Elastic’s earnings this quarter should have been the headline. The San Francisco-based company beat Wall Street forecasts with 20% revenue growth to $415 million, lifted by a 24% jump in its cloud offering. Adjusted earnings rose 71% year-over-year to $0.60 per share, far above analyst expectations. Guidance for fiscal 2026 sales was raised to $1.684 billion. By the numbers, Elastic is executing.
But what dominated attention was the company’s CEO, Ashutosh Kulkarni, and his blunt dismissal of both rivals and narratives that cast Elastic as vulnerable. Asked whether Elastic might be better off inside a larger technology company, Kulkarni was unequivocal: “Those folks don’t know what the hell they’re talking about.”
The sharpest moment came when the conversation turned to Pinecone, a venture-backed company in the AI ecosystem. Pinecone pioneered the idea of a managed vector database and presented itself as a core building block for retrieval-augmented generation (RAG). The company raised $100 million in Series B funding at a $750 million valuation, built integrations with LangChain, Anyscale, and Mistral, and claimed over 20,000 organizations as customers. Today, it is reportedly exploring a sale.
Kulkarni’s verdict is that Pinecone was never a standalone business in the first place. “We’ve always said that vector databases are a feature. They are never going to be a business in and of themselves,” he said.
Elasticsearch to Search AI
Elastic is the “search AI company,” a label that Kulkarni has reinforced with stronger language. Founded around Elasticsearch, an open-source project launched more than a decade ago, the company became known for helping developers search and analyze unstructured data. Unlike structured relational data, unstructured information such as text, logs, and images does not fit neatly into rows and columns. Elastic built tools to make it usable.
“What we have always specialized in is unstructured data, messy data—that’s always been one of the hardest things to analyze and get value out of,” Kulkarni said. In his view, what Google did for the internet, Elastic did for enterprise data: turn noise into something searchable.
With large language models, this role has expanded. Generative AI depends on grounding outputs in reliable, relevant data. Kulkarni pointed to Elastic’s vector database, embedding models, re-ranking methods, and chunking strategies as part of the retrieval and context layer. “Fundamentally you need retrieval, you need some sort of data retrieval to ground these LLMs in the right context,” he said.
Snowflake, Databricks, and the Wrong Kind of Data
In the same interview, Kulkarni drew a deliberate contrast with the industry’s biggest names. Snowflake and Databricks have both made public pushes into AI search. Kulkarni praised them for their strength in structured data and BI workloads but insisted Elastic’s focus is distinct.
“If you have data that you know—you have a good schema for it, you understand what that schema looks like, there are tons of great technologies to put it in,” he said, pointing to Snowflake and Databricks as examples. “When you’re dealing with unstructured information, you need the kind of search index-based platform that Elasticsearch has always been about.”
Elastic is not competing for OLTP or BI workloads. Kulkarni calls its focus “context engineering.”
Pinecone’s Bet and Elastic’s Counter
Pinecone took a different path. Since its founding in 2019, it has promoted the vector database as a category of its own. Its managed platform let developers store high-dimensional vectors with embeddings of text, audio, or images and query them for similarity. In the generative AI boom, this became a common component of RAG workflows.
The company expanded rapidly. Its partner program included LangChain and Mistral. Internally, headcount tripled, and teams adopted Notion as a workspace. Pinecone’s sales team reported going from 50 to 400 weekly touch points after the shift. By 2023, Pinecone had become one of the recognized names in AI infrastructure.
Reports that the company is now exploring a sale highlight the limits of that model. Kulkarni has consistently argued that vectors alone cannot deliver the accuracy or context enterprises require. “If all you do is shove all your data, you vectorize them and think of vector search as a singular technique, you’re going to fail,” he said. Elastic’s approach is hybrid search: combining keyword indexes, vectors, and re-ranking to deliver more relevant results.
While Elastic has raised guidance on top of strong growth, Pinecone is considering strategic options despite name recognition and adoption.
Open Source as Strategy
Elastic’s approach is tied to its open-source foundation. Elasticsearch began as a community project, and the company has kept its core code publicly available. “I’m a huge fan of open source. I think why it sort of creates a democratization effect that is incredibly good for the industry,” Kulkarni said.
Elastic supports open-weight models like Llama and Mistral alongside closed models from hyperscalers, and allows deployment in both public cloud and private data centers. “If you provide choice, that becomes a big differentiator,” Kulkarni said.
He linked that choice to customer trust: “We earn your trust by delivering incremental value every month, every year, as opposed to creating lock-in because we have proprietary formats.” Kulkarni pointed to India as a market where multilingual, multimodal, and privacy-sensitive use cases make that flexibility important.
Elastic and Pinecone now represent different conclusions about AI infrastructure. Elastic emphasizes hybrid methods and open-source choice. Pinecone centered its business on vectors alone. One is raising its outlook after a quarter of solid growth. The other is reportedly weighing a sale despite widespread adoption. Kulkarni has chosen to make the distinction plain: “Vector databases are a feature. They are never going to be a business in and of themselves.”
