Elastic Rearchitects Search Engine for Metrics, Claims 30x Speed Edge Over Prometheus as AI Workloads Reshape Observability Spend
Special call held September 22, 2026 detailed the company's new metrics offering and go-to-market strategy for consolidating Observability spend
Elastic used a special investor call on September 22 to lay out the technical underpinnings and commercial logic behind its metrics offering, launched at the start of the fiscal year in June. The pitch centers on a rearchitected version of Elasticsearch that stores metrics in a columnar format rather than the document store the company has historically used for logs, a shift management says was necessary to compete head-on with dedicated metrics platforms like Prometheus, Mimir, and ClickHouse.
Santosh Krishnan, Senior Vice President of Security and Observability Solutions, disclosed that the project consumed 12 to 18 months of R&D before its June launch, a data point that quantifies the engineering lift behind what the company is positioning as a multiyear competitive moat. "We always get an asymmetric advantage whenever we add innovation to the platform," Krishnan said, arguing that Elastic's existing strength in unstructured log data now extends into a purpose-built metrics engine.
Benchmark Claims Set Up a Direct Confrontation With Incumbents
The most concrete new data point from the call was performance benchmarking: Bahaaldine Azarmi, General Manager of Observability, said Elastic's new columnar store is 30 times faster than Prometheus and Mimir and 8 times faster than ClickHouse, with benchmarks published and reproducible via open-source code. The architecture uses techniques such as "dim filter" to manage high-cardinality data in memory and tuned codecs for storage efficiency, addressing what Azarmi described as the central technical challenge in metrics: unlike logs, metrics arrive with numerous dimensions and labels that must be queried selectively without parsing irrelevant data. Management is framing cardinality — the number of dimensions a system can handle without penalizing customers on cost or performance — as the primary battleground against competing metrics platforms, several of which the company says force customers into data retention trade-offs that create blind spots during incidents.
PromQL Compatibility Nearing Completion
Azarmi disclosed that Elastic has reached 90% compatibility with PromQL, the query language standard among Prometheus users, and is "on track to get to 100%." This is a meaningful data point because it lowers switching costs for the large existing base of Prometheus users, a cohort management explicitly identified as one of three go-to-market targets, alongside the company's existing logs customer base and customers currently being "penalized" on pricing by other metrics vendors for higher retention or cardinality.
AI Workloads Are Framed as the Demand Catalyst, Not Just a Feature
Management repeatedly tied the metrics launch to the operational burden created by agentic AI. Azarmi argued that agent-based workloads are structurally different from traditional application monitoring because reasoning steps, tool calls, and turn counts are unpredictable, unlike the well-defined transaction paths of legacy applications. "AI is really causing an explosion in metrics," he said, pointing to GPU cycles, LLM calls, and agent harnesses as new categories of signal that require monitoring. Krishnan reinforced this on the analyst Q&A, noting that Observability spend is being pulled by AI in two distinct ways: infrastructure monitoring for the compute buildout itself, and a newer category of agent-level monitoring focused on safety and behavior, not just uptime.
Adoption Will Be Gradual, Not a Hockey Stick
On the commercial trajectory, Krishnan was notably candid that near-term adoption will not be a hockey stick. He told analysts that early design partner engagement has been "extremely positive," but cautioned that enterprise sales cycles mean adoption will follow "more of a ramp style" rather than an immediate inflection. Roughly two-thirds of Elastic's existing business today is tied to logs and SIEM, and Krishnan confirmed that the metrics product's near-term opportunity lies primarily in attaching a new sale to that existing SRE-focused customer base, where third-party infrastructure monitoring tools often already carry higher economic value than Elastic's log analytics contracts. Management declined to quantify a specific revenue uplift per customer, with Krishnan saying only, "it's early days... stay tuned."
Spend Consolidation Plus a New Layer of Incremental Demand
Responding to a question from Cantor Fitzgerald's Thomas Blakey on whether the metrics push represents share-shift from competitors or genuinely incremental spend, Krishnan characterized it as both. "There is definitely a spend consolidation aspect to it," he said, but added that AI-driven infrastructure growth is prompting customers to reevaluate existing tooling altogether, creating what he called "a little bit of the gravy on top" beyond simple vendor consolidation. Metrics data volumes coming into the system, he noted, are of a comparable order of magnitude to log volumes, though retention periods for metrics tend to be shorter.
Deductive AI Acquisition and October Product Event Loom as Catalysts
Azarmi flagged the recent acquisition of Deductive AI, a company specializing in agent-driven root-cause-analysis investigation, as feeding into a broader Observability announcement planned for Elastic's October 8 event in New York. He described the company's ambition to let AI agents perform root-cause analysis across metrics, traces, logs, and vectorized knowledge bases within a single integrated platform, a capability he said is only possible because all four data types now sit inside the same architecture. Investors should treat the October event as the next checkpoint for evidence that the metrics launch is translating into bookings, given management's own acknowledgment that quantifiable adoption metrics are not yet available.
Customer Proof Point From Norion Bank
A customer video featured Norion Bank, an eight-year Elastic user, whose Observability engineers said early testing of the new time-series data streams produced "substantial" storage savings within a month of deployment. One engineer offered a pointed articulation of the category's value proposition under budget pressure: "The question often becomes how do we get -- make this cheaper instead of how does this get us out of incidents faster." That tension between cost containment and detection speed is precisely the trade-off Elastic is positioning its consolidated metrics store to resolve, though the bank's commentary also underscored that compliance regimes like DORA and GDPR add retention and audit requirements that complicate simple cost-cutting narratives around telemetry data.