The short answer: choose by where your source systems already run, by the table format your data is stored in and by the cloud contract you already pay, and treat the query engine as the part you can change later. On 27 October 2026, at Oracle AI World in Las Vegas, Oracle and Snowflake announced that the Snowflake AI Data Cloud will be offered as a native, first-party service on Oracle Cloud Infrastructure: bought from Oracle, paid for with Oracle Universal Credits, provisioned from the OCI console and connected natively to Oracle’s applications and databases. No launch date and no region list were given, and both companies describe it as a plan. Even so, it changes the arithmetic for any company whose ERP, HR or finance system runs on Oracle while its warehouse sits in another cloud, because the warehouse can now come to the data instead of the data crossing clouds every night. Location, format and contract now decide more than feature lists do.
What Changed, and When
| Date | What happened | Source |
|---|---|---|
| 6 October 2026 | SDAIA opens a consultation on amendments to Saudi Arabia’s data protection regulations, including a new Article 23 requirement to store personal data within the Kingdom; comments close on 5 November | Clyde & Co, Reed Smith |
| 9 October 2026 | BigQuery’s Lakehouse ingests change data capture from Delta Lake and Apache Iceberg tables (preview); its Apache Hive catalog endpoint (14 October) and table copies across regions (15 October) become generally available | Google Cloud release notes |
| 12 October 2026 | Microsoft announces partner promotions from 2 November: up to 20% off Microsoft Fabric pay-as-you-go capacity for Azure database customers new to Fabric, and up to 30% off Azure database usage for companies moving SQL Server off their own servers | Microsoft Partner Center |
| 13 October 2026 | Databricks makes writes to its managed Delta tables from outside Apache Spark clients generally available, and lets external engines query Unity Catalog views through its REST API | Databricks release notes |
| 16 October 2026 | Snowflake makes Apache Iceberg v3 table upgrades generally available; Time Travel for Iceberg tables follows in preview on 28 October | Snowflake release notes |
| 27 October 2026 | Oracle and Snowflake announce Snowflake as a planned first-party OCI service, bought and billed through Oracle | Oracle, Snowflake |
What Oracle and Snowflake Actually Said
Snowflake runs today on Amazon Web Services, Microsoft Azure and Google Cloud; OCI would be the fourth. Under the plan a customer contracts with Oracle rather than Snowflake, is billed through OCI, can draw on existing Universal Credit commitments, creates the account from the OCI control plane and administers it in Snowflake’s own tooling. The planned integrations cover Oracle Fusion Cloud Applications (ERP, HCM and SCM), NetSuite, cloud-hosted Oracle Database and Autonomous Database, MySQL HeatWave and Oracle NoSQL Database, and Snowflake says they should reduce the custom pipelines a company builds to get that data out. Cross-cloud sharing, replication and failover with accounts on the other three clouds is planned, subject to technical validation. Both companies attach a product-direction disclaimer: nothing is committed, and timing and pricing may change.
Two facts put the announcement in proportion. Oracle data already reaches Snowflake: the Openflow connector for Oracle has been generally available since 4 March 2026. And Snowflake reports more than 14,500 customers, so this is a distribution agreement for an established product. What is new is the commercial path: a company that negotiates one Oracle agreement a year can bring its warehouse spend inside it, and Oracle, which already runs cloud regions in Saudi Arabia and the UAE, gains a reason for that company to stay.
The Platforms Are Opening Their Tables
The quieter change of the month matters more over five years. A data platform has three layers: the files where the data sits, the table format and catalog that describe them, and the engine that runs the queries. For most of the last decade the engine owned the other two, which made switching expensive. In October all three large platforms moved that boundary. Databricks now lets Spark clients outside Databricks create and write its managed Delta tables, and outside engines read its views through a REST API. Snowflake’s Iceberg tables moved to version 3 of the open specification with full support. BigQuery can take a feed of changes from Delta and Iceberg tables held elsewhere, and its catalog exposes a Hive endpoint that other tools understand. A warehouse built this year on Iceberg or Delta tables with an open catalog can be read by a second engine, moved between clouds by copying files, and priced competitively at renewal. One built on a proprietary format can only be renewed.
Where the Data Sits Is Becoming a Legal Question
On 6 October 2026 the Saudi Data and Artificial Intelligence Authority published draft amendments to the Implementing Regulations of the Personal Data Protection Law for consultation until 5 November. The draft adds to Article 23 an express requirement that controllers store personal data within the Kingdom, while still permitting transfers abroad that comply with the law. Clyde & Co and Reed Smith both read the wording as ambiguous: it may mean a copy must be kept in Saudi Arabia, or that in-Kingdom storage is the default and lawful transfers need no local copy. If adopted, the draft takes effect 60 days after publication in the Official Gazette. For a company with Saudi customers or staff and a warehouse in Frankfurt or Virginia, the region a platform runs in has moved from the architect’s desk to the legal department’s. Our article on data residency in Saudi Arabia and the UAE lists the rules in force and the regions open in both countries.
Microsoft’s Answer: Pay Less to Move
Microsoft is discounting rather than opening. From 2 November 2026 its cloud solution provider partners can offer up to 20% off Microsoft Fabric pay-as-you-go capacity to Azure database customers with no prior Fabric spend, and up to 30% off Azure database usage to companies moving SQL Server off their own servers. Its documentation describes a preview assessment inside Azure Data Factory and Synapse that grades each pipeline for upgrade and converts it, with triggers switched off until someone turns them on. For a company already on Azure SQL or Synapse the price of staying just fell; the question is whether Fabric’s OneLake, which stores data as Delta tables, leaves the door open that the others opened this month.
A Decision Sheet for the Next Quarter
- Map the sources before the destination. List every system that will feed the platform, the cloud and region it runs in, and whether it holds personal data of Saudi residents. The warehouse belongs near the heaviest sources; Saudi data needs a home in the Kingdom or a documented reason why not.
- Price the movement. Find the data transfer lines on the last three months of cloud invoices. If the warehouse is in a different cloud from the ERP, that line is what the Oracle and Snowflake plan targets, and it decides whether waiting is worth anything.
- Choose the table format first, the engine second. Require Iceberg or Delta tables with a catalog another engine can read, and ask each vendor to show a second tool writing to and reading from the same table.
- Check which contract already covers it. Universal Credits, an Azure commitment or an AWS agreement may already pay for the platform you are evaluating; bought inside a commitment it is cheaper to start and harder to leave.
- Get dates and regions in writing. Snowflake on OCI has neither. Plan a 2027 go-live on what exists today and treat the OCI service as a later re-platforming option, which an open table format keeps open.
- Keep the pipelines portable. Transformations in SQL and dbt, orchestration in Airflow and tests on every load move between platforms; a pipeline built in one vendor’s proprietary tool does not.
Where SoftwareBiz fits. Our data engineering and modernization service designs and builds the pipelines, warehouses and lakehouses that bring your systems together on BigQuery, Snowflake, Databricks, Microsoft Fabric or Amazon Redshift, on open table formats, with quality tests on every load and the scheduling that keeps the bill predictable, and moves legacy databases to managed platforms with the data validated on both sides. Where the sources are many, our data integration and governance team records the lineage of every field, so the question of where Saudi personal data sits has an answer. Where the decision is which cloud and region to run in, our cloud migration and modernization team plans the landing zone and the move.
What to Do This Quarter
If your ERP runs on Oracle and your reports run somewhere else, work through the six points before renewing either contract, and ask Oracle in writing when Snowflake on OCI will be available in your regions. If the platform decision is open, settle the format and the region this quarter and the engine after. The October announcements show the vendors expect you to be able to leave, and are pricing accordingly.