forecasts worldwide public cloud spending will reach $1.6 trillion by 2028, a 99% increase over 2024. Flexera's 2026 cloud report confirms that hybrid cloud remains the dominant architecture while generative AI accelerates infrastructure spend. Buyers should expect continued budget pressure from AI infrastructure, cloud networking, and governance tooling.

Cribl acquired CardinalOps to add agentic detection engineering to its telemetry platform, pressuring Splunk, Microsoft Sentinel, and Elastic Security by moving value upstream into open telemetry architectures. CIQ upgraded Ascender Pro to automate remediation across enterprise Linux fleets, competing with Red Hat Ansible Automation Platform and SUSE Manager. Both moves reduce operational headcount pressure and lower mean time to repair.

What to Watch

Watch whether Google's Cloud Run Sandboxes gain adoption among teams deploying AI agents in production. The security model will determine whether enterprises trust LLM-generated code in customer-facing workloads. Track IBM's rack-mount packaging uptake among regulated buyers facing mainframe refresh cycles. Monitor DXC Private Cloud+ pricing transparency and whether it forces HPE GreenLake and Dell APEX to adjust consumption models. The broader question is whether the shift toward workload-specific infrastructure and control continues to outpace the hyperscale public-cloud expansion narrative.