Observability has a data problem. AI is about to make it worse.

OpenTelemetry fixed one big part of observability: collecting the data in a standard way. But that win exposed a messier problem underneath. Teams can now gather more logs, traces, and metrics than before, and they still do not have a cheap, practical way to keep, search, and analyze all of it. The result is a lot of expensive blind spots.

Bronto, a Dublin company, thinks that is where the next fight will be won. Trevor Parsons, the firm’s co-founder and co-CEO, says the industry has spent too long polishing the edges instead of rebuilding the economics of telemetry storage. In his view, a small discount or a cleaner dashboard does not solve anything when data volumes keep climbing and budgets are already strained.

The pain shows up in familiar ways. Customers are told to shorten retention from three days to seven days to 30 days, sample their data, or rehydrate it later if they need the full picture. Parsons says that leaves teams paying heavily for access to only a slice of what they generate. Noel Ruane, Bronto’s other co-founder and co-CEO, says vendors have leaned on workarounds and bolted-on features while asking customers to live with blind spots.

OpenTelemetry made the collection problem smaller, but it also made the storage problem impossible to ignore. Severin Neumann, Bronto’s head of community, who also helps maintain OpenTelemetry and sits on its governance committee, says enterprises now need affordable places to put all that telemetry. If they throw away large chunks of it, humans and agents alike lose the ability to make sense of what happened.

Bronto’s answer is a custom data store called BrontoD. The company says it can hold more than 100 times the observability data teams keep now without making the system slower or harder to use. It also says it can handle high-cardinality metrics without the usual cost blowups, make log search sub-second across petabytes, and remove the need to sample traces. Billing is tied more closely to search and analysis than to data sitting idle in storage.

AI is what raises the stakes. Parsons says AI systems produce more telemetry, especially because they are non-deterministic and trace-heavy, and they need longer retention for debugging. At the same time, he argues, the same data gets more valuable because agents can sift through far more of it than any SRE team ever could by hand. The next observability battle is looking less like a dashboard contest and more like a storage bill with teeth.