The main battleground for the control plane that determines enterprise AI adoption is shifting. Centered on Palantir's ontology, Snowflake, Databricks, AWS and Microsoft are each advancing their own strategies.
The focus of competition has shifted from model intelligence to the control plane that connects AI to enterprise operations. At the center of the contest is who will control the infrastructure that integrates three elements: where data is consolidated, how work is carried out and who has the necessary permissions, enabling AI agents to run real-world operations.
Palantir is emphasizing a system built around its long-developed ontology that maps data, permissions and executable business processes in one place. Snowflake, meanwhile, is strengthening the incorporation of context from the data platform, while Databricks has unveiled Genie Ontology, which learns business context. The three companies are targeting the control plane from different starting points.
Cloud companies are also stepping up their entry. AWS has organized forward-deployed engineering teams that work on-site with customers, while Microsoft has established Frontier Company to strengthen its implementation capabilities. Anthropic and OpenAI are also expanding their implementation teams, making the competition to support deployment, rather than merely provide models, increasingly clear.
NVIDIA is not directly joining the contest for control, instead supplying components for agent infrastructure and the underlying compute platform. In Japan, industry-specific AI development using Nemotron is advancing, strengthening the trend toward building dedicated in-house platforms. The three key areas of competition are deep understanding of business operations, the ability to drive a broad range of activities across the map, and rapid adoption.