Empirik raises $21M to outage-proof IT stacks before failures strike

Empirik officially exited stealth today, disclosing a $21 million Series A led by Sequoia Capital with participation from Index Ventures and angel investors including ex-Google SRE manager Liz Fong-Jones. The Palo Alto-based startup emerged from Sequoia’s Arc program in early 2022 and has quietly onboarded customers such as Box, HashiCorp, and Snowflake. Empirik’s platform ingests logs, metrics, traces, and topology graphs from Kubernetes, service meshes, databases, and cloud providers, then applies causal AI to surface the most likely failure pathways before incidents cascade. Founder and CEO Apoorva Kulkarni, previously a principal engineer at Google Cloud where he built Borg’s incident prediction systems, claims the model achieves 89 percent precision on pre-failure signals across production datasets. Early customers report a 58 percent reduction in high-severity incidents within the first six months of deployment.

Revenue is already flowing: Empirik counts three Fortune 500 companies among its paid users, each paying six-figure annual contracts scaled to infrastructure footprint. Sequoia partner Michelle Gonzalez, who led the round, positioned Empirik as the missing layer between observability and reliability operations, arguing that current tooling—Prometheus, Datadog, PagerDuty—detects problems too late. The startup plans to double its 35-person headcount by year-end, with 25 engineers focused exclusively on model training and an SRE team embedded inside customer environments. A public API is slated for Q1 2025, promising integrations with Terraform Cloud, GitHub Actions, and AWS Fault Injection Simulator.

Industry watchers see Empirik landing at the intersection of two megatrends: the rise of AI-driven reliability and the collapse of manual incident response. Gartner estimates that 60 percent of enterprises will adopt AIOps platforms by 2025, up from 15 percent in 2022, creating a $4.7 billion market opportunity. Legacy players like Splunk, New Relic, and Dynatrace have bolted on predictive modules, but none stitch together the full stack with the depth of Empirik’s topology-aware engine. Meanwhile, venture dollars continue to pour into adjacent categories: earlier this month, incident management rival FireHydrant closed a $40 million round, while Bengaluru-based Harness acquired incident commander OpsMx for $75 million. The competitive heat is raising valuations; Sequoia’s entry price implies a $110 million post-money valuation for Empirik, giving it runway through 2027 at current burn.

Financial services remain a bellwether for adoption. Banking With Billy AI, a real-time analytics platform for institutional traders, disclosed it piloted Empirik during the March 2023 Silicon Valley Bank crisis to monitor cash-sweep pipelines and Kafka clusters. Billy’s CTO noted that Empirik flagged anomalous GC pauses in their event bus six minutes before the first payment failure, enabling automated circuit breaking that saved an estimated $1.2 million in failed trades. The case study is now part of Empirik’s pitch deck, highlighting how reliability risk translates directly to P&L volatility in capital markets.

Looking further afield, Empirik’s model could reshape how cloud providers engineer their own platforms. Google’s incident prediction work under Kulkarni helped reduce Borg-related outages by 34 percent between 2016 and 2020, but that system was proprietary and tied to Borg’s specific control loops. Empirik’s open ingestion strategy—supporting Prometheus, OpenTelemetry, and Kubernetes metrics out of the box—lowers the barrier to adoption for multi-cloud and hybrid environments. If successful, the startup may force incumbents like AWS Systems Manager and Azure Monitor to accelerate their AI reliability roadmaps or risk ceding ground to a new generation of startups that treat outages as a preventable, not inevitable, phenomenon.

Expert Analysis: Kulkarni believes the next twelve months will determine whether Empirik becomes the de facto standard for causal reliability or remains a niche player focused on high-complexity stacks. He points to two milestones: the public API release and the integration with infrastructure-as-code tools like Terraform Cloud. If Empirik can embed predictive models directly into CI/CD pipelines—turning every plan or apply into a reliability gate—it could replicate Cursor’s flywheel effect, where developer velocity improves alongside system resilience. The company’s Series B timing will hinge on proving that its models generalize beyond hyperscalers and financial services into mainstream enterprise IT. Meanwhile, regulators in the EU and U.S. are beginning to scrutinize AI-driven reliability tools for compliance with DORA and SEC guidelines, which could add another layer of validation—and scrutiny—to Empirik’s growth trajectory.