Snowflake to Invest $6 Billion in AWS Graviton and AI Infrastructure

Cloud data warehousing giant Snowflake has announced plans to invest a total of $6 billion (approximately 870 billion yen) over the next five years in Amazon Web Services (AWS) custom-designed Graviton CPUs and AI accelerators. As the competition for cloud investments in the era of AI heats up, this move marks a new chapter in the race for data infrastructure dominance.

Strengthening the Long-Standing Partnership

Between Snowflake and AWS The relationship between Snowflake and AWS dates back to the founding of Snowflake in 2011. Since its inception, the company has built its services on AWS infrastructure, and this massive investment aims to deepen that relationship further. Snowflake has already achieved cumulative sales of over $7 billion on the AWS Marketplace, with more than $2 billion in revenue generated during the 2025 calendar year alone. Building on this solid business foundation, Snowflake has decided to ramp up its investments in the AI domain. Snowflake CEO Frank Slootman highlighted the significance of this investment, stating that the company is “enabling enterprises to bring AI directly to governed data, act faster, operate with greater density, and create measurable impact at scale.”

Graviton Adoption Highlights the Resurgence

of CPUs One noteworthy aspect of Snowflake’s strategy is its shift from traditional Intel and AMD CPUs to ARM-based AWS Graviton instances for its computing resources. The latest fifth-generation Graviton processor features 192 ARM Neoverse V3 cores, a 12-channel memory subsystem with an 8800 MT/s memory bandwidth, and offers one of the highest performance levels among server processors built on ARM architecture. In recent years, discussions around AI have primarily centered on GPUs and other AI accelerators. However, CPUs are reclaiming their importance in the actual operation of AI agents. While AI model inference is typically executed on GPUs, the execution of tools and functions such as SQL queries or Python scripts called by these models still heavily relies on CPUs. The performance of individual AI agents is intrinsically limited by how quickly CPUs can respond to requests. As the adoption of agent-based AI grows, the demand for CPU cores is resurging for fundamental structural reasons.

Cortex AI:

Shaping the Future of Data Analytics A key focus of Snowflake’s investment is the enhancement of its Cortex AI platform. Cortex AI provides features like converting natural language into SQL queries, summarizing data, and performing sentiment analysis. For example, users can ask natural language questions such as “What were the top five selling products last month?” Cortex AI then converts the query into SQL, retrieves and analyzes the data from the data warehouse, and delivers the answer. This approach brings advanced data analytics capabilities directly to business users, eliminating the need for data analysts or engineers to be involved in such tasks. As part of the agreement, Snowflake plans to use a combination of GPUs and Graviton CPU cores running on AWS to execute and train generative AI models and services. The envisioned architecture assigns GPUs to handle model training and inference, while Graviton CPUs support data processing and query execution.

Is a $1.2 Billion Annual Investment a Risky

Bet? Snowflake’s planned annual investment of approximately $1.2 billion is a significant financial commitment. How does the company justify it? Snowflake is betting that its AI tools will continue to drive revenue and generate returns that justify the infrastructure spending. Competition in the data warehousing market is intensifying, with rivals like Databricks and Google BigQuery also racing to enhance their AI capabilities. For Snowflake, investing in AI infrastructure is an essential strategic move to maintain a competitive edge. Wall Street’s reaction has been broadly positive. On the day of the announcement, Snowflake’s stock price surged by over 30% in after-hours trading, signaling that investors view this large-scale infrastructure investment as a positive step toward future growth.

Graviton Adoption Extends Beyond Snowflake

The expansion of AWS’s Graviton ecosystem isn’t limited to Snowflake. Meta has also revealed plans to deploy tens of millions of Amazon Graviton 5 CPU cores by April 2026. This multi-year collaboration positions Meta as one of the largest users of AWS’s Graviton platform. ARM-based server CPUs are considered more energy-efficient than traditional x86 processors, contributing to significant cost savings in operating large-scale data centers. As AI workloads continue to surge, increasing power consumption, ARM-based options like Graviton are becoming more attractive to cloud providers from a cost-efficiency standpoint.

Future Outlook Snowflake’s $6 billion

investment is not merely about expanding infrastructure; it signifies a broader vision. In the era of AI agents, data warehouses are evolving beyond mere repositories into operational foundations for AI services. At the core of this transformation lies the synergy between CPUs and AI accelerators. By integrating AI intelligence directly with governed enterprise data, Snowflake is envisioning a future where data analysis becomes more democratized, fundamentally transforming corporate decision-making processes. However, it remains to be seen how this massive upfront investment aligns with the company’s revenue timeline and long-term profitability. ---