groundcover's $100M Series C tests BYOC observability for AI workloads
groundcover raised a $100 million Series C this week, led by One Peak, bringing the four-year-old observability startup's total funding to $160 million. The round is framed as evidence that AI workloads are reshaping observability economics, but the specific story is architectural: groundcover is betting that telemetry should stay inside customer clouds and pricing should track hosts rather than data volume. Whether that bet works depends less on feature parity with incumbents and more on whether BYOC pricing delivers the cost predictability it promises in environments with high telemetry density. groundcover's own briefing materials flag caveats about what metadata still leaves customer environments and where per-host pricing underperforms, and that is where enterprise evaluation actually happens.
The pricing claim is the load-bearing one. Most observability vendors charge by data ingested, which creates a direct conflict between cost and visibility when telemetry volume grows. groundcover's argument is that decoupling price from volume lets enterprises retain everything and decide what to analyze later, rather than sampling or dropping data at the collection layer to control spend. The company says it has more than 250 paying customers, tripled annual recurring revenue over the past year, and is increasingly replacing established observability platforms inside enterprise environments. Those are company-reported figures, and the source does not present independent validation. Read as commercial momentum they suggest traction; read as architectural proof they do not establish that BYOC delivers under enterprise workload conditions.
The pricing model is not universally cheaper. groundcover's own briefing acknowledges that per-host economics work best for organizations with high telemetry density, such as dense Kubernetes fleets generating large volumes of operational data per host. For lightly utilized fleets, the math can shift the other way. That is a real constraint, not a marketing gloss, because it means the BYOC pitch fits a specific infrastructure profile and not the long tail of enterprise estates. Procurement teams evaluating the model will need to model their own telemetry density, not assume the savings generalize.
The data-residency story carries a similar caveat. groundcover describes its bring-your-own-cloud model as keeping the customer data plane inside the customer's AWS, Microsoft Azure, or Google Cloud environment while the vendor operates a managed control plane and user experience. That is a real architectural difference from fully SaaS observability, where telemetry is processed and stored inside the vendor's infrastructure. But the company's own briefing materials recommend scrutinizing exactly what metadata leaves customer environments in standard BYOC deployments, rather than assuming that no operational data ever reaches vendor infrastructure. The verification work for that claim is on the enterprise side, and groundcover has not published the kind of independent assessment that would close the question.
The eBPF component is increasingly table stakes. Most observability vendors now incorporate eBPF into their collection layer because it enables kernel-level visibility without requiring code instrumentation. groundcover's stated differentiator is the combination: eBPF-first collection, customer-controlled storage, OpenTelemetry compatibility, and host-based pricing inside a single product. groundcover's own research materials acknowledge that none of these technologies individually represents a competitive moat. The claimed moat is integration, which is harder for competitors to replicate but also harder for groundcover to demonstrate without independent benchmarks.
The agent-as-user framing is the most speculative part of the pitch. groundcover describes observability as infrastructure for autonomous software, with the Agent Mode product allowing engineers to investigate incidents using natural language across logs, metrics, traces and Kubernetes events. The longer-term vision is that observability becomes a feedback mechanism for coding agents: production context flows back to the systems that generate code, and the loop closes. Today, humans remain in the loop for production changes. Whether agent-to-observability feedback becomes a meaningful workflow category, or stays a marketing frame, is a question the source does not test.
The market context is hostile. Datadog alone generated more than $3 billion in annual revenue in 2025, and Dynatrace, Cisco's Splunk business, Grafana Labs and New Relic all maintain enterprise support organizations and partner ecosystems that newer entrants cannot easily replicate. Gartner currently tracks more than one hundred observability products, and nearly every major vendor now markets AI-powered operational capabilities. The Series C validates groundcover's commercial trajectory, not the architectural thesis. The thesis gets tested in procurement, where groundcover's own caveats about metadata leakage and workload-specific economics are the constraints teams will need to verify independently.
The specific claim that BYOC + eBPF + host-based pricing is the right architecture for AI-driven enterprise observability is groundcover's argument, not an industry conclusion. The source-noted caveats about metadata leakage and workload-specific economics are the load-bearing constraints, and the verification work falls on enterprise procurement rather than the vendor: groundcover has not published independent workload benchmarks, third-party metadata audits, or migration data that would let evaluators skip the step. Whether that verification becomes standard practice will determine whether the architectural bet compounds or stalls.