A standard that makes collection cheap and portable does not make storage cheap. It moves the bill downstream. OpenTelemetry took the proprietary agent out of the path and standardised how telemetry gets emitted [1], and more emitted telemetry has to land somewhere. Bronto's Severin Neumann, who maintains OpenTelemetry and sits on its governance committee, says the instrumentation problem is fixed and the open question is what enterprises do with the volume, because data thrown away cannot be made sense of later by humans or agents [9].
The number worth pulling out of the New Stack interview is not the 20 to 30 percent of infrastructure spend that co-CEO Trevor Parsons dismisses as a minimizing framing of the cost [3]. It is the bar he sets for anyone claiming to fix it. Being "a couple of times cheaper or 50% cheaper" will not do, he says, given how fast volumes are growing [5]. Both of those thresholds describe roughly halving the price, so the floor he is setting sits above a 2x improvement [13]. That is an argument about compounding volume, and it applies to his own product as much as to the incumbents he accuses of optimising at the edges to protect margins [8].
The compromises he lists are all decisions made at write time. Retention windows get cut, data gets sampled, data gets rehydrated [4]. The retention figures he cites, three days against 30 [4], are a factor of ten apart in how far back a team can ask a question [12]. A sampling rule is a bet placed before the incident that produces the trace you want, and no amount of query sophistication recovers a span that was never written. That is the actual shape of the blind spot: not a missing dashboard but a missing row.
It is also the part of the argument to hold at arm's length. Every voice in the piece works for Bronto [11], a Dublin firm selling a data observability platform on the thesis that the next platform fight happens at the data layer rather than the dashboard layer [7]. The infrastructure-spend percentages arrive without independent sourcing, and the diagnosis fits the product neatly. What lifts it above positioning is that it is testable in a way dashboard claims are not: cost per queried gigabyte at a fixed retention window is measurable, and a team can run that comparison against its own invoice.
Co-CEO Noel Ruane's version of the pressure is that agents and applications emit more logs, traces and metrics every day while vendors offer workarounds and ask teams to live with the gaps [6]. Teams running agent workloads will hit that first, and they will hit it as a retention setting somebody chose months earlier for reasons that had nothing to do with debugging an agent.
