Agentic AI Is Pushing Enterprise AI Procurement Toward Cloud Platforms

Google's latest State of AI Infrastructure report surveyed more than 1,400 senior IT leaders and found that 83% of organizations need infrastructure upgrades to support production-scale agentic AI. While that finding has drawn significant attention, another trend in the report highlights how enterprise AI procurement is evolving. According to the survey, 78% of organizations now source their generative AI solutions from their primary cloud provider, up 30 percentage points from 2025. The increase suggests enterprises are increasingly adopting AI capabilities through the cloud platforms that already host their infrastructure, rather than assembling separate AI tools from multiple vendors. Agentic AI Is Changing Enterprise Requirements During the first wave of enterprise generative AI adoption, organizations often evaluated foundation models, retrieval systems, orchestration platforms and governance tools independently as they experimented with AI. Agentic AI introduces additional operational requirements. Unlike chatbots, AI agents interact with enterprise applications, access business data, execute workflows and coordinate tasks across systems with greater autonomy. Google's report found that 79% of technology leaders identified security , governance and MLOps as the biggest challenges in scaling AI inference. As enterprises deploy larger numbers of AI agents, they also need centralised visibility into agent identities, permissions and audit trails. Managing those controls across multiple AI platforms can add operational complexity, particularly in regulated industries. Cost and Operations Are Also Influencing Procurement The report also highlights infrastructure costs associated with production AI. According to Google, 62% of organizations experience an " inference tax ", driven by factors such as data egress fees, storage growth and underutilised specialised hardware. Many also cited operational complexity as a hidden cost of scaling AI. Running AI workloads alongside existing compute, storage and enterprise data on the same cloud platform can reduce integration overhead and simplify operations. Regulatory requirements are also shaping infrastructure decisions. The survey found that 52% of firms operate hybrid multicloud environments, while 48% prioritise infrastructure that supports data residency requirements. These requirements are encouraging enterprises to adopt platforms that support both public cloud and on-premises deployments while meeting compliance obligations. Cloud Providers Are Expanding AI Platform Capabilities The procurement trend also aligns with how major cloud providers are evolving their enterprise AI offerings . Rather than competing solely on access to foundation models, vendors are expanding capabilities around identity, governance, security and compliance. One common trend is treating AI agents as distinct digital identities instead of extensions of human users. Microsoft introduced Entra Agent ID to manage agent identities within its ecosystem. Amazon Web Services launched AgentCore Identity, which authenticates AI agents as workload identities. Google similarly assigns agents unique cryptographic identities through its Gemini Enterprise Agent Platform. Governance capabilities are also moving into the platform layer. AWS provides AgentCore Policy for applying security policies to AI agents. Google is developing Agent Gateway, currently in preview, to govern interactions between AI agents, enterprise applications and external tools. Salesforce has introduced Agent Fabric to provide identity and policy management across Agentforce and third-party AI agents. Compliance capabilities are also being integrated into enterprise AI platforms. Microsoft has expanded Purview to audit Copilot interactions and enterprise data access. Oracle has embedded AI agents within Fusion Cloud Applications, allowing them to inherit existing role-based access controls, approval workflows and audit processes. Oracle's AI management framework has also achieved ISO/IEC 42001 certification for AI governance. Another emerging trend is support for multi-vendor AI environments. Google allows organizations to govern third-party agents from Oracle, Salesforce and ServiceNow using the same identity and audit framework as Gemini agents. Salesforce has taken a similar approach by extending governance across Agentforce, Data Cloud, MuleSoft and Informatica. These developments indicate that cloud providers are increasingly competing on the broader enterprise AI platform rather than on foundation models alone. Enterprise AI Buying Decisions Are Broadening Google's survey suggests that enterprise AI procurement is increasingly being evaluated as a platform decision rather than a model-selection exercise. Alongside foundation model performance, organizations are also considering factors such as identity management, governance, compliance, infrastructure integration and deployment across hybrid environments. As enterprise

Agentic AI Is Pushing Enterprise AI Procurement Toward Cloud Platforms

Agentic AI Is Changing Enterprise Requirements

Keep reading. It's free.

Related stories