Fixing Observability Pricing for the AI Era

#observability #pricing #devops #ai

Authored by Trevor Parsons

Today, we're announcing new pricing for Bronto.

TL;DR:

$0.10 per GB ingested, $1 per TB searched — for any signal: logs, traces, or metrics. No per-host or per-metric charges. Retained for 12 months, always hot, with sub-second search.
It's designed to be disruptive, simple, predictable, and to align cost with customer value. It's over 100x more efficient than legacy pricing.

We're addressing a foundational issue our industry has failed to tackle: current pricing models aren't fit for purpose. They're out of whack from an overall cost perspective, don't align cost with value, and are designed to favor the vendor, not the customer. They're also hard to understand and wildly unpredictable, with teams regularly getting hit with nasty overage surprises.

Cost has been the biggest issue for observability customers for the past 15+ years, and vendors continue to willfully ignore it.

Observability costs routinely run at 20–30% of total infra spend. That leads to a familiar set of workarounds: dropping high-volume logs, cutting retention to 3, 7, or 15 days, sampling, archiving and rehydrating, building pipelines just to throw data away before it lands, or rolling your own observability on open source and inheriting all the management overhead.

One example that stuck with us: a company resized its hosts onto bigger AWS instances just to fight per-host pricing — even though that wasn't the right architecture for their system. They changed their infrastructure to suit their observability bill. That's how distorted this pricing has become.

Instead of tackling the problem head-on, vendors keep adding "features" on top of datastores that aren't fit for purpose. The latest wave is AI agents and automated workflows. Those capabilities are genuinely powerful and will help teams manage complex systems, reduce MTTR, and improve root-cause analysis — but if you build AI capabilities on top of fundamentally broken foundations, the cost problem only gets worse, especially as AI workloads drive even higher data volumes.

Legacy Pricing Is Broken

There are two core problems with legacy pricing.

First, it's roughly 10x too expensive no matter how you slice it. Vendors charge dollars per GB ingested and stored, when per-GB pricing needs to be at the level of cents per GB — so teams stop architecting around the cost. Observability spend should drop from ~30% of infra spend to under 5%, low enough that you stop engineering around it. On top of that, dollar-per-GB pricing is typically for only days of hot retention, meaning you're paying dearly for access to a sliver of your data — again, at least an order of magnitude off for an AI world where historical analysis over months or years should be the norm.

Second, it's a value problem. You pay to ingest and store, so the vendor gets paid whether or not you ever get value from the data. You may never log in, search, or add an alert — the vendor still gets paid. In logging especially, people describe their provider as an expensive datastore they never actually use. The model was built for the vendor, not the customer.

Enter Bronto — Pricing Built for the Customer, Not the Vendor

Built on BrontoDB, our custom-built observability datastore, Bronto drives massive efficiency in ingesting, storing, and analyzing observability data.

Bronto pricing: cents per GB, not dollars per GB.

$0.10 per GB ingested — any signal, logs, traces, or metrics. No per-host or per-metric charges. Retained for 12 months, always hot, with sub-second search.
$1 per TB searched — 5x cheaper than scanning the same data on AWS Athena, which runs $5 per TB.

If you ingest data and never search it, you pay very little. You pay more only when you get more value by searching across more data — incentives that line up with yours, not against them.

For enterprise plans, typical DevOps usage at scale comes in under a combined ingest-and-search cost of about $0.20 per GB (roughly $0.10 for ingest, ~$0.10 for search at the $1/TB rate). A free trial lets you verify exactly where you'd land with your own data.

100x–1000x More Efficient

Bronto's entry plan is built for startups, solo builders, and teams building something new: $25/month for 1TB ingested with 20x search, at 12-month retention — roughly $0.025 per GB for any signal, with no per-host costs.

Datadog runs about $2.60 per GB for 30-day retention. A team ingesting 1TB/month might pay around $2,600 with Datadog for 30 days of retention versus $25 with Bronto — 100x cheaper, with over 10x the retention on top, which works out to roughly 1000x more efficient. That's before even factoring in cost explosions from things like high-cardinality metrics.

At larger volumes, Bronto's $0.10/GB ingested plus $1/TB searched comes out to around $0.20 per GB all-in for a typical DevOps profile, with 12 months of retention — versus Datadog's $2.60 per GB at 30 days (assuming ~1KB events and 100% log indexing). Roughly 100x more efficient.

Simple. Predictable. No surprises.

Simple — one per-GB cost for any signal ingested, one per-TB cost for what you search, with 12-month retention by default.
Predictable — entry plans have generous built-in headroom; larger plans get a usage dashboard or direct support from the team.
No surprises — built-in usage tracking and alerting mean no end-of-month shocks, and data never stops flowing when you hit a limit — you'll just get a heads-up.

What Bronto Delivers

All your data in one place, with full coverage and no blind spots. Always-hot data at 12-month default retention. Seamless cross-correlation across logs, traces, and metrics without hopping between separate datastores (think Prometheus, Tempo, and Loki, each limited and disjointed in its own way). And AI on top, via Bronto's MCP server, Bronto Vibe, or Bronto's built-in investigation capabilities.