From Model Hype to Enterprise Data Foundations
Enterprise AI infrastructure spending is shifting from headline-grabbing models toward data platforms because companies are learning that real AI value depends on clean, connected, and governed data that can reliably feed many different models and applications across the business. Databricks’ recent funding round, totaling more than USD 7 billion (approx. RM32.2 billion) and including USD 5 billion (approx. RM23.0 billion) in equity at a USD 134 billion (approx. RM615.2 billion) valuation, is a clear signal of this new priority. According to Inc., this followed a USD 1 billion (approx. RM4.6 billion) raise at a valuation above USD 100 billion (approx. RM459.3 billion) only months earlier, underscoring investor conviction that data platform investment is becoming central to enterprise AI strategies. While OpenAI and others dominate the model conversation, the real competitive battleground is shifting to the platforms that prepare and operationalize data for AI model deployment.
Databricks, Snowflake and the New Enterprise AI Infrastructure Layer
Databricks and Snowflake are fighting to own the layer where enterprise data is prepared, governed, and exposed for AI model deployment and analytics. Databricks grew from Apache Spark roots into its Data Lakehouse concept, unifying structured and unstructured data so enterprises can use text, images, documents, and transactional records in the same environment. As enterprises move beyond experimentation with tools like ChatGPT, they want AI systems that plug into finance, supply chain, HR, customer operations, and industry workflows. That requires platforms that can combine ERP data with email threads, PDFs, and logs in a governed way. Snowflake, originally focused on structured cloud data, is expanding into open-source databases, unstructured information, and AI tools to stay relevant. The rivalry shows that enterprise AI infrastructure is no longer about storage alone; it is about building the default place where operational data becomes AI-ready.
Why Data Platform Investment Now Outweighs Model Building
Enterprises are realizing that models, even cutting-edge generative ones, cannot deliver sustainable value without solid data foundations. Most organizational information is unstructured, and turning that content into usable signals for AI requires data engineering, governance, and consistent access controls. Databricks’ rise displays how much the market values this layer: the company disclosed an annual revenue run rate of USD 5.4 billion (approx. RM24.8 billion), with fourth-quarter revenue growing more than 65% year over year. Models can often be sourced from external providers, but the data platform where corporate information is cleaned, cataloged, and monitored is harder to replace. This is pushing enterprise data spending toward platforms that can serve many models at once, not only one-off generative AI experiments. In effect, the durable asset is the data fabric; models are becoming plug-ins that sit on top of it.
Acquisitions, Agentic AI and the Push Toward Production
Both Databricks and Snowflake are buying AI and data startups to turn their platforms into full enterprise AI infrastructure stacks. Databricks acquired MosaicML for about USD 1.3 billion (approx. RM6.0 billion) to add generative AI model training and customization, and later bought Arcion, Neon, Mooncake Labs, and Tabular to connect high-speed transaction data and cross-database interoperability. Snowflake purchased Neeva and brought its CEO into the company while also partnering with large language model vendors. These moves show the battle is not only about where data sits, but how it powers agents, analytics, and automation. Agentic AI, where autonomous systems act on business processes, depends on timely, well-governed data. As Databricks’ CTO of Neural Networks Hanlin Tang notes, companies are already seeing real value from such systems, which means enterprises are investing in production-grade platforms rather than short-lived model pilots.
Beyond the ChatGPT Era: Building Durable AI Capabilities
The rush that followed ChatGPT’s release pushed many enterprises to experiment with generative models before they had solid data foundations. That phase is ending. Organizations now understand that AI model deployment at scale requires a shared platform where structured ERP records, semi-structured logs, and unstructured content can be integrated and governed once, then reused across many AI services. Databricks’ and Snowflake’s strategies highlight this pivot: they are embedding model customization, security, and transaction-level access into their core platforms, making them central to long-term AI roadmaps. For CIOs and ERP leaders, the lesson is that the most important AI spending decision is not which foundation model to adopt, but which data platform will anchor their enterprise AI infrastructure. Models will change frequently; the data platform is the asset that must endure and keep delivering return on enterprise data spending.
