4 trends that shaped data management, analytics in 2025

MCP support is a must

One of the hindrances to some agentic AI initiatives is the ability to securely and easily connect agents with external data sources.

It's the way [Open Database Connectivity] was maybe 20 to 30 years ago. If you didn't support ODBC, you were less likely to have tools that worked with your technology. We've reached that point with MCP as well, that if you're a data management vendor, you've got to support MCP.

AI pilots, despite the ongoing efforts of data management and analytics vendors to simplify development, fail at an overwhelming rate. Depending on the source, it's estimated that up to 80% of all AI projects never reach production.

Common barriers include a lack of proper talent to build and manage AI models and applications, outdated technology that can't meet the performance requirements of AI systems and the high cost of building cutting-edge tools.

Data issues, however, are perhaps the primary problems preventing successful AI development. And one of the major ones is being able to securely connect agents and other AI applications with proper data sources, both internal proprietary data sources and external ones such as LLMs.

Not only is combining proprietary data with external sources risky from a security standpoint -- whenever data is moved between environments, there is risk of exposure -- but it is a complex and time-consuming undertaking. Doing it each time a new agent is developed can be prohibitive.

MCP is not a panacea, but it does virtually eliminate the issues related to connecting agents with data sources.

Launched by AI developer Anthropic in November 2024, MCP is a set of open source code that standardizes how enterprises build MCP servers that connect AI models with databases, data warehouses, data lakes and data lakehouses. When data management and analytics vendors provide support for MCP, they enable customers to use the prebuilt code to easily make connections.

As a result, as agentic AI development gained momentum in 2025, adding MCP support became a significant trend among data management and analytics providers. By the end of the year, it became a must for data management vendors, in particular, or they risked putting themselves at a competitive disadvantage, according to David Menninger, an analyst at ISG Software Research.

"We've reached that point," he said in September. "It's the way [Open Database Connectivity] was maybe 20 to 30 years ago. If you didn't support ODBC, you were less likely to have tools that worked with your technology. We've reached that point with MCP as well, that if you're a data management vendor, you've got to support MCP."

As many vendors added agentic AI development frameworks -- Databricks, Snowflake and Oracle among them -- MCP support was included. However, not all data management providers have created a full development environment, so they offer MCP support independently.

"It's a table-stakes functionality," said Sumeet Agrawal, vice president of product management at Informatica, in September. "It's like five years ago if you asked if API support was important. The answer is yes, it's fundamental. I believe that MCP is extremely fundamental for any data management provider, and if they don't do that, they will fall behind."

Heightened interest in semantic modeling

While simply and securely connecting agents to the data sources that feed them is one hindrance to successfully developing agentic AI applications, another is the data itself.

Inconsistencies and inaccuracies in datasets stall AI projects. Ensuring that data is properly prepared to inform AI tools is, therefore, a critical element of any AI pilot. 

One means of doing so is semantic modeling.

Semantic models are sets of common definitions that help govern data. By standardizing how metadata is codified when data is ingested from its source or transformed, data becomes consistent and discoverable.

Google Cloud, through the  Looker platform it acquired in 2019, is one vendor that has long provided semantic modeling capabilities. Strategy -- formerly known as MicroStrategy before it began placing as much emphasis on bitcoin investment as it does analytics -- is another. Meanwhile, specialists such as DBT Labs and Cube have made semantic modeling their focus.

Until recently, they were among a minority of data management and analytics vendors.

AI tools require massive amounts of high-quality data to properly perform. With AI the main development focus for many enterprises, semantic modeling is a rising trend in data management and analytics.

"You can't orchestrate AI agents that access different data sources without metadata, so it's increasingly important," said Rita Sallam, an analyst at Gartner, in June. "Now, metadata gets its new day in the sun, a new prominence, as a key to success."

Kevin Petrie, an analyst at BARC U.S., similarly noted the significant role semantic layers can play in AI development.

"A unified semantic layer can help overcome [data quality problems], enabling AI applications to consume diverse inputs to generate rich outputs," he said in September.

To date, one of the barriers to more widespread use of semantic layers has been the proprietary nature of each vendor's semantic modeling capabilities. Because they differ from one another, if an enterprise uses more than one platform for its data management or analytics needs, its data becomes fragmented.

To try to standardize semantic modeling, the Open Semantic Interchange -- a consortium of data management and analytics vendors including Snowflake, Salesforce and ThoughtSpot -- formed in September. Its aim, just as Anthropic's MCP provides an open standard for connecting agents with data sources, is to develop an open standard for defining data, according to Josh Klahr, director of analytics product management at Snowflake.

"Every company has struggled with fragmented, inconsistent semantic definitions for years," he said. "But until now, the pain was largely hidden inside BI tools and analytics teams. What's changed is the explosive demand for AI and agentic analytics. Suddenly, those inconsistencies aren't just slowing down dashboards. They're undermining the accuracy and trustworthiness of AI systems."

Informa TechTarget

Purchasing PostgreSQL capabilities

The growing interest in semantic layers among data management and analytics vendors over the past year has been mirrored by a rising trend in acquiring PostgreSQL databases to support AI development.

Databricks was the first of three vendors to acquire PostgreSQL database capabilities, purchasing Neon in May. Archrival Snowflake quickly followed in June with the acquisition of Crunchy Data. Finally, in October, Redpanda purchased Oxla to add PostgreSQL database capabilities.

PostgreSQL databases are built on the open source PostgreSQL format, which is designed to enable users to securely and efficiently store and prepare data for analytics and AI applications.

Key tenets of PostgreSQL databases are flexibility and versatility, which have made them the most popular database format, according to the 2024 Developer Survey by Stack Overflow; they now outpace the open source MySQL format and databases from Microsoft, MongoDB and Redis.

In addition to traditional relational database capabilities for handling operational and transactional workflows, PostgreSQL databases can also manage geospatial, time series, JSON and vector database workloads.

"As organizations build AI solutions, they need a database that can handle transactional data at scale," said Stephen Catanzano, an analyst at Omdia, a division of InformTechTarget,  in June. "Postgres fits this role perfectly."

Meanwhile, the decision to buy PostgreSQL database capabilities rather than build them internally stems from how difficult it is to develop them, according to Tyler Akidau, Redpanda's chief technology officer.

"Anyone who has ever tried to build a SQL engine will tell you it's a massive undertaking," he said when the vendor acquired Oxla. "This acquisition gives us years of advantage over building it ourselves."

Regarding the impetus for adding PostgreSQL database capabilities, Akidau cited the development of agentic AI.

"It's the move that makes its real-time data accessible to both humans and agents through familiar, governed queries. … Redpanda can [now support] multimodal data access for agentic reasoning," he said.

Consolidation rises as funding falls

Many enterprises are concerned about vendor lock-in, so they avoid using a single provider for their data management and analytics needs. Others prefer using the best tools for tasks such as data integration, data preparation and data governance, so they similarly use tools from various vendors to build their data stacks.

But the cost of AI development, which is significantly higher than the cost of building traditional data products, such as reports and dashboards, may be changing that. Instead, to save the expense of integrating disparate tools and maintaining complex systems, more organizations may opt for a single, end-to-end provider.

Meanwhile, seizing on an opportunity, hyperscale cloud providers such as AWS, Google Cloud and Microsoft, along with data platform providers including Databricks and Snowflake, are adding more managed services to cover all aspects of data management and analytics.

As a result, specialists are suddenly finding it more difficult to remain independent.

"Independent vendors are getting squeezed from two directions," said Sanjeev Mohan, founder and principal of analyst firm SanjMo, in early December. "Staying independent requires massive R&D investment to keep pace with both threats simultaneously."

Funding was abundant throughout the 2010s and into the early 2020s. Then it stopped in mid-2022. Throughout the second half of the year and all of 2023, only Databricks and a few others were able to attract much funding.

In 2024, while funding didn't reach pre-2022 levels, vendors such as Cribl, Aerospike and Sigma each raised more than $100 million. Others, including Ocient and Coalesce, attracted about $50 million.

In 2025, funding again disappeared and, instead, consolidation became a trend among data management and analytics vendors.

In addition to established vendors acquiring startups to add new capabilities, longtime data integration specialist Informatica was acquired by Salesforce for $8 billion, data streaming vendor Confluent was purchased for $11 billion by IBM, and data integration specialist Fivetran merged with data preparation vendor DBT Labs.

"What's happening is that the big companies feel like they need to own every component now, from data ingestion all the way through AI, analytics and streaming data," Catanzano said when Confluent was bought on Dec. 8. "Before, it was, 'We're good with an ecosystem,' and now we're seeing platform plays. You can see the pattern."

Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than 25 years of experience. He covers analytics and data management.