Snowflake’s $6B AWS Commitment and the New AI Infrastructure Playbook
Snowflake’s USD 6 billion (approx. RM27.6 billion) AWS commitment is a long-term cloud infrastructure spending plan in which a software vendor agrees to secure large amounts of compute capacity, including Graviton CPUs and AI accelerators chips, so it can run and sell advanced AI services at scale. Announced as a multi‑year strategic collaboration, the deal is Snowflake’s largest cloud spend commitment so far and centers on AWS Graviton compute for general workloads plus GPU‑accelerated EC2 instances for AI model training and inference. According to The New Stack, Snowflake’s agreement will support both its traditional data warehousing and its growing AI portfolio. This structure marks a clear shift away from a pure software licensing model toward owning more of the enterprise AI infrastructure story, with Snowflake taking on procurement risk in exchange for predictable access to cloud capacity.
From Software Licensing to Shared Hardware Risk
Snowflake’s AWS deal shows how AI infrastructure is moving directly into software balance sheets. In the earlier cloud era, software companies tried to avoid deep infrastructure commitments because they raised capital needs and added operational complexity. Now, serious AI offerings depend on dependable access to GPUs and CPUs, which often requires large, multi‑year commitments before revenue fully matches usage. Snowflake’s USD 6 billion (approx. RM27.6 billion) pledge fits this pattern: it locks in compute so the company can promise customers reliable enterprise AI infrastructure while tying its success to efficient use of that capacity. Startup Fortune notes that AWS is not only Snowflake’s primary cloud provider but also a major distribution channel via AWS Marketplace, where lifetime sales have crossed USD 7 billion (approx. RM32.2 billion). That tighter go‑to‑market alignment comes with higher exposure to infrastructure costs and gross‑margin pressure.
Graviton CPUs, AI Accelerators and the Cloud Chip Stack
A central feature of the Snowflake AWS investment is the mix of Graviton CPUs and AI accelerators chips that underpin its AI strategy. Snowflake has shifted more compute from Intel and AMD to Amazon’s Arm‑based Graviton processors, now in their fifth generation and packing 192 Arm Neoverse V3 cores with high‑speed memory channels. These CPUs handle the non‑model tasks around AI, such as SQL queries and Python functions that agentic AI systems call repeatedly. For model training and inference, Snowflake will rely on GPU‑accelerated EC2 instances, a signal that NVIDIA hardware remains central even as AWS promotes its in‑house designs. The Register reports that the company will run its GenAI models and services on this combined stack, allowing Cortex AI functions like natural language‑to‑SQL and sentiment analysis to operate close to customer data while AWS further consolidates chip design, infrastructure, and software layers.
Natoma, Cortex AI and the Race to Enterprise AI Platforms
Snowflake’s parallel acquisition of Natoma, alongside its AWS commitment, underlines a dual strategy: build higher‑value AI capabilities while securing the compute base needed to run them. Under CEO Sridhar Ramaswamy, Snowflake is repositioning from a cloud data warehouse into “the platform for the AI era,” with Cortex AI supporting text‑to‑SQL, summarization, sentiment analysis, and AI coding agents operating directly on governed data. This approach targets what Ramaswamy describes as the “agentic enterprise,” where AI systems coordinate workflows and business outcomes rather than only answer questions. To deliver that vision, Snowflake must guarantee that enterprise AI infrastructure will be available when customers scale workloads, which explains the willingness to commit billions to AWS capacity. The bet is that AI‑accelerated insights and automation will drive enough new usage to justify both the Natoma technology integration and the sizable AWS spend.