Don’t just say data cloud: Snowflake puts AI first and believes its mission is to lead enterprises into a future where AI finds its way into data easily, securely, and as affordably as possible.

“You enter the AI ​​data cloud“, booms from the loudspeakers of the Moscone Conference Center in San Francisco, at the beginning of the big keynote speech of the Snowflake Summit.

The new CEO Sridhar Ramaswamy was able to shine on stage at the opening of the conference the day before. For the first time, he introduced himself in detail to customers and partners, placing great emphasis on his experience, training and love of drums.

Today, Ramaswami acts more as a hype man for Snowflake’s big star: Benoit Dageville, President of Product, but above all co-founder of Snowflake and fan of the better ski vacation. Snow is never far away at the Snowflake Summit, even if the sun is doing its best outside the conference center in San Francisco.

Snowflake as Cloud OS

Dageville makes no secret of it: Snowflake is now a modern cloud operating system that is ideal for building and running your AI apps. “The Snowflake architecture is really the best for all your AI applications,” he says confidently.

The focus on AI is omnipresent at this edition of the summit. Last year, Snowflake jumped on the AI ​​hypemobile, but today AI has become core to the Snowflake mission. Snowflake calls itself “the AI ​​data cloud company.”

“We recently adopted this term,” explains Baris Gultekin, head of AI at Snowflake. “We have felt the impact of AI very clearly since last year. It is important to define ourselves this way from now on because we believe that there is no AI strategy without a data strategy.”

Five pillars

But don’t just say AI. Christian Kleinerman, EVP of Product, explains: “It’s easy to create a chatbot generally is correct. But who wants that anyway?” The requirements for AI in the corporate environment are much higher than those for consumer AI. Dageville sees five crucial pillars on stage that must support AI at the corporate level:

• A good data basis;
• Elastic computing power on demand;
• Access to the right models;
• Security and management at all levels;
• Collaboration internally and externally.

First, Dageville opens five open doors. The story becomes more concrete as he explains how Snowflake is ready to support organizations in the five areas today. The approximately 15,000 participants can get a picture of Snowflake as an end-to-end platform for everything related to AI.

Twice complete

Data is the foundation of Snowflake. “Our platform is Data complete”claims Dageville. Snowflake can actually handle structured, semi-structured and unstructured data, as well as transactional and relational data. “It’s not just that you can store and share all kinds of data that’s important, but also how easy it is,” he adds. “Everything works seamlessly.”

The next pillar is computing power and you will never guess, Snowflake is it mathematically completed. The Snowflake platform allows you to work with both CPU processing power and GPU processing power, depending on your needs. The latter is a novelty that Dageville introduced last year at the Summit in Las Vegas together with Nvidia CEO Jensen Huang. This year, the two parties are further expanding their integration, with more possibilities for inference acceleration via Nvidia Triton servers.

Models for every taste

For access to the right models, Dageville refers to Cortex AI. This platform offers access to various LLMs, including those from Meta and Mistral. Snowflake itself has also built a powerful LLM with Arctic. Arctic is based on an innovative approach that makes the LLM more efficient than its competitors, both in terms of training and use.

Snowflake also focuses on efficiency, including by providing small LLMs optimized for specific tasks. Consider Arctic Tilt, which powers Document AI. This model has less than a billion parameters, but still delivers top results in interpreting unstructured data from documents. Snowflake sees the availability of large and small models with different strengths as a major advantage.

Secure data cloud

When it comes to data security and governance, Dageville doesn’t mince words. Snowflake was designed to be a solution for efficient and secure data management with a zero-copy approach that ensures that all data remains subject to the correct access rights. Because AI apps run within the Snowflake platform, these governance rules apply automatically.

Finally, collaboration is discussed. This aspect makes Snowflake not a data warehouse or lakehouse specialist, but a data cloud provider. The platform not only enables the internal and external exchange of data, regardless of whether it is anonymized or not and always under full control, but customers can also share and sell applications on the data.

Dageville: “Snowflake is the only platform that allows you to share data and AI apps between clouds around the world. You never have to move the data because applications are built on top of the data.”

Not ripped out of the pot

The argument makes sense. Snowflake has evolved into a data specialist in the cloud. Even before the AI ​​hype, the Snowflake platform offered a way to bring applications to the data and not the other way around. This architecture is ideally suited for AI applications in a cloud-native context.

Snowflake has its Container Services and its native apps that enable companies to run their own applications, LLMs or third-party apps alongside their data, whether accelerated by Nvidia hardware or not. The added value of the approach is clear.

The San Francisco summit focused on numerous announcements, but all of them fit into the AI ​​Data Cloud vision. Some new features benefit developers by giving them more tools to work with their data, while others make it easier to integrate data into Snowflake. Snowflake continues to innovate in the AI ​​space as well.

A well-deserved stamp

Every company calls itself an AI company these days, which sometimes makes the term ring hollow. However, Dageville, Ramaswamy and their collaborators at this conference show that adding “AI” to the “data cloud” is more than just a smart move by the marketing team. Snowflake has earned the AI ​​stamp, as has the momentum among customers and partners that is almost palpable at the Moscone Center.