Tsuga raises $35M Series A to keep AI observability inside customers' own cloud infrastructure

Tsuga Secures $35M to Challenge Legacy Observability Pricing

Tsuga, a Paris-based startup building observability software for AI workloads, has closed a $35m Series A funding round just six months after emerging from stealth. The round was led by Singular, with participation from returning investor General Catalyst, alongside new investors DST Global and Quantumlight, plus Picus and Databricks Ventures 1

1

. This brings the company's total capital raised to approximately $45m since its December 2024 launch, a fundraising pace that reflects strong investor appetite for AI infrastructure solutions 2

2

.

Founded by Gabriel-James Safar and Sebastien Deprez, both former Datadog executives who sold their previous company Madumbo to Datadog in 2019, Tsuga targets a fundamental problem the founders witnessed firsthand: legacy observability platforms charge customers based on data volume, a model that breaks down as AI agents generate exponentially more telemetry data 1

1

.

Bring Your Own Cloud Model Addresses Cost and Compliance

The core innovation behind Tsuga lies in its Bring Your Own Cloud model, which inverts the traditional observability architecture. Instead of ingesting customer telemetry into a vendor's cloud and charging per byte, Tsuga deploys directly inside the customer's environment, ensuring data never leaves the customer's security perimeter 1

1

. This approach eliminates the ingestion tax that compounds as AI systems scale, while simultaneously addressing data residency requirements that matter increasingly to regulated industries and European customers 2

2

.

The platform enables engineering teams to maintain stronger control over costs, scalability, and data sovereignty—critical concerns as autonomous AI agents generate telemetry at volumes legacy platforms weren't designed to handle 2

2

. Forward-deployed engineers work alongside client teams to tune deployments and reduce the volume of data processed and retained, while automated root-cause analysis runs on complete, unsampled data 1

1

.

Strategic Databricks Partnership and Early Traction

The Databricks investment signals more than financial backing—it represents a strategic partnership where Tsuga customers can route observability data directly into Databricks for further analysis 1

1

. This integration fits Databricks' broader push into security and operational data, betting that the data layer and AI observability layer belong together inside customer environments.

Since launching in December 2024, Tsuga reports several million dollars in contracted annual recurring revenue with average contract values in the six figures. Its customer base includes Le Monde, Camunda, Buk, and Black Forest Labs . Le Monde used the platform to monitor infrastructure during French municipal elections, while Camunda and Buk run it across multi-cloud setups with strict data-residency requirements 1

1

.

Scaling AI Agents and Go-to-Market Efforts

The Series A funding will fuel expansion of Tsuga's team, development of its Skills library and agent-building toolchain, and scaling of its forward-deployed engineering model 1

1

. The platform includes a bundled MCP server and command-line tool that let engineering teams build their own AI agents on top of the observability layer, all within their security boundary 1

1

. The funds will specifically accelerate go-to-market efforts and the rollout of AI agents designed to power a new generation of autonomous systems 2

2

.

CEO Safar framed the challenge bluntly: "The incumbents built good businesses on a model that no longer works. Sending your telemetry to a vendor's cloud made sense when data volumes were manageable and AI was not writing and deploying your code. Neither of those things is true any more" 2

2

. The critical question ahead is whether Tsuga's hands-on, embedded engineering approach can scale economically as the company grows, particularly as competitors in the active observability market bolt AI features onto existing architectures rather than rebuilding from the ground up 1

1

.