What happened
- Teradata announced general availability of Elastic Python Compute in AI Studio. The environment runs Python tasks in parallel with Ray and lets teams train, develop and serve machine-learning workloads from the data platform.
- The company says teams can bring compute to the data, host models locally and avoid external token-based inference calls. The offer is available on AWS and Microsoft Azure; Google Cloud is expected in the fourth quarter of 2026.
- Teradata says its on-premises Factory deployment is also supported.
Why it matters
- Moving data for analysis is an architectural decision and a risk decision. For a company processing financial, health or telecommunications information, each copy raises questions about permissions, residency, retention, cost and auditability. Running code close to the database may reduce friction, but it does not by itself solve model quality or determine who is accountable for an automated decision.
- Teradata presented customer examples in banking, telecommunications and the public sector. It attributes a speed improvement to a bank pipeline and describes a telecom operator analyzing records inside its own infrastructure. These are supplier-reported cases, with no independent evaluation in the announcement. They should not be read as guaranteed outcomes for every deployment.
- For teams in Chile and the region, the practical appeal is testing models against data already covered by internal controls without first creating an export path to another service. The choice still requires checking where models run, where backups live, who holds encryption keys and the actual residency of every component. Saying data stays within a platform is not a substitute for a technical map.
What’s next
- Teradata offers the capability on AWS and Azure.
- Google Cloud availability is planned for Q4 2026; the company did not give a more precise date or separate pricing for the feature in this announcement.
Bottom line
The question is not how many models a team can launch. It is whether it can repeat the process with protected data, visible costs and a useful audit trail.
Sources
Written by Mamífero. Edited by Rodrigo Cornejo. This draft was assisted by AI and requires human editing before publication. See how we select and verify these notes.