Observability startup groundcover announced a $100 million funding round this week led by One Peak, bringing the company’s total external capital to $160 million. According to groundcover, it has more than 250 paying customers, tripled its annual recurring revenue (ARR) over the past year, and is increasingly replacing legacy observability platforms inside enterprises.
Those figures are company‑reported, and while they point to momentum in a competitive market, independent verification of the specific claims is limited.
Why groundcover says observability needs a new architecture
groundcover’s central argument is that artificial intelligence has changed the basic assumptions behind modern observability. Rather than only adding AI features to existing platforms, the four‑year‑old company argues the underlying architecture must evolve because AI systems are becoming more autonomous, generate far more telemetry, and directly participate in software operations.
Traditionally observability has been a post‑production discipline: deploy, monitor logs/metrics/traces, debug incidents and improve reliability. AI‑assisted development and autonomous agents accelerate deployments, increase code churn, and create new telemetry layers (prompt execution, model latency, token usage, retrieval pipelines, tool invocations and agent behavior). As organizations run agents that execute multi‑step workflows and interact with production systems, the telemetry those activities produce becomes more valuable and harder to discard.
Telemetry growth, pricing tension and the BYOC model
Many incumbent vendors charge based on data ingestion or storage. That has driven practices like sampling traces, shortening retention and restricting which data is collected—cost‑saving measures that also reduce visibility. groundcover contends enterprises increasingly want to retain complete telemetry rather than discard it.
To that end, groundcover offers a bring‑your‑own‑cloud (BYOC) architecture: customers keep the data plane (telemetry storage and processing) inside their own AWS, Microsoft Azure or Google Cloud environments while groundcover runs a managed control plane and user experience. A fully self‑hosted deployment option is available as well.
Because customers already pay for their cloud infrastructure, groundcover avoids ingestion‑based pricing and instead charges primarily by monitored hosts. The company argues host‑based pricing is more predictable and lets teams keep full telemetry for operational analysis, compliance and AI‑native troubleshooting. CEO and co‑founder Shahar Azulay summed it up: “We don't price by data volume. We price by the size of the infrastructure.”
That approach is not universally cheaper — organizations with many lightly used hosts may see different economics than dense Kubernetes environments producing high telemetry volume. groundcover acknowledges the per‑host model is most advantageous where telemetry density is high.
eBPF as a technical differentiator
A second core element of groundcover’s strategy is eBPF, a Linux kernel technology that enables deep observation of network traffic, system calls and application behavior with minimal manual instrumentation. eBPF can speed deployments and broaden visibility across cloud‑native infrastructure, which matters as AI systems generate complex cross‑service interactions.
While many vendors now incorporate eBPF, groundcover argues its differentiation comes from combining automatic eBPF collection with customer‑controlled storage (BYOC), OpenTelemetry compatibility and host‑based pricing within a single platform. The company acknowledges no single technology is an unassailable moat; the claimed edge is the combination.
Observability for AI agents, not just human operators
groundcover also frames observability as infrastructure for autonomous software development. Historically, observability platforms helped human operators troubleshoot production incidents. groundcover expects future platforms will increasingly serve AI agents as both users and consumers of operational context.
Its Agent Mode product lets engineers query incidents in natural language across logs, metrics, traces and Kubernetes events. More broadly, Azulay envisions observability providing continuous context that coding agents can use to evaluate changes, detect regressions and propose or enact fixes — initially with human approval.
Today humans remain in the loop: Agent Mode surfaces recommendations, but production changes still require human sign‑off. Azulay expects autonomy to increase gradually as organizations grow comfortable with AI participation in operational workflows.
Competition, cautions and what’s at stake
groundcover is entering a crowded market dominated by established vendors such as Datadog, Dynatrace, Splunk (Cisco’s business), Grafana Labs and New Relic, which together represent years of product maturity and large revenues. For example, Datadog generated more than $3 billion in annual revenue in 2025. Gartner currently tracks over one hundred observability products.
groundcover does not claim it will instantly outscale incumbents. Its pitch is that AI may trigger an architectural inflection similar to shifts seen during the move to cloud‑native computing. According to the company, many customers first adopt groundcover to cut observability costs and then stay to access richer telemetry and AI‑native workflows. The firm says deployments often replace incumbent platforms rather than run alongside them, but it has not published independent migration studies to validate that pattern.
The company’s self‑reported growth metrics and vendor‑authored customer case studies should be treated cautiously until corroborated by independent data. The briefing also recommends examining exactly what metadata leaves customer environments under BYOC deployments instead of assuming no operational data ever reaches vendor infrastructure.
Viewed narrowly, groundcover’s Series C is a large infrastructure funding round. Viewed more broadly, it signals a debate about how observability should evolve as software increasingly writes, tests and operates itself. If AI continues to drive rapid growth in operational data, assumptions about telemetry collection, pricing and storage could come under pressure—potentially forcing vendors that charge for data ingestion to revise their economics. New entrants that design around changing assumptions from the start have an opening; groundcover’s bet is that a BYOC, eBPF‑first, host‑priced platform built for autonomous software can capture that opportunity. Whether that architectural wager endures will depend on enterprise adoption over the coming years.
