| TL;DR:OpenSearchCon NA 2026’s nine keynotes showed OpenSearch becoming the trust layer for agentic AI. Production adopters like CERN, Uber, and SAP validated it, research confirmed rising awareness and vendor neutrality, and security emerged as the next frontier. |
AWS’s Tia White put a name to this at the top of the conference: speed, collaboration and transparency, the three things she asked the community to build the next five years around. Nine keynotes later, that’s not just a nice phrase, it’s what we watched happen. Watch any single talk from OpenSearchCon North America 2026 and you’ll hear a good story about one organization’s work. Watch them back to back, and a different picture emerges: a project that has quietly stopped being a search engine and started becoming the trust layer underneath the agentic AI era.
The timing wasn’t incidental. The same week we hosted OpenSearchCon, we announced three new general members: Adelean, Intel and Sidecar, along with two newly certified Long-Term Support providers: Seacom and Sidecar. We also reported that contributor participation grew 31% year over year and project downloads surpassed 2.4 billion, up 140% from last year. That’s not just background noise to the keynotes.
From content to context
The clearest thread running through this year’s keynotes was a reframe of what retrieval even means, one our community is living out in real time. AI agents don’t just need to find the right document anymore, they need the right evidence, the access controls that come with it, and a memory of what’s already been decided. That’s a fundamentally different job than keyword search was built to do, and speaker after speaker described OpenSearch growing into it: as the place agents go to reason, not just the place humans go to search.
Built for the extremes, not the demo
What made this year’s program credible rather than aspirational was who showed up to validate it. CERN running search infrastructure for the largest scientific instrument on Earth. Uber serving search queries at a scale most companies will never touch. SAP running roughly 18,000 production instances under strict regulatory compliance. These aren’t hypothetical use cases or roadmap slides, they’re organizations in our community running OpenSearch today, at the kind of scale that exposes every weakness a platform has, and choosing to keep building on it anyway.
Vendor neutral, and increasingly proven to be
For a project that started inside one company, the data we shared this year made a real case that it no longer belongs to one. Our research showed awareness climbing from 68% to 89% in two years and production deployment nearly doubling, with 24% of non-adopters associating the project primarily with its founding contributor. TThis week’s membership news reinforced that: Intel and two specialist consultancies chose to join the Foundation’s governance, a commitment to the project itself.
Security is catching up to capability, on purpose
Perhaps the most important undercurrent of the week was a warning delivered from more than one stage, one we take seriously: as this platform becomes the place where an organization’s knowledge actually lives, securing it can no longer be treated as a checkbox. Encryption in transit and at rest is necessary, but it’s not sufficient once the data itself becomes the target. That shift in thinking, and real engineering work addressing it, was on display across multiple keynotes this year.
The takeaway
This is no longer a story about whether OpenSearch will be adopted. The harder questions are already on the table: how agents should receive context, how to secure data that now holds everything a company knows, how to keep a project this large moving at the speed its users need. What this year’s keynotes showed is that those questions are being answered in public. Speed looked like Uber’s load shedding work shipping in a release and Aiven’s cloud support landing in a matter of months. Collaboration looked like SAP, SAS Institute and AWS carrying FIPS support across the finish line together. Transparency looked like publishing our research for anyone to read. Tia White asked for all three on the first morning. By the last keynote, the community had already answered her, and we’re proud to be the neutral home where that work keeps happening.
The keynotes
Nine keynotes, nine reasons to make time this week. Here’s where to start.
Tia White, AWS – The next five: Speed, collaboration, and the future we build together
The conference began with a question every mature open source project eventually faces. OpenSearch met the mission it was founded on, and White makes a clear case for what this community should do next as the ground shifts from human search to agent-driven retrieval. Watch for her vision of relevance tuning built by five or six organizations in parallel, and ask yourself whether your team is solving alone something the community could solve together.
Watch now: https://youtu.be/4VaCooTfsr8?si=fe9MpmOLR3eyvxSK
Bill Pearson, Intel – Intel + OpenSearch: Co-designing the future of search and analytics
White’s call for collaboration across company lines found its first answer on the very next talk. Pearson walks through two years of Intel performance work, all of it upstream and available to anyone running OpenSearch, including a vector search engine that breaks the usual tradeoff between speed, memory and accuracy. Then he explains why Intel decided to join the Foundation. Worth watching for the benchmarks alone, and for a hardware company’s take on why openness wins.
Watch: https://youtu.be/eibywF7NYHY?si=iQOL9JHTEFGyHuka
Socrates Papadopoulos, CERN – From particle collisions to queries: Ten years of search at CERN
If Intel shows how fast OpenSearch can go, CERN shows how far it can stretch. Their starting point is a petabyte of data per second. This talk shows what OpenSearch looks like when it supports the Large Hadron Collider, a global computing grid and one of the largest security indexes anywhere, and it’s one of the best tours you’ll find of how many jobs a single platform can do at once. Stay for the part about operators talking to their monitoring data through the OpenSearch MCP server.
Watch: https://youtu.be/2S6mIhKK-8o?si=gLOUGZH-z1oS2V1j
Pushkala Pattabhiraman, IBM – From little r to big R, why OpenSearch is the retrieval engine behind the AI era
CERN’s operators talking to their data is a glimpse of where search is heading, and this talk explains what it takes to get there. Pattabhiraman uses one deceptively ordinary example, a customer returning a laptop, to show why finding the right content isn’t enough anymore and what it actually takes to give an agent context it can trust. You’ll come away with a framework you can apply to your own architecture, and a line worth remembering: memory is evidence, not authority.
Watch: https://youtu.be/JIPOf6sBFAU?si=gsmf2mSq6aarrOsA
Dmitri Kan and Tony Piazza, Aiven – Our ambition for opensearch on aiven: everywhere for everyone
Of course, a retrieval layer for agents only matters if you can run it wherever your data lives. Kan makes the case that OpenSearch belongs on every cloud and at every budget, down to a free tier with no strings attached. Piazza tells the story of how one engineer’s contribution to the migration assistant opened OpenSearch up to Google Cloud and AWS Terraform users, with Azure next. If you’re planning a migration, watch this before you start.
Watch: https://youtu.be/h-GHsINYwWU?si=GNpwUj2FdO7Cf_-2
Karthik Mariappan, Portal26 – OpenSearch as AI’s memory: the encryption questions nobody’s asking
The more places OpenSearch runs, and the more an organization puts into it, the higher the stakes. This is the talk to send your security team. As OpenSearch becomes the place where a company’s knowledge lives, Mariappan asks a pointed question: if someone had your admin credentials right now, would your encryption actually protect anything? His live demo, reconstructing text from vector embeddings, is reason enough to watch.
Watch: https://youtu.be/Z6_EypGV1uI?si=vTDIOgLYSR80iWYQ
Bianca Lewis and Carl Meadows, OpenSearch Software Foundation – OpenSearch at 5: Growth, Trust, and What’s Next
After two days of individual stories, Lewis and Meadows zoomed out to the market as a whole to answer where OpenSearch actually stands with data from our largest research project yet, covering awareness, production adoption, AI plans and the vendor neutrality question head on, including a candid take from one of the people who started the project at AWS. If you need to make the case for OpenSearch inside your organization, start here.
Watch: https://youtu.be/04eyZZmuVLk?si=1EyY2r418Y5SNbT8
Mingmin Chen, Uber – Uber’s Journey to Modernize Search Platform with Innovation and Embracing the Ecosystem
The research shows where the market is going. Uber shows what it looks like when you’re already there. Every time you open the Uber app, you’re using OpenSearch, and Chen shows how it powers rides, maps and Uber Eats while handling the kind of traffic spikes that would take most systems down. The real value is in the engineering detail: how Uber built resilience, scale and load shedding, then contributed it back so everyone else gets it too.
Watch: https://youtu.be/XJMhIkjBx9k?si=85xP3bhLPaxVDTKV
Wolfgang Theilmann, SAP – Enterprise Observability With OpenSearch
Uber’s story is about scale. SAP adds the other thing enterprises can’t compromise on: trust. Theilmann asks what observability looks like when AI agents are running the business, answers from 18,000 production instances, then tells the story of FIPS 140-3 support, a years-long effort across SAP, SAS Institute and AWS that shows exactly how big features get built in an open community. Essential viewing for anyone in a regulated industry.
Watch: https://youtu.be/44SNmv3msjA?si=kEq3Jva11JdJDuhu
To view the recorded keynotes and sessions, view the full playlist here:
https://youtube.com/playlist?list=PLOZAsMYNWb40&si=C0_lSyyp6vSlmurz
