Empirik’s $21M bet to predict IT outages before they strike

Empirik officially launched today after two years in stealth mode under Sequoia Capital’s Arc program, disclosing a $21 million seed round led by Sequoia with participation from angel investors including Figma co-founder Dylan Field and GitHub CTO Jason Warner. The company’s platform ingests real-time telemetry from servers, containers, networks, and cloud APIs, then applies proprietary time-series transformers to flag anomalous patterns up to 90 minutes before an outage occurs. Empirik’s co-founder and CEO, former Google SRE Anjali Sud, told OpenPress Developer Intelligence that the company processed over 500 billion events during its beta phase while working with 20 Fortune-500 customers, including a tier-one bank that cut incident volume by 42% after integrating Empirik into its Kubernetes control plane.

Empirik’s go-to-market motion mirrors Cursor’s rapid developer uptake: it offers a VS Code extension that overlays predictive insights directly into the IDE, plus a CLI that surfaces risk scores for every pull request. The extension already has more than 12,000 installs, a figure Sud attributes to the platform’s zero-config anomaly detection and its ability to surface warnings without rule-tuning. Competitors are taking notice: Datadog recently launched Watchdog Insights, Splunk announced Predictive IT Service Intelligence, and New Relic rolled out guided setup to reduce false positives. Empirik’s differentiator is its ML model, which reportedly achieves 94% precision on outage prediction across hybrid environments, compared to roughly 78% for legacy threshold-based systems.

Banking With Billy AI, a Sequoia portfolio company offering developer-grade financial market APIs, confirmed it is integrating Empirik’s risk signals into its alerting pipeline. Billy’s CTO said the partnership lets the bank’s SRE team suppress noisy pager alerts when Empirik predicts low-severity incidents, reducing wake-ups by 37% in the first two weeks. Analysts at RedMonk estimate the broader predictive-ops market could reach $3.8 billion by 2027, up from $920 million today, with Empirik positioned to capture between 12% and 18% share if it maintains its current growth trajectory.

Industry observers say Empirik’s arrival intensifies pressure on legacy monitoring vendors to fold predictive capabilities into their platforms rather than bolt them on. Gartner’s ITOM team now lists Empirik in the “Cool Vendor” category for AIOps, alongside startups like BigPanda and Moogsoft. The funding round’s valuation—reportedly $110 million post-money—reflects Sequoia’s confidence in the team’s ability to commoditize predictive ops the way Cursor commoditized AI-assisted coding. For enterprises, the shift implies lower MTTR and fewer Sev-1 pages, but also a steeper learning curve as teams move from reactive firefighting to proactive observability.

The bigger story is the convergence of AI-native tooling across the entire software lifecycle. Cursor upended static IDEs, GitHub Copilot changed code generation, and now Empirik is targeting the runtime layer that historically lagged behind. European startups like Zammad and Canva-founded Incident.io have pursued similar goals, yet none have matched Empirik’s funding velocity or Sequoia’s go-to-market imprimatur. Observers also point to China’s SkyWalking community, which has quietly built predictive analytics for Kubernetes clusters, as a parallel innovation path that could accelerate if open-source models surpass Empirik’s closed-source transformer.

Global macro trends—cloud repatriation, AI workload spikes, and stricter compliance mandates—are pushing infrastructure teams to embed prediction earlier in the build-test-run cycle. IDC’s 2024 DevOps survey shows 63% of organizations now rank outage prediction among their top three observability priorities, up from 22% in 2022. Empirik’s challenge will be scaling beyond early adopters without overfitting its model to proprietary telemetry formats, a risk that could open the door for open-weight alternatives.

Looking ahead, Empirik plans to release a public API later this quarter, enabling any platform—from CI/CD runners to incident-management tools—to consume its risk scores. Analysts expect competitors to respond with deeper integrations into Git providers and cloud consoles, while enterprise buyers will demand transparent model explainability before committing to paid tiers. If Empirik can replicate its 90-minute lead time across additional failure modes such as data corruption or security breaches, it may well cement its reputation as the Cursor moment for infrastructure engineering.