I’ve been attending Snowflake Summits for years, and this year’s event was the most forward-looking I’ve seen. If you’re building for the future of data and AI, the 2025 Snowflake Summit felt like a line in the sand. AI is no longer experimental. Enterprises are putting it into production. And data strategy? It’s evolving just as fast.
I had the opportunity to connect with Snowflake’s Cortex team, chat with old friends like Josh Klahr, and listen in on keynotes and customer stories that point to the shape of things to come.
Here are my biggest takeaways from Snowflake Summit 2025 and why I believe the conversation around semantic layers is about to reach a whole new level.
The AI Tipping Point: From Startups to Enterprises
During the opening keynote panel at Snowflake Summit 2025, Sam Altman joined Snowflake CEO Sridhar Ramaswamy and NYSE President Lynn Martin for a thought-provoking discussion moderated by Sarah Guo, Founder and Managing Partner of Conviction. Altman delivered a clear message: AI is no longer just for startups; it’s ready for enterprise scale.
He stressed that real returns from AI will come to organizations willing to scale compute, prioritize high-impact use cases, and adopt AI-native architectures.
At AtScale, we’re seeing this shift play out every day. Enterprises are deploying AI agents, natural language interfaces, and predictive models. However, many still struggle to scale these efforts because their semantic layer foundation lacks the structure and consistency required for trustworthy, enterprise-grade AI.
This is why Snowflake’s Semantic Views announcement was so noteworthy. They shared that Cortex Analyst is already achieving 90% accuracy. That’s promising. But as anyone who has deployed an LLM to interact with enterprise data knows, accuracy is only as good as the quality and consistency of your semantic definitions.
That’s where AtScale comes in. Our semantic layer integrates directly with Cortex, giving AI models governed definitions, consistent metrics, and reusable logic across domains. With AtScale as the foundation, enterprises can achieve 100% accurate responses from Cortex-powered natural language queries, backed by trusted data and a universal semantic model.
Cortex, Cortex AI SQL, and the Role of Semantics
Snowflake is moving fast with Cortex:
- Cortex Analyst: Natural language interface for enterprise data
- Cortex AI SQL: Extensions for querying unstructured data with SQL
- Semantic Views: The foundation that defines how AI (and BI) can interpret enterprise data consistently
At a surface level, this mirrors the work we’ve been doing at AtScale for years, building semantic scaffolding so that humans and machines can both ask smart questions and get trustworthy answers.
But here’s where things diverge.
Semantic Views vs. AtScale’s Semantic Layer: What’s the Difference?
After the product keynote, a question echoed across conversations: How do Semantic Views in Snowflake compare to the AtScale Semantic Layer?
It’s a fair and important question.
Snowflake Semantic Views are scoped specifically to the Snowflake platform. They allow users to define reusable logic and business metrics directly within the Snowflake environment, logic that tools like Cortex and other SQL-based applications can then consume.
This is a strong validation of the semantic layer concept. It reflects an industry-wide shift: We’re moving beyond fragile, hand-coded logic buried in BI dashboards toward governed, reusable metrics that are accessible across teams. By embedding semantics closer to the data, Snowflake is taking a meaningful step toward simplifying how models and metrics are applied.
But here’s the key limitation: semantic views only live inside Snowflake.
That means if your organization uses multiple data platforms or relies on a mix of BI tools and AI interfaces, it still risks creating semantic silos. Bringing semantics closer to the data is valuable, but keeping them locked into a single platform recreates the very inconsistencies that semantic layers were designed to solve.
The AtScale Difference: Open Semantics. Avoiding Lock-In.
AtScale was built to avoid semantic silos, not create new ones. Our semantic layer sits above your data platforms (Snowflake, Databricks, BigQuery, Redshift, and more) and integrates natively with BI tools and AI agents alike.
With AtScale:
- You define your metrics once, and they’re available across tools like Excel, Power BI, Tableau, and even NLQ and LLM interfaces
- Your logic is platform-agnostic, meaning no vendor lock-in or rework when your stack evolves
- Your semantic model is versioned, documented, and treats semantics like code (because that’s how they scale)
Semantic Views are a step in the right direction, but they’re platform-specific semantics, not a universal semantic layer.
If you’re building for long-term AI and BI interoperability, you need open semantics that transcend tools, clouds, and query languages. That’s what we do at AtScale, and it’s why we’re working closely with partners like Databricks, Google, and yes, Snowflake.
Snowflake’s Expanding Ecosystem: Signals Worth Watching
Beyond AI, Snowflake made several announcements that caught my eye:
- Postgres + Crunchy Data Acquisition: Snowflake is doubling down on Postgres compatibility, which expands their reach to more app-centric and operational workloads
- Snowpipe Streaming + Snowpipe Pricing Update: More flexible ingestion and pricing models make streaming use cases more viable
- Internal Marketplace: Enterprises can now build and share internal data products, aligning with the growing trend of data productization
- Lineage with Horizon: Governance and visibility are finally catching up with innovation
- Support for Git-based Workspaces and DBT Projects: A signal that modern data engineering practices are coming to the forefront
These moves show that Snowflake is serious about becoming an all-in-one data, AI, and applications platform. But the more capabilities they add, the more critical it becomes to abstract logic away from the platform.
That’s where an independent semantic layer becomes a business imperative.
Customer Stories That Hit Home
Some of the best moments of the Summit came from customers sharing real-world applications:
- Ericsson: Using Snowflake and Cortex to streamline customer support
- NYSE President Lynn Martin: Discussed how they process over 1.2 trillion incoming order messages while ensuring market integrity, highlighting how precision and governance are table stakes for enterprise-grade data platforms
These examples reminded me of the enterprises we work with at AtScale, organizations like TELUS, Bluemercury, and a leading home improvement retailer that are using semantic layers to align KPIs, scale natural language access, and deliver governed self-service at scale.
The Road Ahead: Why Open Semantics Matter More Than Ever
The semantic layer has moved from theory to necessity. Snowflake’s embrace of semantics proves that the future of analytics, human or AI-driven, relies on well-governed, universally understood data logic.
But we can’t afford to silo that logic within individual platforms. Just as we outgrew hardcoded SQL in dashboards, we’ll outgrow platform-specific semantic views.
AtScale’s commitment is to open semantics-governed models that can power GenAI agents, dashboards, and decision engines, no matter where your data lives or your teams’ tools.
Final Thoughts
Snowflake Summit 2025 confirmed what many of us already believed: AI is real, it’s here, and it’s changing how we work. But it also raised the stakes for managing and governing the data that feeds it.
As AI gets smarter, the semantic layer becomes more essential, not less.
We’re excited to see Snowflake embrace this direction. And we’re even more excited to help enterprises go further, faster, with a semantic foundation that’s open, universal, and built to last.
Until next year,
Dave Mariani
CTO & Co-founder, AtScale
SHARE
Case Study: Vodafone Portugal Modernizes Data Analytics