- Two former Datadog engineers, who previously sold their startup to Datadog, have secured $35 million in Series A funding. Their goal is to build an observability platform that keeps telemetry data inside each customer’s own cloud.
- Tsuga runs directly in the customer’s cloud account, so there’s no need to send data through a vendor’s systems. This setup avoids the per-byte fees that have made older observability platforms too expensive for today’s AI workloads.
- Just six months after launching publicly, Tsuga has already brought in several million dollars in revenue, with average contracts worth six figures. Its clients include Black Forest Labs and the French media group Le Monde.
Gabriel-James Safar and Sébastien Deprez sold their startup Madumbo to Datadog in 2019. Safar became a senior director of product management, while Deprez led the engineering team for Datadog’s fourth product. After leaving in 2023, they decided the business model they helped build was outdated. In 2024, they started Tsuga in Paris and have now raised $35 million in Series A funding to bring their new idea to life.
Singular led the Series A funding round, joined by General Catalyst and new investors DST Global Partners, QuantumLight, Databricks Ventures, and Picus Capital. With this round, Tsuga’s total funding reaches $45 million, following a $10 million seed round led by General Catalyst last November.
“The incumbents built good businesses on a model that no longer works. Every customer we speak to is paying more for observability than two years ago and getting less reliable coverage. We built Tsuga to end that,” says Safar, the company’s co-founder and CEO.
The global observability market made $4.1 billion in 2024 and is expected to grow to $18.1 billion by 2034, with a 16% annual growth rate. Still, the economy is getting worse. AI agent loops, automated deployments, and token interactions generate large volumes of telemetry data that older platforms can’t manage. As costs rise, engineering teams sample more data to save money, leading to gaps in coverage.
Why giving away your data is a problem
For years, observability companies have used the same approach: they collect customer telemetry, store it in their own cloud, and raise prices as data volumes grow. This worked in the past, but with AI, the model no longer makes sense.
Tsuga solves this problem with its design. Rather than moving telemetry to its own systems, Tsuga runs inside the customer’s cloud account on AWS, Azure, Google Cloud, and regional sovereign clouds. Data never leaves the customer’s environment.
Pricing is simple and transparent, with one rate per gigabyte. Tsuga’s engineers also help each customer’s team optimise their setup and reduce data usage, which saves money.
This setup gives customers a single view of all their data, including application and infrastructure traces, AI agent traces, prompt and token details, confidence scores, and agent call graphs. All of this stays inside the customer’s security perimeter and uses complete, unsampled data.
For Black Forest Labs, this means they can automate root cause analysis without missing information. For regulated companies like Le Monde, their observability data stays within legal boundaries.
“By staying out of the data path entirely, Tsuga removes every structural disadvantage the rest of the market is built on. The infrastructure tax disappears. The sampling compromise disappears. The AI governance gap disappears. What is left is a better platform at a lower cost, inside the customer’s own environment,” says Henri Tilloy, partner at Singular, which led the round.
Challenging Datadog from within
The founders’ time at Datadog is a key part of their story. Safar and Deprez built and sold Madumbo to Datadog, worked there for years, and then left to challenge the business model they once supported. Their inside experience gives them a unique understanding of the platform’s weaknesses.
Tsuga has tough competition. Datadog is worth over $30 billion, and both Dynatrace and New Relic are widely used by large companies and work well with older systems. Still, none of them can solve the cost and data sovereignty problems of the AI era without changing how they do business. This gives Tsuga a chance to stand out.
Alexandre Momeni, a partner at General Catalyst, which backed Tsuga at seed and invested again in this round, puts it simply: “Tsuga has the architectural scalability, flexibility, and distributed platform support that the coming generation of AI-native applications and agents require in their observability platform. This is a step-function change only possible because of their innovation across the full stack, from the storage layer to the business model.”
Databricks Ventures’ joining the round is important. Databricks supports the idea that enterprises should own their data and computing resources. Backing a platform that brings this idea to observability aligns with their strategy and gives Tsuga a partner with strong ties to large enterprise data teams.
Six months after coming out of stealth, Tsuga’s customers now include Le Monde, Camunda, Buk, and Black Forest Labs, with average contract values in the six figures. The new funding will help Tsuga grow its business and speed up work on its AI agent observability platform.
A big question is whether enterprise customers, accustomed to Datadog’s depth and Dynatrace’s wide range of features, will trust a young company with their deployments, even with experienced founders. Trust in infrastructure takes a long time to earn but can be lost quickly. Tsuga’s founders know this from their past experience and are now working to build that trust for their own company.
