From Experiments to Locked-In Enterprise AI Infrastructure
Enterprise AI infrastructure refers to the long-term hardware, cloud services, and data platforms that support large-scale AI applications across an organization, turning isolated pilots into durable, production systems. After years of experiments, major vendors and enterprises are now signing multibillion-dollar AI infrastructure contracts that bind workloads to specific clouds for many years. This shift is reshaping cloud AI adoption: instead of testing a few models, enterprises are committing to full stacks that combine data platforms, GPUs, agentic frameworks, and managed services. That means AI agents, data warehouses, and core business applications all share the same underlying cloud fabric. The result is a new kind of enterprise AI ecosystem where infrastructure decisions determine which providers power future agents, analytics, and line-of-business systems—and how hard it will be to switch later.
Snowflake and AWS Bet Big on Enterprise Agentic AI
Snowflake’s expanded strategic collaboration with AWS shows how cloud providers are using long-term commitments to anchor enterprise agentic AI. Snowflake has agreed to a USD 6 billion (approx. RM27.6 billion) multi-year infrastructure commitment to AWS, focused on Graviton compute and AI workloads. The deal builds on more than USD 7 billion (approx. RM32.2 billion) in lifetime AWS Marketplace sales and more than USD 2 billion (approx. RM9.2 billion) in 2025 sales through the marketplace, indicating that customers are already buying Snowflake’s data and AI capabilities at scale. Snowflake Cortex AI lets customers run text-to-SQL, summarization, and other AI applications directly on governed data, while AWS Graviton and GPU-accelerated instances provide the performance layer. Together, they are positioning agentic AI—agents that reason over data and coordinate workflows—as a default pattern for enterprise AI workloads, embedded within a unified data cloud.
Oracle’s USD 67 Billion AI Infrastructure Wave
Oracle’s latest results highlight how AI infrastructure contracts are scaling from hundreds of millions to tens of billions. The company reported USD 67 billion (approx. RM308.2 billion) in AI infrastructure contracts in its fourth fiscal quarter, as customers move to production-grade agentic solutions. According to Oracle, remaining performance obligations reached USD 638 billion (approx. RM2,935.0 billion), up 363% year over year, giving it clear visibility into future revenue tied to long-term deals. Four customers each signed contracts worth more than USD 8 billion (approx. RM36.8 billion) in that quarter alone. On the hardware side, Oracle delivered more than 1.2 gigawatts of data center capacity, with GPU utilization at 97.5%. Oracle links this demand to agentic coding and AI agents embedded in its applications, suggesting that as enterprise AI becomes embedded in day-to-day operations, infrastructure demand becomes both durable and harder to unwind.
Bundled Data, AI, and Agents: The New Enterprise Stack
Both Snowflake and Oracle show how cloud providers are bundling AI capabilities with data platforms and applications to create integrated enterprise AI ecosystems. Snowflake’s approach centers on bringing foundation models to where enterprise data already lives, so AI applications run inside its governed data environment rather than on separate systems. Oracle is building an application-led stack: it reports more than 1,000 AI agents across its suites, outcome-based pricing models, and token bundles that connect customers to advanced reasoning models. These bundles link data, models, and business processes tightly to a provider’s infrastructure. For enterprises, the appeal is clear: fewer integration points, faster deployment, and consistent governance. But the trade-off is that once agents, data pipelines, and workload migrations are all optimized for a single cloud, the practical cost of switching providers grows sharply over time.
Lock-In Risk and the Future of Cloud AI Adoption
Long-term AI infrastructure contracts are reshaping cloud AI adoption by turning cloud providers into de facto operating systems for enterprise AI. Multi-year commitments like Snowflake’s USD 6 billion (approx. RM27.6 billion) AWS deal and Oracle’s USD 67 billion (approx. RM308.2 billion) in AI infrastructure contracts extend far beyond traditional licensing. They include regional expansions, marketplace procurement channels, and customer success programs that encourage enterprises to move all critical data and agentic workloads into a single environment. This strategy reduces switching risk for providers, but increases dependency for customers, who must weigh predictable capacity and integrated tools against future flexibility. As vendors add more agentic AI, outcome-based pricing, and multi-cloud database options, the next competitive edge may come from offering strong enterprise AI infrastructure with clearer exit paths—so organizations can commit at scale without closing off future choices.
