From AI Experiments to Enterprise-Scale Agents
Snowflake’s latest moves signal a shift in enterprise AI agents from isolated pilots toward production-scale autonomous systems deployment that connects models, data, and actions on a governed platform. Instead of treating AI as a separate experimentation layer, Snowflake is positioning its AI Data Cloud as the core environment where agents reason over trusted data and trigger workflows across business systems. CEO Sridhar Ramaswamy describes this as entering the “agentic enterprise,” where intelligence is only useful when it drives action on top of reliable data. That framing explains why Snowflake is aligning AI infrastructure, coding agents, and governance into a single control plane. Enterprises want AI that can operate across CRMs, analytics, and operations while respecting policy, privacy, and identity boundaries. The result is a race to build data governance platforms that can keep up with far more autonomous AI behavior than yesterday’s chatbots.
A $6B AWS Infrastructure Commitment Built for Agent Workloads
Snowflake plans to spend USD 6 billion (approx. RM27.6 billion) on AWS Graviton CPUs and AI accelerators over the next five years, a clear AI infrastructure investment tuned for agent-heavy workloads. The company has already shifted much of its compute from Intel and AMD chips to Amazon’s Arm-based Graviton instances, whose latest generation packs 192 Arm Neoverse V3 cores and high-speed memory. While large models still run on GPUs, the surrounding tools—SQL queries, Python code, and orchestration logic—live on CPUs, which now gate the responsiveness of each agent. Under the extended AWS deal, Snowflake will run and train its GenAI services, including Cortex AI for natural language to SQL, summarization, and sentiment analysis, on a blend of GPUs and Graviton cores. According to Amazon, Snowflake’s lifetime AWS Marketplace sales crossed USD 7 billion (approx. RM32.2 billion), underscoring the commercial stakes behind this infrastructure bet.
Natoma and MCP: Governance as the Agent Control Layer
Snowflake’s plan to acquire Natoma targets the growing governance gap as enterprises deploy autonomous systems that can call APIs, touch production data, and trigger workflows. Natoma’s Model Context Protocol (MCP) infrastructure—an emerging standard for connecting AI systems to tools and data—will give Snowflake a natively integrated identity and governance layer for AI agents. That layer is designed to control which MCP tools agents can access across SaaS apps, cloud infrastructure, on‑prem systems, CRMs, email, Jira, Slack, and internal APIs. As organizations worry about fragmented policies, shadow AI, and data leakage, the acquisition brings governance closer to where actions occur. Instead of bolting compliance on after the fact, Snowflake aims to have policy, access control, and observability embedded in the same environment that runs the agents. In effect, data governance platforms and AI orchestration are converging into a single architectural concern.
CoCo and Anthropic: Coding Agents with Built-In Governance
On the development side, Snowflake is expanding its agentic control plane through Snowflake CoCo, the coding agent previously known as Cortex Code. CoCo gives builders a governed environment to automate workflows, develop applications, and operationalize AI on enterprise data using outcome-focused prompts. New capabilities bring CoCo into desktop, mobile, Slack, VS Code, Claude Code, and Microsoft Excel, so enterprise AI agents can be designed and iterated from familiar tools. As a core part of Snowflake’s platform, CoCo understands schemas, governance rules, and business context by default, reducing the risk that AI-generated code violates policy or misuses data. Snowflake Datastream then supplies real-time data from Apache Kafka into the same platform, keeping agents aligned with current events instead of stale snapshots. Customers like Fanatics, Thomson Reuters, and WHOOP show how this model can accelerate production-ready AI while keeping development and governance tightly coupled.
Why Consolidated Platforms Are Winning the Agentic Enterprise
Snowflake’s AWS expansion and Natoma acquisition highlight where enterprise AI agents are running into friction: fragmented tools for marketing orchestration, data governance, and interoperability. As organizations try to move beyond copilots, they discover that autonomous systems deployment needs more than powerful models—it requires integrated identity, policy, observability, and action routing. That is pushing buyers toward consolidation on platforms that combine AI infrastructure investment, governed data, and agent tooling in one place. Snowflake’s pitch is a unified control plane where CoCo, Cortex Agents, and Snowflake Intelligence share common governance and semantics across clouds and tools. For enterprises, the promise is fewer silos between marketing stacks, analytics, and operational systems, and less custom glue code to keep policies aligned. As agentic AI matures, success will depend less on individual models and more on how well platforms coordinate data, governance, and actions end to end.
