Data Management News for the Week of October 23; Updates from Denodo, Couchbase, Teradata & More

A BARC study commissioned by Actian found that 76% of organizations say deploying AI outside the cloud has become more important over the past year, even as 77% use cloud infrastructure for some AI workloads. Based on 329 respondents surveyed in June and July, the research identifies data sensitivity, mission-critical processes, cybersecurity, and provider dependency as key drivers. The findings point to deliberate workload placement across cloud, on-premises, and edge environments—not wholesale cloud retreat—with Actian emphasizing enforceable policies and auditable controls wherever AI operates.

Read on for more→ New BARC Research Commissioned by Actian Finds 76% Say Deploying AI Outside the Cloud Is Now More Important

Boomi Acquires Elementary To Add Data Observability And Reliability For Enterprise AI

Boomi has acquired Elementary Data in a deal valued by CTech at approximately $40–45 million in cash and stock, adding data observability to its integration and governance platform. Elementary combines quality monitoring, column-level lineage, shared metadata, and an MCP server that exposes data context and health to AI tools. All 17 employees will join Boomi, which plans to continue supporting both Elementary Cloud and its open-source project without immediate changes to the repository or release process.

Read on for more→ Boomi Acquires Israeli Startup Elementary for $40–45 Million

Couchbase Adds Managed Agent Memory, Semantic Caching, And Bedrock Integration

Couchbase has expanded its AI Data Plane with managed Agent Memory on AWS, a managed MCP server, semantic caching, and native Amazon Bedrock integration in its Model Service. The capabilities preserve context between sessions and reuse responses to similar requests, reducing repeated model processing. Couchbase reports 40–60% fewer model calls in semantic-caching tests and 33–45% fewer memory tokens than leading alternatives on the LoCoMo benchmark; these measure different workloads, not a combined savings rate.

Read on for more→ Couchbase Adds Managed Agent Memory and Semantic Caching to AI Data Plane

Denodo Study Models 290% Return From A Governed AI Data Layer

A Veqtor8 study published by Denodo projects a 290% return on investment over three years from using a governed AI data layer to support enterprise agents. Based on practitioner and partner interviews, the composite model scales from 100 to 900 agents across 100 distributed sources and reports threefold faster deployment and 68% lower AI operating costs. The modeled savings come from reusable data products, centralized policies, and smaller context payloads rather than custom integrations for every agent; the figures represent a composite scenario, not guaranteed customer results.

Read on for more→ Denodo Study Finds 290% AI Data Layer Return on Investment

Grafana Speeds Log Search And Expands Automated Telemetry Cost Controls

Grafana Labs has made faster full-text search generally available in Grafana Cloud Logs, using technology from its Logline acquisition to locate rare values such as request and customer IDs more quickly. Adaptive Logs Archive, now in public preview, routes selected logs to customer-owned object storage in an open format for lower-cost retention. Other generally available updates automatically apply metric-cardinality recommendations and let developers retrieve traces dropped by adaptive sampling for 24 hours, balancing telemetry savings with investigation needs as AI workloads generate more data.

Read on for more→ Grafana Labs Brings Faster Full-Text Search and Smarter Data Optimization to Grafana Cloud at ObservabilityCON 2026

Neo4j Introduces Context Engine To Connect Agent Reasoning With Business Meaning And Policies

Neo4j has introduced Context Engine in private preview, combining business definitions, processes, enterprise data, access rules, and operational memory in a connected graph. A six-layer ontology tells agents what business concepts mean, where relevant data resides, how to access it, and what previous decisions can teach them. Context Studio provides a shared management interface, while governed GraphRAG retrieval and connected memory aim to make answers and actions explainable without requiring customers to replace their existing LLMs or agent frameworks.

Read on for more→ Neo4j Context Engine: Rapidly Build Dependable Enterprise Agents

Proofpoint Agrees To Acquire Varonis In $6.3 Billion Data And AI Security Deal

Proofpoint has entered a definitive agreement to acquire Varonis in an all-cash transaction valued at approximately $6.3 billion, with shareholders receiving $49.75 per share. The companies plan to connect communications, identities, permissions, sensitive data, and AI activity in a unified security platform that helps organizations control access and prevent data loss. Following closing, Proofpoint intends to integrate capabilities and align go-to-market efforts, extending its human and agent cybersecurity business with Varonis’ data visibility, exposure reduction, and AI security technology.

Read on for more→ Proofpoint to Acquire Varonis in $6.3B Transaction

Qlik Brings Governed Business Context And Analytics Workflows Into Claude

Qlik’s MCP server is now available through Anthropic’s Claude Directory, giving customers a simpler way to connect Claude with governed data, business metrics, lineage, and existing analytical logic in Qlik Cloud. Users can search apps and datasets, ask natural-language questions, inspect expressions and scripts, and create dashboards and visualizations without rebuilding business definitions in prompts.

Read on for more→ Qlik Brings Governed Business Context and Analytics Workflows to Claude

Starburst Introduces GPU-Native SQL Engine Without Requiring Query Or Application Changes

Starburst has announced limited general availability of Torin, a Trino-based engine for Starburst Enterprise that executes SQL natively on NVIDIA GPUs using cuDF. Torin complements the existing CPU engine without requiring changes to Trino SQL, applications, agents, or governance controls. Starburst reports up to sevenfold faster performance in its TPC-H testing and 3.4 times the ClickBench throughput of a CPU baseline at similar hourly hardware cost; the vendor-run, single-node development benchmarks exclude software licensing and do not guarantee production results.

Read on for more→ Starburst Introduces Torin, a GPU-Native Analytics Engine Built on NVIDIA cuDF

Teradata Combines Identity Reduction, Data-Layer Security, And Immutable AI Recovery

Teradata has announced privacy, security, and resilience capabilities that protect sensitive information while preserving its usefulness for enterprise AI. Anonomatic-powered Poly-Anonymization separates identity from analytical records, while native masking, classification, tagging, and column-level encryption enforce protection closer to the data. Additional capabilities include post-quantum protection for data in transit, Immuta-based attribute-driven authorization, real-time replication, and immutable backups.

Read on for more→ Teradata Introduces New Privacy, Security, and Resiliency Capabilities for Enterprise AI Data Protection