What Snowflake’s $6B AWS Commitment Really Means
Snowflake’s $6 billion (approx. RM27.6 billion) AWS commitment is a long-term infrastructure deal to run its data warehouse and AI services on Amazon’s Graviton CPUs and AI accelerators so enterprises can keep analytics and generative AI close to governed data instead of sending information to external tools. This expanded Snowflake AWS deal is framed as a way to reduce friction between customer data and a growing set of AI services built on AWS, particularly through platforms like Cortex AI that sit directly on top of Snowflake data. The company is already shifting more compute from Intel and AMD chips to Arm-based Graviton instances, reflecting how data warehouse AI acceleration increasingly depends on high-core-count CPUs as much as on GPUs. With lifetime AWS Marketplace sales above USD 7 billion (approx. RM32.2 billion), Snowflake is wagering that higher performance and tighter integration will translate into future AI-driven revenue.
Graviton CPUs, GPUs and the Cost of Faster Enterprise Analytics
Snowflake’s spend is anchored on AWS Graviton, which now reaches 192 Arm Neoverse V3 cores per chip and high memory bandwidth to feed AI-adjacent workloads. While models still run on GPUs, the day-to-day work of enterprise analytics infrastructure—SQL queries, Python functions, data transformation and coordination for AI agents—remains CPU-bound. By standardising on Graviton for these tasks and combining it with AWS GPU-accelerated instances for training and inference, Snowflake is building a tightly coupled stack for data warehouse AI acceleration. The upside is performance: more cores per dollar and better density for large numbers of concurrent AI requests. The downside is concentration risk and a large fixed infrastructure outlay. Snowflake is betting that its agent-focused services, such as natural language to SQL, summarisation and sentiment analysis inside Cortex AI, will draw enough workloads to justify USD 1.2 billion (approx. RM5.5 billion) in annual cloud spend.
Natoma and the Rise of Agentic AI Governance
In parallel with the infrastructure deal, Snowflake’s move to acquire Natoma shows how quickly enterprise priorities are shifting from AI experiments to controllable, action-taking agents. Natoma’s Model Context Protocol (MCP) infrastructure is designed to connect AI systems to tools and data sources while enforcing governance and identity, addressing fears of shadow AI, fragmented controls and data leakage. Snowflake plans to embed a “natively integrated governance and identity layer” for its AI agents and services—Cortex Agents, Snowflake Intelligence and Cortex Code—so they can safely work across SaaS apps, cloud infrastructure, CRMs, email, Jira, Slack and internal APIs. For enterprises, this links AI reasoning directly to operational systems without losing oversight. Snowflake cites internal research claiming 96% of organisations still face major barriers to scaling AI, suggesting that governance, not raw model quality, is now the main bottleneck to broader adoption.
Strategic Pivot in Cloud Data Platform Competition
Together, the Natoma deal and AWS infrastructure expansion signal a clear strategic pivot: Snowflake wants to be the default control plane for enterprise AI agents running on governed data. This marks an escalation in cloud data platform competition, where Databricks and others are also vying for data engineering and AI mindshare. Snowflake is deepening its cloud-native positioning by tying its roadmap tightly to AWS silicon, marketplace distribution and joint go-to-market programs. According to Amazon, Snowflake’s AWS Marketplace sales exceeded USD 2 billion (approx. RM9.2 billion) in 2025 alone, showing strong demand for its existing analytics stack. The open question is long-term infrastructure ROI. Locking into Graviton and AWS accelerators may deliver performance and co-selling benefits now, but it reduces optionality across other clouds and chips. If agentic AI workloads consolidate on a few platforms, Snowflake’s $6B bet could either secure a durable foothold—or constrain future flexibility.
