Snowflake adds a cross-region inference setting to increase the availability of AI models.
In Europe, AI models are regularly made available later than in other parts of the world. Before a launch is considered, the companies behind AI models want to be sure that the models comply with all regulatory and legislative requirements.
This can put companies at a competitive disadvantage by limiting innovation opportunities. Snowflake now offers a solution for these companies. The new setting, “cross-region inference,” allows developers to access Cortex AI. Requests for the AI model are handled without paying egress fees.
Data processing in another continent
To give companies access to the AI model without releasing the model in a particular region, the request will be processed in another region. For example, a request from a European company will be processed in America. To use the setting, developers must activate cross-region and determine in which region they want to process the data.
According to Snowflake, data privacy is guaranteed in all cases. That can stay within the confines of one cloud provider if the region where the company is located and the region where the data is to be processed both use AWS. In other cases, encrypted protocols will transfer the data over the internet. Data can currently only be processed on AWS, even with users who activate cross-region for Azure or Google Cloud.
Arun Agarwal, senior product marketing manager at Snowflake, further notes that there may be some latency due to data transmission and processing. “We expect network latency to be negligible compared to LLM-inference latency. We recommend you test your specific usage scenario while inference between regions is enabled.”
Companies that have now accessed Cortex AI through the new capability will only pay for the use of the LLM. New LLMs will be added as soon as Snowflake releases them.
Tip! Snowflake makes AI Data Cloud the brain of any business