Google Cloud has spent greater than two years educating Gemini to learn Oracle saved procedures and rewrite them for PostgreSQL. On August 11, 2026, it prolonged that work once more, giving Gemini in Database Migration Service the power to map a complete schema’s relationships earlier than changing a single line of code. The replace seems incremental by itself. The race it belongs to shouldn’t be.
A Functionality Two Years within the Making
The latest model of Google Cloud’s Database Migration Service (DMS) provides what the corporate calls full schema context evaluation. As a substitute of translating one saved process at a time, Gemini now critiques desk relationships, knowledge varieties, and cross-procedure dependencies throughout a complete database earlier than producing transformed code. The console shows supply and goal code facet by facet, marks every transformed object as Transformed, Warning, or Motion Required, and requires an individual to evaluate and validate the output earlier than something strikes to staging or manufacturing.
The form of conversion Gemini handles is the half that used to require a database specialist fluent in each dialects. Oracle’s DECODE operate, a typical option to write conditional logic inside a saved process, has no direct PostgreSQL equal; Gemini rewrites it as a CASE expression and explains why. Oracle’s NVL turns into PostgreSQL’s COALESCE. Multiply that sample throughout the saved procedures a legacy Oracle or SQL Server database accumulates over ten or fifteen years, and the enchantment of automating it turns into apparent.
None of that is Google’s first try on the drawback, both. The corporate launched Gemini-assisted code conversion for Oracle-to-PostgreSQL migrations in preview at Google Cloud Subsequent in April 2024, prolonged it to SQL Server sources a yr later, and took the core conversion options to normal availability in September 2025. A associated GA milestone, Gemini-powered conversion high quality assessments, adopted in Might 2026. August’s replace is nearer to a fourth or fifth iteration than a debut. “Google Cloud’s Database Migration Service simplifies the method of modernizing databases,” Shashank Srivastava, a software program engineering supervisor at Wayfair, stated when Google added SQL Server help in 2025. “This makes the migration course of much less guide and time-consuming, permitting groups to spend extra time on improvement and fewer on infrastructure.”
The Final Mile Each Cloud Vendor Is Now Preventing Over
Database migrations hardly ever stall on tables and columns. Rule-based conversion instruments have dealt with that half nicely for years. They stall on procedural code: the saved procedures, triggers, and customized capabilities written in a vendor’s proprietary SQL dialect, dense with enterprise logic no person needs to rewrite by hand. Amazon constructed its personal reply into AWS Database Migration Service in December 2024, including generative AI by Amazon Bedrock to its Schema Conversion software. In a single instance AWS printed at launch, rule-based conversion alone translated 100% of storage objects however solely 57% of code objects; including generative AI introduced code protection to 100%. AWS says the AI-assisted software now robotically converts as much as 90% of schema objects from business databases, and it has since prolonged the function to further supply databases and areas by 2025 and into 2026.
Microsoft has approached the identical drawback from the applying layer. Its GitHub Copilot modernization tooling, up to date as not too long ago as June 2026, now helps Java purposes rewrite Oracle SQL for PostgreSQL and swap in managed-identity authentication as a part of the identical migration. That function stays in preview. Individually, Microsoft previewed an Azure Copilot Migration Agent in March 2026 that automates VMware discovery and landing-zone planning for broader infrastructure strikes, although it nonetheless fingers the precise cutover to Azure Migrate. Three hyperscalers have now constructed generative AI straight into the purpose the place migrations historically received caught, converging on the identical technical reply inside roughly two years of one another. A bottleneck that cussed, mounted by each main cloud supplier in the identical quick window, was costing all three of them actual enterprise offers, not simply engineering time.
Whoever Converts Your Code First Often Retains It
Framing issues right here. Gemini in DMS doesn’t convert Oracle or SQL Server code into some impartial, transportable format. It converts it into PostgreSQL working on AlloyDB or Cloud SQL, each Google merchandise. AWS’s software converts into Aurora or RDS. Microsoft’s modernization tooling factors at Azure. Every hyperscaler’s AI migration assistant solves an actual technical drawback, and it additionally occurs to be the simplest software out there for making a switching resolution everlasting earlier than a buyer has completed evaluating options. The engineering is real. So is the inducement behind it.
For many IT groups going through a real deadline, that commerce is affordable. It nonetheless deserves extra scrutiny than a brand new AI function normally will get. Oracle’s PL/SQL and PostgreSQL’s PL/pgSQL differ in how they deal with NULL comparisons, exception scoping, and implicit transaction boundaries, variations that hardly ever floor in a demo however can quietly change what a monetary calculation or a listing verify returns as soon as the code is working in manufacturing. Google’s personal interface concedes the purpose: the side-by-side code evaluate, the Warning and Motion Required labels, and the requirement that an individual approve each transformed object earlier than deployment are all implicit admissions that Gemini’s output nonetheless wants somebody who understands the supply database checking its work. That could be a wise design alternative, and a quiet admission too: automated conversion of business-critical logic shouldn’t be but one thing to approve on religion, whichever cloud is doing the changing.
None of this makes AI-assisted migration a nasty wager. Three hyperscalers converging on the identical repair inside about two years says the underlying drawback, changing years of amassed saved procedures with out months of guide labor, was actual and costly sufficient to justify the funding every of them made. What’s value watching subsequent is which cloud proves its conversion accuracy below unbiased scrutiny fairly than in a launch put up, as a result of that’s the declare that may truly transfer enterprises nonetheless working on Oracle and SQL Server. A quick, assured migration to the flawed database continues to be a migration. It’s only a costlier mistake to undo.
Google Cloud has spent greater than two years educating Gemini to learn Oracle saved procedures and rewrite them for PostgreSQL. On August 11, 2026, it prolonged that work once more, giving Gemini in Database Migration Service the power to map a complete schema’s relationships earlier than changing a single line of code. The replace seems incremental by itself. The race it belongs to shouldn’t be.
A Functionality Two Years within the Making
The latest model of Google Cloud’s Database Migration Service (DMS) provides what the corporate calls full schema context evaluation. As a substitute of translating one saved process at a time, Gemini now critiques desk relationships, knowledge varieties, and cross-procedure dependencies throughout a complete database earlier than producing transformed code. The console shows supply and goal code facet by facet, marks every transformed object as Transformed, Warning, or Motion Required, and requires an individual to evaluate and validate the output earlier than something strikes to staging or manufacturing.
The form of conversion Gemini handles is the half that used to require a database specialist fluent in each dialects. Oracle’s DECODE operate, a typical option to write conditional logic inside a saved process, has no direct PostgreSQL equal; Gemini rewrites it as a CASE expression and explains why. Oracle’s NVL turns into PostgreSQL’s COALESCE. Multiply that sample throughout the saved procedures a legacy Oracle or SQL Server database accumulates over ten or fifteen years, and the enchantment of automating it turns into apparent.
None of that is Google’s first try on the drawback, both. The corporate launched Gemini-assisted code conversion for Oracle-to-PostgreSQL migrations in preview at Google Cloud Subsequent in April 2024, prolonged it to SQL Server sources a yr later, and took the core conversion options to normal availability in September 2025. A associated GA milestone, Gemini-powered conversion high quality assessments, adopted in Might 2026. August’s replace is nearer to a fourth or fifth iteration than a debut. “Google Cloud’s Database Migration Service simplifies the method of modernizing databases,” Shashank Srivastava, a software program engineering supervisor at Wayfair, stated when Google added SQL Server help in 2025. “This makes the migration course of much less guide and time-consuming, permitting groups to spend extra time on improvement and fewer on infrastructure.”
The Final Mile Each Cloud Vendor Is Now Preventing Over
Database migrations hardly ever stall on tables and columns. Rule-based conversion instruments have dealt with that half nicely for years. They stall on procedural code: the saved procedures, triggers, and customized capabilities written in a vendor’s proprietary SQL dialect, dense with enterprise logic no person needs to rewrite by hand. Amazon constructed its personal reply into AWS Database Migration Service in December 2024, including generative AI by Amazon Bedrock to its Schema Conversion software. In a single instance AWS printed at launch, rule-based conversion alone translated 100% of storage objects however solely 57% of code objects; including generative AI introduced code protection to 100%. AWS says the AI-assisted software now robotically converts as much as 90% of schema objects from business databases, and it has since prolonged the function to further supply databases and areas by 2025 and into 2026.
Microsoft has approached the identical drawback from the applying layer. Its GitHub Copilot modernization tooling, up to date as not too long ago as June 2026, now helps Java purposes rewrite Oracle SQL for PostgreSQL and swap in managed-identity authentication as a part of the identical migration. That function stays in preview. Individually, Microsoft previewed an Azure Copilot Migration Agent in March 2026 that automates VMware discovery and landing-zone planning for broader infrastructure strikes, although it nonetheless fingers the precise cutover to Azure Migrate. Three hyperscalers have now constructed generative AI straight into the purpose the place migrations historically received caught, converging on the identical technical reply inside roughly two years of one another. A bottleneck that cussed, mounted by each main cloud supplier in the identical quick window, was costing all three of them actual enterprise offers, not simply engineering time.
Whoever Converts Your Code First Often Retains It
Framing issues right here. Gemini in DMS doesn’t convert Oracle or SQL Server code into some impartial, transportable format. It converts it into PostgreSQL working on AlloyDB or Cloud SQL, each Google merchandise. AWS’s software converts into Aurora or RDS. Microsoft’s modernization tooling factors at Azure. Every hyperscaler’s AI migration assistant solves an actual technical drawback, and it additionally occurs to be the simplest software out there for making a switching resolution everlasting earlier than a buyer has completed evaluating options. The engineering is real. So is the inducement behind it.
For many IT groups going through a real deadline, that commerce is affordable. It nonetheless deserves extra scrutiny than a brand new AI function normally will get. Oracle’s PL/SQL and PostgreSQL’s PL/pgSQL differ in how they deal with NULL comparisons, exception scoping, and implicit transaction boundaries, variations that hardly ever floor in a demo however can quietly change what a monetary calculation or a listing verify returns as soon as the code is working in manufacturing. Google’s personal interface concedes the purpose: the side-by-side code evaluate, the Warning and Motion Required labels, and the requirement that an individual approve each transformed object earlier than deployment are all implicit admissions that Gemini’s output nonetheless wants somebody who understands the supply database checking its work. That could be a wise design alternative, and a quiet admission too: automated conversion of business-critical logic shouldn’t be but one thing to approve on religion, whichever cloud is doing the changing.
None of this makes AI-assisted migration a nasty wager. Three hyperscalers converging on the identical repair inside about two years says the underlying drawback, changing years of amassed saved procedures with out months of guide labor, was actual and costly sufficient to justify the funding every of them made. What’s value watching subsequent is which cloud proves its conversion accuracy below unbiased scrutiny fairly than in a launch put up, as a result of that’s the declare that may truly transfer enterprises nonetheless working on Oracle and SQL Server. A quick, assured migration to the flawed database continues to be a migration. It’s only a costlier mistake to undo.
