Overview of Big Technology Podcast (episode with Sridhar Ramaswamy)
This episode features Sridhar Ramaswamy (CEO of Snowflake) in a wide-ranging conversation with host Alex Kantrowitz about the current state of the AI race, the practical reality of agentic AI in enterprise, Snowflake’s product strategy (Snowflake Intelligence and CortexCode), shadow AI (employee-led adoption), market reactions to AI, and the evolving role of open-source and international model makers.
Key takeaways
- The AI landscape is extremely fluid — “the AI race changes every month.” Leads that model makers gain can be large but short-lived.
- Top model makers today (OpenAI, Anthropic, Google/Gemini) have made big strides, but competition from optimized approaches and open-source models is accelerating.
- Enterprise value of AI comes from making work actionable and grounded in trusted data — not just from the model itself.
- Snowflake’s strategy: be the data-and-agent platform that connects enterprise data, models, and operational systems while enabling interoperability (not becoming a passive data pipe).
- Shadow AI (employees adopting consumer AI tools) is a primary, bottom-up driver of enterprise adoption — companies should embrace and govern it, not only block it.
- Market multiples for software are under pressure because AI-as-a-bolt-on to SaaS does not automatically create value — differentiation will come from workflows, integrations, and measurable outcomes.
Topics discussed
- State of the AI race (OpenAI vs Google vs Anthropic vs open-source)
- Hardware and model axes: NVIDIA/OpenAI collaborations vs Google’s TPU + Gemini/DeepMind stack
- How Snowflake is applying agentic AI in enterprise: Snowflake Intelligence, sales agent, CortexCode
- Concrete enterprise use cases (pricing optimization for a manufacturer; customer support debugging)
- Shadow AI / personal agents built by employees and their security implications
- SaaS valuation compression and market expectations for AI-driven growth
- The role of open-source and non-U.S. models in accelerating global innovation
Product and technical details (Snowflake)
- Snowflake Intelligence: an agentic product focused on enterprise data — early traction: >2,000 customers shortly after GA.
- CortexCode: Snowflake’s data coding agent (launched to GA) to automate tasks like data ingestion, database setup, model building, and deploying agentic apps via natural language and reusable "skills".
- Internal use cases:
- Sales agent that aggregates customer info, meeting notes, use cases — programmable and used by Sridhar.
- Support tooling: agentic tools plus a “builder-user” model reduced debugging time by ~10x in complex cases.
- Snowflake emphasizes integration (APIs/MCPs) so data products and agents can be accessed across other platforms and UIs, preserving customer choice.
Notable quotes / insights
- “The AI race changes every month.”
- “A great new model can sometimes end up producing a lead that's like a year long, which is an eternity in today's world.”
- “What does work look like in the future?” — framing AI as a means to shift work to agent-driven briefs, recommendations, and actions.
- “The model is everything and nothing else matters — we approach it from the viewpoint of the entirety of the experience.”
- On shadow AI: bottom-up adoption persists because these tools often deliver 10x faster results on tedious work.
Concrete examples & data points mentioned
- Snowflake is a ~$59 billion public data cloud company.
- Snowflake Intelligence reached >2,000 customers within months of GA.
- OpenAI (chat) web visits grew ~50% Jan 2025–Jan 2026; Google’s Gemini/chat visits grew ~647% in the same period (illustrating rapid Google uptake).
- Anthropic took about two years after GPT-4 to reach comparable quality; leads can take substantial time to replicate.
- Snowflake announced a $200M partnership with OpenAI (enterprise collaboration).
- Internal Snowflake metric: up to 10x reduction in time to debug complex support cases using agent tools.
Predictions & implications (from Sridhar)
- 2026: Shadow AI (individual/employee adoption of consumer tools) will continue to drive enterprise adoption from the bottom up.
- Big tech’s monopoly on model-building may loosen as new training methods and open-source models emerge (e.g., DeepSeek-style approaches and regional model makers).
- Winners will be those who combine strong models with productized workflows, integrations, and measurable value — not those who only ship a model.
- Interoperability will matter: enterprises will demand choices (data can be exposed to enterprise chat tools, Snowflake agents, or third-party UIs).
Practical recommendations for enterprises
- Embrace shadow AI constructively:
- Identify and empower “AI champions” who pilot tools safely and spread best practices.
- Provide secure, sandboxed environments (e.g., managed cloud instances) instead of allowing unregulated local deployments.
- Prioritize agent-focused workflows that connect to trusted enterprise data and operational systems (not just chat interfaces).
- Invest in guardrails: data access controls, auditing, and approval workflows before broad rollouts.
- Measure ROI on AI projects: prioritize projects with clear actionability (pricing optimization, incident resolution, sales enablement).
- Maintain interoperability: avoid vendor lock-in where your data becomes a dumb backend to a single model/UI.
What to watch next
- Market movement: how software/SaaS multiples evolve as AI features mature and buyers demand value, not just models.
- New training/efficiency breakthroughs (DeepSeek-style) and impact from open-source and non-U.S. model labs.
- Enterprise adoption trends: how many firms shift from pilot to production agentic workflows and which industries show the fastest measurable gains.
- The balance between central large-model platforms vs specialized vertical or company-specific agents.
Bottom line
Sridhar frames the present moment as early but transformative. The key battleground isn’t only who builds the best foundational model, but who turns models + enterprise data + integrations into repeatable, guarded, measurable workflows that let users act faster and better. Shadow AI and agentic platforms are forcing fast adoption; successful companies will pair productized agent experiences with robust governance and interoperability.