Observability startup groundcover announced a $100 million Series C funding round led by One Peak this week, pushing its total funding to $160 million. The four-year-old company is challenging incumbents like Datadog and Dynatrace by shifting telemetry storage inside enterprise cloud environments using a bring-your-own-cloud architecture, aiming to solve the exploding data costs of autonomous AI agents.
Rewriting the Economics of Enterprise Telemetry
For years, traditional observability platforms operated on a predictable SaaS playbook. Engineers deployed applications, streamed logs, metrics, and traces to vendor-managed infrastructure, and paid heavy bills based strictly on data volume ingestion. According to groundcover co-founder and CEO Shahar Azulay, that model is breaking down as enterprises scale up autonomous software development.
Coding assistants and multi-step AI agents generate unprecedented volumes of operational data. They call external tools, execute complex retrieval pipelines, and trigger endless chains of microservices. When every prompt execution and token consumption cycle generates telemetry, organizations face a brutal choice: sample their traces aggressively, shorten retention windows, or watch their monitoring bills spiral out of control. Groundcover’s $100 million round—which brings its total backing to $160 million and follows a period where the company says it tripled its annual recurring revenue with over 250 paying customers—is a bet that enterprises are ready to break away from those ingestion-based pricing models altogether.
Why Bring-Your-Own-Cloud Changes the Data Plane
Instead of routing telemetry into vendor-locked SaaS repositories, groundcover relies on a bring-your-own-cloud (BYOC) architecture. The data plane, encompassing telemetry storage and heavy processing, stays securely tucked inside the customer’s own AWS, Microsoft Azure, or Google Cloud environment. Groundcover simply supplies the managed control plane and user interface.
This structural pivot alters how companies pay for observation. Because customers already pay for their underlying cloud footprint, groundcover prices its platform by monitored hosts rather than raw data gigabytes. As Azulay noted during a recent briefing, organizations can finally retain complete telemetry without rationing logs or trimming traces just to appease finance teams.
Yet, this architectural choice introduces trade-offs. Host-based economics favor dense Kubernetes environments running massive workloads on relatively few nodes. For lightly utilized fleets spread across sprawling cloud instances, the math might look entirely different. Furthermore, enterprise security teams must audit precisely what metadata leaves the secure cloud perimeter during standard BYOC operations.
Leveraging eBPF for Zero-Instrumentation Observability
At the hardware-adjacent layer of groundcover’s technical stack sits eBPF. By running programs directly inside the Linux kernel, the platform observes network traffic, system calls, and application behavior without demanding manual code instrumentation from developers.
In a modern cloud-native stack running Kubernetes and complex APIs, manual tracing is a maintenance nightmare. eBPF abstracts that friction away. By combining automatic kernel-level collection with OpenTelemetry compatibility and host-based BYOC pricing, groundcover attempts to carve out a distinct identity in a crowded market.
Competition here is fierce. Industry heavyweights like Datadog, which generated more than $3 billion in annual revenue in 2025, alongside Dynatrace, New Relic, Splunk, and Grafana, have spent years building vast partner ecosystems and mature enterprise integrations. While those giants have rushed to release AI assistants and root cause analysis tools, groundcover argues that adding features to a legacy SaaS silo misses the deeper architectural necessity: moving the data plane itself.
Observability as the Feedback Loop for Autonomous Software
Beyond traditional debugging, groundcover is positioning its platform as foundational infrastructure for autonomous agents. Through features like its Agent Mode, human operators can investigate production incidents using natural language queries that span metrics, logs, traces, and Kubernetes events.
The longer-term vision stretches much further. As software increasingly writes, tests, and deploys itself, observability data becomes the critical feedback loop that coding agents need to understand production realities. Instead of operating purely as a post-mortem dashboard for human engineers, the telemetry layer acts as the contextual memory that allows AI systems to evaluate regressions, adjust workflows, and recommend fixes.
For now, humans remain firmly in the loop, approving code changes and managing deployment gates. But as enterprise architectures grapple with the telemetry explosion triggered by autonomous workloads, the battle lines in observability are shifting. The next era of software monitoring will likely be decided not by who has the flashiest dashboards, but by who controls the operational data layer that intelligent software relies upon to survive.
