Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
For decades, the data landscape was relatively static. Relational databases (hello Oracle!) were the default and dominated, organizing information into familiar columns and rows.
This stability has eroded as successive waves have introduced NoSQL document stores, graph databases, and, more recently, vector systems. In the age of agentic AI, data infrastructure is once again evolving and evolving more rapidly than at any time in recent memory.
As we enter 2026, one lesson has become inescapable: data matters more than ever.
RAG is dead. Long live the RAG
Perhaps the most important trend for 2025 and one that will continue to be debated through 2026 (and perhaps beyond) is the role of RAG.
The problem is that the original RAG pipeline architecture looks a lot like basic search. Retrieval finds the result of a specific query, at a specific time. It’s also often limited to a single data source, or at least that’s how RAG pipelines have been built in the past (the past being before June 2025).
These limitations have led a growing number of conga sellers to claim that RAG is dying, on the verge of extinction, or already dead.
What is emerging, however, are alternative approaches (such as contextual memory), as well as nuanced and improved approaches to RAG. For example, Snowflake recently announced its agentic document analysis technology, which extends the traditional RAG data pipeline to enable the analysis of thousands of sources, without the need for structured data first. There are also many other RAG-like approaches emerging, including GraphRAG whose use and capacities will probably only increase in 2026.
So now, RAG is not (entirely) dead, at least not yet. Organizations will still find use cases in 2026 where data recovery is necessary and an enhanced version of RAG will likely still be suitable. In 2026, companies should evaluate use cases individually. Traditional RAG works for static knowledge retrieval, while enhanced approaches such as GraphRAG are suitable for complex and multi-source queries.
Contextual memory is table stakes for agentic AI
While RAG won’t completely disappear in 2026, one approach that will likely surpass it in terms of its use for agentic AI is contextual memory, also known as agentic or long context memory. This technology allows LLMs to store and access relevant information over extended periods of time.
Several systems of this type will emerge during 2025, including Hindsight, A-MEM framework, General agentic memory (GAM), LangMem and Memobase. RAG will remain useful for static data, but agentic memory is essential for adaptive assistants and agentic AI workflows that need to learn from feedback, maintain their state, and adapt over time.
In 2026, contextual memory will no longer be a new technique; this will become table stakes for many operational agentic AI deployments.
Use cases for purpose-built vector databases will change
At the start of the modern era of generative AI, purpose-built vector databases (like Pinecone and Milvus, among others) were all the rage.
For an LLM (usually but not exclusively through RAG) to access new information, it must access the data. The best way to do this is to encode the data as vectors, which is a numerical representation of what the data represents.
By 2025, what became painfully obvious was that vectors were no longer a specific database type but rather a specific data type that could be integrated into an existing multi-model database. So instead of requiring an organization to use a purpose-built system, they could simply use an existing database that supports vectors. For example, Oracle supports vectors, as do all databases offered by Google.
Oh, and it gets better. Amazon S3, long the de facto leader in cloud-based object storage, is now allows users to store vectorsthus eliminating the need for a single, dedicated vector database. This doesn’t mean that object storage replaces vector search engines (performance, indexing, and filtering are still important), but it narrows the set of use cases where specialized systems are required.
No, this doesn’t mean that purpose-built vector databases are dead. Just as with RAG, there will still be use cases for purpose-built vector databases in 2026. What will change is that the use cases will likely be somewhat restricted for organizations that need the highest levels of performance or specific optimization that a general-purpose solution does not support.
PostgreSQL bottom-up
As 2026 begins, what’s old becomes new again. The open source database PostgreSQL will be 40 years old in 2026, but it will be more relevant than ever.
During the year 2025, the supremacy of PostgreSQL as the must-have database for creating any type of GenAI solution became obvious. Snowflake spent $250 million to acquire Crunchy Data, the PostgreSQL database provider; Data bricks spent $1 billion on Neon; and Supabase raised a $100 million Series E, giving it a valuation of $5 billion.
All this money is a clear signal that businesses are defaulting to PostgreSQL. The reasons are many, including open source, flexibility and performance. For flavor coding (a primary use case for Supabase and Neon in particular), PostgreSQL is the standard.
Expect greater growth and adoption of PostgreSQL in 2026 as more organizations come to the same conclusions as Snowflake and Databricks.
Data researchers will continue to find new ways to solve already solved problems
It is likely that there will be more innovation to solve problems that many organizations already think are solved problems.
In 2025, we have seen many innovations, such as the idea that an AI is capable of analyzing data from an unstructured data source like a PDF. This is a capability that has existed for several years, but has proven more difficult to operationalize on a large scale than many expected. Databricks now has an advanced parser, and other vendors, including Mistral, have emerged with their own enhancements.
The same is true with natural language translation to SQL. Although some might have thought that this was a solved problem, it is a problem that continued to see innovation in 2025 and will see more in 2026.
It is essential that businesses remain vigilant in 2026. Don’t assume that fundamental features like parsing or natural language to SQL are fully resolved. Continue to evaluate new approaches that may significantly outperform existing tools.
Acquisitions, investments and consolidation will continue
2025 was an important year for significant investments in data providers.
Meta invested 14.3 billion dollars at data labeling provider Scale AI; IBM announced its intention to acquire data streaming provider Confluent for $11 billion; and Salesforce recovered Informatica for 8 billion dollars.
Organizations should expect the pace of acquisitions of all sizes to continue in 2026, as large vendors realize the fundamental importance of data to the success of agentic AI.
The impact of acquisitions and consolidations on businesses in 2026 is difficult to predict. This can lead to vendor lock-in, but also potentially to an expansion of platform capabilities.
In 2026, the question will not be whether companies are using AI, but rather whether their data systems are capable of supporting it. As agentic AI matures, durable data infrastructure (not intelligent prompts or ephemeral architectures) will determine which deployments evolve and which quietly shut down.
