OpenSearch in 2025: Much more than an Elasticsearch fork

Open source has never been more popular. It’s also never been more contentious.

With hundreds of billions of dollars available to the companies that best turn open source into easy-to-use cloud services, vendors have fiercely competed for enterprise dollars. This has led to a spate of licensing changes from the companies that develop the open source software, and forks from the clouds that want to package and sell it. But something interesting is happening: These forks may start as clone wars, but they’re increasingly innovative projects in their own right. I’ve recently written about OpenTofu hitting its stride against Terraform, but OpenSearch, which has its big community event this week in Amsterdam, is an even bigger success story.  

Born from the fire of Elastic’s 2021 license change, OpenSearch spent its first few years stabilizing and proving it could (and should) continue to exist. In the past year, OpenSearch has actively forged its own identity as a truly independent and innovative force in enterprise search, one that is quickly evolving to be much more than an Elasticsearch look-alike.

Moving beyond the fork

To understand OpenSearch’s recent path, a quick rewind is essential. In early 2021, Elastic ditched the Apache License (ALv2) for new Elasticsearch and Kibana versions, opting for the Server Side Public License (SSPL) and the Elastic License (ELv2). The goal? Keep AWS and other cloud vendors from offering Elasticsearch as a service without Elastic getting a cut. AWS, whose managed service relied heavily on the ALv2 codebase, responded swiftly, forking Elasticsearch 7.10.2 and Kibana 7.10.2. They stripped Elastic’s proprietary code and telemetry, launching the OpenSearch project under ALv2. It was a bold move but it left a lot of uncertainty: AWS didn’t have expertise in running a community-driven project, and only had a bit more experience managing its own open source projects (such as Firecracker).

Frankly, the odds weren’t great that AWS would succeed. And yet it has.

In 2023 I noted some of OpenSearch’s early successes as it expanded its community and won over some early customers. But it’s really the events in the past year that have demonstrated just how far AWS has come in learning how to contribute to open source in big ways, setting up OpenSearch as a serious contender in enterprise search.

Even though most open source projects have very limited contributor pools and often are the handiwork of a single developer (or a single company), it’s easier to attract volunteer contributors when a project sits within a neutral foundation. As such, AWS demonstrated how serious it was about OpenSearch’s open source success when it moved the project to the Linux Foundation in late 2024, establishing the OpenSearch Software Foundation (OSSF). This wasn’t just admin shuffling; it was strategic. Placing the project within a neutral foundation directly addressed concerns about AWS controlling the project. Suddenly the Technical Steering Committee (TSC) boasted representatives from SAP, Uber, Oracle, Bytedance, and others. Additionally, OpenSearch now can claim more than 1,400 unique contributors (over 350 active), hundreds of maintainers across dozens of organizations, and activity spanning more than 100 GitHub repositories by early 2025. Critically, the percentage of contributions and maintainers from outside AWS has significantly increased, signaling progress towards genuine diversification.

For AWS, whose Leadership Principles almost demand control over customer outcomes (“Deliver Results,” etc.), this is a revolutionary change in how it does business. 

Getting better all the time

Clearly, OpenSearch is on the correct path. With governance solidifying, OpenSearch has pursued aggressive development, guided by a public road map, pushing beyond its roots to tackle modern data challenges, especially in AI/vector search and observability. OpenSearch has significantly moved beyond mere Elasticsearch compatibility. Driven by user needs, OpenSearch has added vector similarity search, hybrid search combining keyword and semantic methods, and built-in neural search capabilities. In 2024 alone, OpenSearch made major strides—adding integration with Facebook’s FAISS, SIMD hardware acceleration, and vector quantization for high-performance semantic searches.

Performance and scalability improvements have also been dramatic. Query speeds increased significantly (up to six times faster than early versions), thanks to extensive optimizations. New features, such as segment replication, have boosted data ingestion throughput by approximately 25%. Additionally, remote-backed storage now enables cost-efficient indexing directly into cloud object storage services, critical for enterprises dealing with petabyte-scale data sets.

This isn’t a community hoping to play catch-up. This is a strategic bid for leadership.

It’s one thing to write good code. It’s quite another to convince enterprises to use it. In this area, there’s growing evidence that OpenSearch is gaining enterprise ground. Just measuring use (without concern for whether it’s paid adoption), by the end of 2023 OpenSearch had surpassed 300 million cumulative downloads, clearly signaling mainstream adoption. AWS, for its part, touts “tens of thousands” of customers (which may be true, but that number includes users of older Elasticsearch versions). Although it’s hard to find public examples of large enterprises adopting OpenSearch, past and future OpenSearchCon events reveal LINE, Coursera, and other significant users (though most of the talks are still given by AWS employees). Job postings show that Fidelity Investments, Warner Bros, and others are OpenSearch users. Plus, a Linux Foundation report found 46% of surveyed users run OpenSearch as a managed service, indicating significant cloud uptake. High demand (87%) for better interoperability also suggests users see it as part of a broader stack.

The long shadow of Elasticsearch

Despite progress, OpenSearch faces challenges, primarily the constant comparison with Elasticsearch. For example, Elastic often claims performance advantages (40% to 140% faster). However, a March 2025 Trail of Bits benchmark comparing OpenSearch 2.17.1 and Elasticsearch 8.15.4 found OpenSearch faster overall on the “Big 5” workload and moderately faster in Vectorsearch (default engines), though results varied. Benchmarks are notoriously unreliable gauges; your mileage may vary.

Nor can OpenSearch still claim to be the open source alternative to Elasticsearch. In late 2024, Elastic added an AGPLv3 license option alongside SSPL and ELv2. Skeptics viewed this return to open source as a cynical response to OpenSearch’s momentum, but in my own conversations with Shay Banon, Elastic’s cofounder, the company had always wanted to return to an OSI-approved license: “I personally always wanted to get back to open source, even when we changed the license. We were hoping AWS would fork and would let us move back when enough time has passed.” Whatever the motivation, Elasticsearch is now just as open source as OpenSearch.

That comparison no longer really matters. OpenSearch has proven it’s more than AWS’ knee-jerk reaction to supply chain risk. OpenSearch is building its own identity, focused on next-gen workloads. Still, OpenSearch’s big challenge is still the process of converting its open governance and permissive licensing into an ecosystem that builds superior search to Elasticsearch or other competitors. There’s a long way to go, but its progress in the past few years, and particularly in 2024, suggests OpenSearch is here to stay—and to win.