Tsuga raises $35m Series A to kill per-byte observability pricing in AI-era clouds

Tsuga just raised $35m for its observability push, and it is aiming at a problem most cloud teams feel every month: telemetry costs that scale like a meter is running while your AI workload thinks. The Paris startup, founded by two former Datadog hands, is building observability software for the age of AI agents. Its bet is simple to say and hard to execute: end the per-byte pricing model that turns exploding data into exploding bills.

This $35m Series A arrives barely six months after Tsuga came out of stealth. That timing matters because it tells you the company is not just “building something cool.” It is moving fast enough to attract institutional capital, while the market is already straining under the economics of modern workloads. AI workloads are not just bigger apps. They behave differently. They generate more intermediate signals, more traces, more logs, more events, and more attempts. More attempts means more telemetry. More telemetry means higher usage-based charges if your tooling charges per byte.

To understand why Tsuga’s pitch resonates now, you have to look at what per-byte pricing does inside enterprises. Observability is supposed to reduce downtime, debug failures faster, and keep performance stable. But when your bill grows with raw data volume, telemetry becomes an expense you manage like cloud compute rather than an engineering practice you scale. Teams start making tradeoffs they should not have to make: sampling less, dropping logs, shortening retention, or trying to design around the tool’s billing mechanics. That is fine until your AI system starts failing in novel ways, because the failures you do not see are the ones that cost the most.

Tsuga’s “end per-byte pricing” stance is also a direct response to how AI changes the shape of observability. Traditional systems generate steady, predictable event streams. AI agents can be bursty and iterative. They may call tools repeatedly, generate traces across multiple steps, and produce large amounts of intermediate output that teams want to observe for safety, debugging, and performance tuning. Even when the ultimate user-facing action is a single “request,” the agent’s internal journey can be dozens of smaller actions. Telemetry volume can rise without the business impact rising at the same pace, and the cost model does not care why the bytes are there.

Now bring in the company’s origin story. Tsuga was founded by two former Datadog hands. That detail matters because it suggests the founders have seen the observability market from inside a major player. Datadog is a well-known benchmark for how observability is packaged and billed. A startup built by former Datadog operators carries institutional knowledge about what buyers expect, what usage-based models are buying you, and where customers run into friction. It is not proof of a strategy, but it is a credibility signal that the team understands the incentives on both sides: the vendor’s need to align revenue with value, and the customer’s need to predict costs.

The capital raise itself, a $35m Series A, indicates Tsuga is past the “stealth experiment” phase and into the “prove it commercially” phase. Coming out of stealth only six months ago usually means the market-facing groundwork and product iteration happened quickly. A Series A also typically implies more than product readiness. It suggests the company needs to scale go-to-market, deepen integrations, and build a pricing and packaging approach that can win against incumbents and alternatives that already sit in customer stacks.

There is also a broader governance angle worth flagging for decision-makers. Observability data can include sensitive operational details, sometimes even traces that correlate to user behavior indirectly through system interactions. As regulations and internal compliance programs become more strict about data handling, the cost and control of storing telemetry become more than just a finance question. A pricing model that pushes teams to reduce retention or sampling can undermine compliance objectives like auditability and incident investigation. In that environment, the ability to collect enough telemetry to meet operational and compliance needs without a runaway per-byte meter becomes strategically valuable.

Second-order, Tsuga’s move could pressure peers to revisit how they price observability for AI workloads. If teams start demanding predictable cost models aligned with “signals that matter” rather than raw volume, vendors may face pressure to evolve packaging, offer alternative billing units, or restructure what counts as billable telemetry. Even if Tsuga does not win every account, the argument it is making now can shift procurement criteria in the next budget cycle.

Strategically, Tsuga is positioning itself at the exact intersection where many organizations are stuck: they want deeper observability for AI systems, but the economics of per-byte pricing can make “more visibility” feel like a tradeoff against cost control. The $35m Series A, coming so soon after stealth, reads like an attempt to seize that moment. For executives, the question is whether their current observability spend is buying actionable insight or simply paying for bytes. Tsuga is trying to make that bill stop scaling with how much your AI system tries.