From Models to Middleware: Redefining Enterprise AI Priorities
Enterprise AI data infrastructure is the set of platforms and services that collect, store, process, govern, and deliver data so that AI models and agents can use it reliably inside business systems at scale. The latest signals from ClickHouse and Databricks show this layer overtaking raw model and compute spending in strategic importance. While model vendors dominate headlines, these AI data platforms sit between foundation models and day‑to‑day operations, turning scattered records, logs, and documents into usable context. Their value lies in making AI model utilization effective: connecting transactional systems, analytics, and real-time streams into pipelines that agents can act on. As enterprises move past early pilots toward production workloads, budgets are shifting toward this middleware, where data quality, latency, and governance now decide whether AI projects succeed or stall.
ClickHouse Revenue Growth Signals an AI Agent Data Stampede
ClickHouse’s latest numbers put a spotlight on how AI agent demand is reshaping data architecture choices. The company has passed USD 250 million (approx. RM1,150 million) in annualised revenue run rate, tripling year over year, and its leadership expects that figure to reach the high nine digits by the end of the year. That pace of ClickHouse revenue growth, backed by a USD 15 billion (approx. RM69,000 million) valuation from its recent Series D, reflects rising demand for databases tuned to analytical and AI workloads rather than generic storage. More than 4,000 customers, including Anthropic, Meta, and Capital One, use its open-source engine and managed cloud services to power high-volume queries and real-time analytics that AI agents depend on. Acquisitions such as Langfuse, which tracks and evaluates AI agent performance, further position ClickHouse as core enterprise data infrastructure for agentic systems.
Databricks Valuation Shows Data Platforms Now Anchor AI Strategy
If ClickHouse reflects demand, the Databricks valuation shows where long-term capital is flowing in enterprise AI. Databricks recently closed more than USD 7 billion (approx. RM32,200 million) in new funding, including USD 5 billion (approx. RM23,000 million) in equity at a USD 134 billion (approx. RM615,000 million) valuation, only months after a USD 1 billion (approx. RM4,600 million) raise above USD 100 billion (approx. RM460,000 million). According to Inc., Databricks’ annual revenue run rate reached USD 5.4 billion (approx. RM24,800 million), with fourth-quarter revenue growing more than 65% year over year. That growth at scale is unusual and underlines investor belief that AI data platforms are now the strategic control point for enterprise AI. By combining its Spark heritage with a "Data Lakehouse" that spans structured and unstructured data, Databricks has become the place where raw corporate data is transformed into material that models, analytics, and agents can use.
The Middleware Layer: Where AI Model Utilization Is Won or Lost
Taken together, ClickHouse and Databricks highlight a common pattern: enterprises are no longer buying AI as isolated tools; they are buying platforms that raise AI model utilization. Databricks and Snowflake are fighting to control the layer that prepares, governs, and connects data with models and applications, while ClickHouse focuses on high-speed analytical workloads that power agents and real-time decisioning. This is classic middleware: neither raw models nor end-user apps, but the enterprise data infrastructure that makes both useful. The acquisitions race—MosaicML and Tabular for Databricks, Langfuse for ClickHouse—shows data platforms expanding into model customization, observability, and transactional access. For CIOs and data leaders, the competitive landscape now centers on which AI data platforms can handle complex pipelines, combine structured ERP records with documents and messages, and support low-latency, scaled workloads without locking AI projects into a single model vendor.
AI Agents Push Real-Time, End‑to‑End Data Architectures to the Fore
AI agents are accelerating this shift from compute-centric to data-centric investment. Agents need continuous access to clean, current data across many systems—customer histories, transaction logs, operational metrics, unstructured documents—to act reliably. Platforms like ClickHouse, with columnar storage optimized for fast analytical queries, and Databricks, with its Lakehouse spanning multiple data types, are emerging as the backbone for these workloads. They can support multi-hop pipelines where data is ingested, refined, and fed into models, then logged and evaluated in near real time. As a result, AI adoption is being decided less by who has the most advanced model and more by who can stand up an architecture where agents can read, write, and reason over enterprise data safely. The real winner in the enterprise AI competition is the data layer that keeps agents supplied with trustworthy, timely context.