Snowflake and Amazon Web Services this week unveiled a multi-year agreement that includes a $6 billion Snowflake commitment for AWS compute and AI infrastructure over five years. Snowflake said the spending will lean heavily on AWS Graviton processors alongside GPU infrastructure for AI training and inference \u2013 an allocation that spotlights how enterprise AI is shifting from sporadic pilots to persistent, production-grade workloads.. Strong Quarter. The infrastructure push arrived alongside one of Snowflake\u2019s strongest quarters in years. The company reported first-quarter fiscal 2027 revenue of $1.39 billion, up 26% year over year, while product revenue climbed 26% to $996.8 million. Snowflake also raised its fiscal 2027 product revenue forecast to $4.325 billion. Shares jumped following the earnings release and AWS announcement as investors responded to rising AI-driven consumption and expanding enterprise workloads.. Related:AWS Networking Boss on Hollow-Core Fiber, AI, and Data Center Innovations. From Pilots to Persistent Loads. Over the past two years, many enterprise AI deployments lived inside pilots, copilots, and internal experiments \u2013 short GPU bursts tied to proofs of concept. That profile is changing. Enterprises are now implementing agentic AI \u2013 systems that continuously query operational data, coordinate workflows, trigger actions, and persist across business processes rather than rely on isolated prompts. In practical terms, this turns AI into steadier infrastructure tenant within hyperscale environments.. Persistent inference workloads consume infrastructure differently. They stretch GPU utilization, increase east-west traffic, expand memory requirements, tighten storage coupling, and create continuous orchestration activity. AI infrastructure is starting to look less like intermittent experimentation and more like core operational capacity.. Dave McCarthy, research vice president for cloud and edge infrastructure services at IDC, said Snowflake\u2019s $6 billion AWS commitment reflects how quickly enterprise AI workloads are settling into long-duration infrastructure demand.. \u201cSnowflake\u2019s $6 billion commitment to AWS is a clear signal that enterprise AI is crossing the chasm from experimental science projects to persistent, foundational infrastructure,\u201d McCarthy said. \u201cWe are moving away from ad-hoc queries and isolated proofs-of-concept toward an era of \u2018always-on\u2019 data gravity.\u201d. McCarthy added that Snowflake\u2019s emphasis on AWS Graviton processors reflects growing pressure to control inference economics as enterprise AI deployments expand.. \u201cBy anchoring this massive spend around custom ARM-based silicon like Amazon Graviton, Snowflake is acknowledging that the long-term viability of enterprise AI relies completely on structural cost-efficiency and predictable, sustained compute economics at scale,\u201d he said.. Related:Quantum Progress Runs Through the Data Center \u2013 AWS Shows Why. Anay Nawathe, director at ISG, said the agreement points toward a broader transition from experimental AI deployments to longer-duration infrastructure demand. However, he cautioned against treating the full $6 billion commitment as pure inference consumption.. \u201cYes, this deal directionally reflects a shift towards always-on, long-duration hyperscale infrastructure demand,\u201d Nawathe said. \u201cThe deal size alone allows AWS to make the investments to support long-term AI capacity.\u201d. At the same time, Nawathe noted, \u201cthis deal is evidence of long-duration capacity, but not actual consumption at enterprises without further detail.\u201d He also said operationalizing agentic AI is placing growing pressure on governance, semantic layers, and infrastructure observability.. \u201cThere needs to be clear visibility into the data sources, access to the data, and the context behind the data,\u201d Nawathe said. \u201cThe infrastructure layer is also becoming more heterogeneous to allow workloads to run where they perform best for their purpose. Observability now becomes critical.\u201d. Related:Amazon, Anthropic Strengthen AI Ties With $5B Investment. The Control Layer Battle. Roy Illsley, chief analyst at Omdia, said growing enterprise demand for AI inference does not necessarily mean all workloads will land inside hyperscale cloud environments.. \u201cThis is the current big debate, as AI inferencing is definitely the next big growth market, but where this will be run is the debate,\u201d Illsley said.. Illsley said enterprises increasingly want to bring AI closer to existing data rather than move large datasets into separate AI environments, a trend that benefits both cloud providers and on-premises infrastructure vendors.. \u201cOne thing that seems to be resonating with enterprise customers is that they should bring AI to their data, not move their data to the AI,\u201d he said.. As a result, Illsley expects inference workloads to spread across cloud, hosted, telecom, and on-premises environments depending on customer economics, availability requirements, and risk-management strategies.. Snowflake CEO Sridhar Ramaswamy framed the partnership around what the company calls the \u201cagentic enterprise,\u201d where AI systems move beyond answering questions and begin operating directly against governed enterprise data and workflows.. \u201cAI has generated enormous excitement, but for enterprises, the real challenge and opportunity is turning intelligence into action,\u201d Ramaswamy said in a statement.. The partnership intensifies the fight over where enterprise AI systems actually operate. Enterprises increasingly want AI systems positioned alongside governed operational data rather than having to shuttle sensitive information between disconnected model platforms, orchestration layers, and storage environments. Vendors now race to collapse governance, analytics, orchestration, and AI execution into tightly integrated operational stacks.. Snowflake wants to become that operational layer. The company said Snowflake Cortex AI enables enterprises to run workloads, including text-to-SQL, summarization, sentiment analysis, and entity extraction, directly within Snowflake environments while maintaining governance and security controls over underlying datasets.. The AWS expansion also sharpens Snowflake\u2019s competition with Databricks, hyperscaler-native AI stacks, and a growing ecosystem of enterprise AI platform vendors competing to control where enterprise agents run and where inference workloads accumulate.. Inference Economics. Snowflake said it will continue scaling use of AWS Graviton processors alongside GPU-accelerated EC2 instances. That mix reflects growing pressure to optimize inference efficiency and to spread AI workloads across CPU and GPU resources rather than relying exclusively on high-cost accelerated compute.. Training built the first wave of AI infrastructure; inference increasingly drives the second.. The companies also expanded joint AWS Marketplace initiatives. Snowflake said it has surpassed $7 billion in lifetime AWS Marketplace sales, including more than $2 billion during calendar year 2025.. The company also highlighted continued AWS regional expansion, including deployments underway in New Zealand, South Africa, and Thailand, as well as the AWS European Sovereign Cloud initiative, as enterprises push for local data residency and sovereign AI controls.. Customers highlighted in the announcement included Fetch and Hex, both of which use Snowflake on AWS to deploy AI applications tied directly to governed enterprise data.. For AWS, the partnership reinforces how deeply enterprise AI workloads now shape long-term cloud growth.. For the broader data center industry, the more consequential signal sits lower in the stack: enterprise AI is beginning to settle into continuous operational demand instead of temporary experimental spikes.<\/p>\n
\u00a0<\/p>","protected":false},"excerpt":{"rendered":"
Snowflake and Amazon Web Services this week unveiled a multi-year agreement that includes a $6 billion Snowflake commitment for AWS…<\/p>\n","protected":false},"author":2,"featured_media":335,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11],"tags":[],"class_list":["post-1840","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-data-center"],"_links":{"self":[{"href":"https:\/\/trustedainews.com\/index.php?rest_route=\/wp\/v2\/posts\/1840","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/trustedainews.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/trustedainews.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/trustedainews.com\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/trustedainews.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=1840"}],"version-history":[{"count":0,"href":"https:\/\/trustedainews.com\/index.php?rest_route=\/wp\/v2\/posts\/1840\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/trustedainews.com\/index.php?rest_route=\/wp\/v2\/media\/335"}],"wp:attachment":[{"href":"https:\/\/trustedainews.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1840"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/trustedainews.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=1840"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/trustedainews.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=1840"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}