The observability market is fracturing along a new fault line: data sovereignty versus vendor lock-in. As AI agents generate exponentially more telemetry data, enterprises are discovering that traditional monitoring platforms' per-byte ingestion models become prohibitively expensive while creating compliance nightmares in regulated industries.

Paris-based Tsuga raised $35 million in Series A funding on June 23, 2026, led by Singular with participation from General Catalyst, DST Global Partners, Quantumlight, Picus, and Databricks Ventures. The round comes just seven months after the company emerged from stealth with a $10 million seed led by General Catalyst, bringing total funding to approximately $45 million. Founded in 2024 by Gabriel-James Safar and Sébastien Deprez — both former engineers who sold their previous company Madumbo to Datadog in 2019 — Tsuga operates an AI-native observability platform with a fundamental architectural difference: it deploys entirely inside customer cloud environments rather than ingesting data into vendor infrastructure.

The 'Bring Your Own Cloud' GTM Wedge

Tsuga's core differentiation isn't just technical — it's a complete inversion of the observability business model. Legacy platforms like Datadog, Splunk, and Dynatrace built their economics around ingesting customer telemetry into their own infrastructure, charging per byte processed. This model worked when applications generated manageable data volumes, but AI workloads break the math. A single AI agent can generate thousands of telemetry events per interaction, making traditional observability costs scale exponentially with AI adoption.

The company's 'Bring Your Own Cloud' architecture sidesteps this entirely. Tsuga's observability stack runs inside the customer's existing cloud environment, meaning telemetry never leaves their security perimeter. This eliminates per-byte ingestion charges while solving data-residency requirements that make third-party SaaS problematic under European frameworks like GDPR or in regulated industries like financial services. The platform can run automated root-cause analysis on complete, unsampled data because storage and compute costs remain within the customer's existing cloud spend.

This architectural choice creates a structural advantage as AI agents proliferate. Where legacy platforms face a choice between sampling data (losing observability) or charging prohibitive rates, Tsuga can process complete telemetry streams without vendor lock-in on data costs. The approach particularly resonates with European enterprises navigating data sovereignty requirements — a market dynamic that American observability vendors struggle to address with centralized SaaS models.

Tsuga's go-to-market reflects this positioning through an enterprise sales motion with forward-deployed engineering. Rather than self-serve onboarding, engineers work alongside client teams to tune the deployment within their specific cloud architecture. This high-touch model aligns with the platform's enterprise ICP while building deep technical relationships that competitors can't easily replicate. Named customers include Le Monde, which used the platform to monitor infrastructure through French municipal elections, alongside Camunda, Buk, and Black Forest Labs.

Enterprise Sales With Strategic Partnership Leverage

The Series A investor composition reveals Tsuga's partnership-led distribution strategy. Databricks Ventures' participation doubles as a product integration, allowing customers to route observability data directly into Databricks for further analysis. This creates a natural upsell path while positioning Tsuga within Databricks' broader AI infrastructure ecosystem — a critical advantage as enterprises consolidate AI toolchains around fewer vendors.

The forward-deployed engineering model serves multiple GTM functions beyond customer success. These embedded engineers become product intelligence assets, identifying feature gaps and integration opportunities that inform roadmap priorities. They also create switching costs through deep technical integration, making Tsuga sticky even as competitors attempt to match the data sovereignty positioning. The model scales through what the company calls its Skills library and agent-building toolchain, allowing customers to customize monitoring for their specific AI workloads.

Tsuga reported several million dollars in contracted ARR with average contract values in the six figures, suggesting the platform commands premium pricing despite eliminating per-byte charges. This pricing power reflects the value of avoiding data egress costs and compliance complexity rather than competing on cost alone. The enterprise sales cycle likely extends 6-12 months given the technical evaluation required, but the forward-deployed model helps accelerate proof-of-concept deployments.

The rapid funding timeline — $45 million raised in under seven months from stealth — indicates intense investor appetite for AI infrastructure companies that solve both technical and regulatory challenges. European VCs particularly value companies that can compete with American incumbents on data sovereignty grounds, creating natural geographic moats that pure-play technical solutions lack.

Category Implications: Observability Unbundles From Ingestion

Tsuga's traction signals a broader unbundling of observability from data ingestion economics. Traditional monitoring platforms built competitive moats around proprietary data formats and vendor-specific query languages, making it expensive for customers to switch once telemetry accumulated in vendor systems. But AI workloads generate too much data for this model to remain economically viable, forcing a architectural reset.

The shift mirrors earlier enterprise software transitions where on-premise deployments gave way to cloud-native architectures, then cloud-native gave way to multi-cloud strategies. Observability is following a similar path: centralized SaaS platforms worked for traditional applications, but AI-era data volumes and sovereignty requirements are driving demand for distributed, customer-controlled deployments. This creates opportunities for startups that design around new constraints rather than retrofitting legacy architectures.

The European regulatory environment accelerates this transition. GDPR, the Digital Markets Act, and emerging AI governance frameworks create compliance complexity that American observability vendors struggle to navigate with centralized data processing. Tsuga's Paris base positions the company to capture European enterprises that need AI observability without cross-border data transfers, a market segment that Datadog and Splunk can't easily address without fundamental architecture changes.

The Series A timing also reflects AI infrastructure investment patterns. Venture capital is flowing toward companies that solve AI adoption blockers rather than core AI capabilities themselves. Observability sits at this intersection — enterprises need monitoring before deploying AI agents at scale, but existing tools weren't designed for AI-generated telemetry volumes. This creates a narrow window for startups to establish market position before incumbents adapt their architectures.

What Founders Can Take From This

Regulatory arbitrage creates durable competitive advantages. Tsuga's data sovereignty positioning isn't just a feature — it's a structural moat that American competitors can't easily replicate without rebuilding their entire platform architecture. Founders should identify regulatory requirements that incumbents struggle to meet, then design solutions around those constraints from day one.

Partnership-led distribution scales faster than pure sales motions. Databricks Ventures' participation creates immediate customer access while validating technical integration. Rather than competing for attention in crowded categories, founders can leverage strategic partnerships to embed within existing enterprise workflows and budgets.

AI infrastructure timing requires solving adoption blockers, not core capabilities. Tsuga doesn't build AI models — it solves the observability challenges that prevent AI deployment at scale. The fastest path to AI infrastructure success often involves identifying operational bottlenecks rather than advancing core AI research.

The Forward-Deployed Engineering Playbook

Tsuga's customer engagement model offers a blueprint for technical infrastructure startups navigating complex enterprise sales cycles. Forward-deployed engineers serve as product intelligence assets while creating switching costs through deep integration work. This approach works particularly well for platforms that require significant customization or operate within regulated environments where standardized deployments aren't feasible.

The model also addresses a common startup challenge: how to gather product feedback from enterprise customers who can't easily articulate technical requirements. Embedded engineers observe actual usage patterns and integration challenges, providing product insights that traditional customer success teams might miss. This intelligence becomes increasingly valuable as the platform scales beyond initial use cases.

The Series A capital will fund expansion to 100 people, with much of that hiring likely focused on scaling the forward-deployed engineering capacity. This suggests Tsuga sees customer engagement as a core competency rather than a temporary go-to-market tactic — a strategic choice that could differentiate the company as observability becomes increasingly commoditized on pure technical features.

The question now is whether Tsuga can maintain its architectural advantages as AI workloads evolve and incumbents respond. Legacy observability vendors have deep enterprise relationships and substantial engineering resources to rebuild around data sovereignty requirements. But they also have existing customer bases built around ingestion-based pricing models, creating innovator's dilemma dynamics that could preserve Tsuga's window to establish market position. The company's ability to scale forward-deployed engineering while maintaining technical differentiation will determine whether data sovereignty becomes a sustainable competitive moat or a temporary market arbitrage opportunity.