# Materialize Product Update | September '26 Canonical: https://brew.new/browse/templates/email/pt1_k97gzg0wtph68gxmdp5kaz1px98fqem2 Brand: materialize.com Category: welcome ![Preview of Materialize Product Update | September '26](https://cdn.brew.new/email-preview-b6b4eeb53341f936-tracking_r57n14j3knggdjz0vtdnd58q4s8fnks0-1791159678660.png) ## Email content View in browser Materialize As Materialize deployments get much bigger, both in terms of size and number of users, we're hearing that our customers want more predictable performance when serving Materialize directly, more support for managing large numbers of users, and better connectivity for harmonizing live data with batch analytics pipelines. We've delivered those and more in September. Internally, we've been loving our live context graph. Anyone in the company and their agents can have secure, governed access to the context they need (and are permitted to see) to move faster and automate more of their core workloads. Any update in our business is reflected throughout the context graph in about a second, creating a tremendous foundation for us to support customers, engage with teams, and generally deliver a better experience by moving faster and with confidence for reactive as well as proactive work. We've also welcomed Zepz, a fintech holding company that operates digital cross-border money transfer platforms, including WorldRemit and Sendwave, as a customer. They are using Materialize to streamline their microservices as well as deliver an online feature store that creates features from dozens of operational databases. You can read the case study here. You'll find the full details in the release notes, but here are some of my favorite updates! Improved serving performance Predictable query latency under load. Heavy SELECT queries now run on separate threads, so they no longer block small lookups queued behind them. In our tests, p99 latency for single-key lookups running next to heavy scans dropped from 3.64 seconds to 37 milliseconds (98x improvement). Concurrent heavy scans also finish faster. Monitor and right-size your clusters Track hydration history. Every hydration is now recorded with how long it took and its peak memory and disk use. Create a cluster at a generous size, hydrate once, then size down with confidence. A better Clusters page. The Console now shows per-replica CPU, memory, disk, and heap usage, with filters to spot unhealthy clusters quickly. We’ve sped up the page too; on environments with 2,000 objects, the page now loads in 25–50 ms instead of about 600 ms. Build with your coding agents Claude Code and Codex plugins. Install all of the Materialize agent skills with one command, and keep them up to date automatically. Improved connectors Sink to Databricks on Azure (public preview). Following last month’s AWS support, the Iceberg sink now writes to Unity Catalog tables on Azure Data Lake Storage. Vended credentials for GCP BigLake. Iceberg sinks to BigLake can now authenticate with vended credentials as well. Safely drop upstream Postgres constraints. The new EXCLUDE CONSTRAINTS option lets you drop an upstream primary key, unique, or not-null constraint without stalling your source in Materialize. Day 2 Operations Advanced SSO on Self-Managed, including OIDC, SAML, SCIM, and group-to-role mapping. This means you can update access automatically as new people join, leave, or change teams. Group-to-role mapping (generally available, Cloud). Much like you can on Self-Managed, you can now sync group membership from your IdP to Materialize on Materialize Cloud. Operational dashboards (Self-Managed). New Grafana dashboards let you monitor your Materialize installation, including following an upgrade as it rolls out. Query History (Self-Managed). The Console’s Query History view is now available for Self-Managed deployments, to debug query latency and performance. New guides Export data to Snowflake and Databricks, using the Iceberg sink Size clusters appropriately for hydration, use autoscaling to speed it up, and finally reduce hydration resource requirements Protect sensitive columns in your data products Understand the lifecycle of a source and a sink A sneak peek at what's next: far better out-of-core performance, so you can use more disk and run much larger working sets for the same budget, with hot data still served from memory. You can read about how we're building it here. You can reply to me directly with any feedback or questions. I read everything! Nate Stewart, CEO Materialize slack-1 LinkedIn X social-icons Materialize, Inc, 436 Lafayette Street, Floor 6, New York, NY 10003, United States Unsubscribe Manage preferences [Open and remix this design](https://brew.new/browse/templates/email/pt1_k97gzg0wtph68gxmdp5kaz1px98fqem2) [Browse email designs](https://brew.new/browse/templates)