Datadog Bits AI vs Dynatrace Davis AI vs New Relic AI: Real AI Observability Cost in 2026
Short answer
None of the three vendors appears to charge separately for its AI capability specifically — Bits AI, Davis AI, and New Relic's AI features are all bundled into the base observability platform rather than sold as a distinct, separately metered add-on in available public pricing information. This answers the category's core question: the "extra" cost of AI here is effectively baked into which platform and tier a company is already paying for, not a separate line item layered on top. What genuinely differs, and differs enormously, is the underlying platform cost itself: Datadog bills per host per product with separate charges for nearly every capability, Dynatrace bills per Host Unit at a reported 15–25% premium over Datadog but with no per-seat fee at all, and New Relic bills primarily on data ingest with per-user or per-concurrent-user-hour seat charges that are reported to make it 3–5x cheaper than Datadog at typical scale — but genuinely expensive for a large team specifically because of how its user-based pricing compounds.
Pricing at a glance
| Datadog (+ Bits AI) | Dynatrace (+ Davis AI) | New Relic (+ New Relic AI) | |
|---|---|---|---|
| Public rate card | Yes — modular, per-product pricing published | Yes — Host Unit-based pricing published, alongside a separate consumption-based (logs/traces/metrics per GB) pricing surface | Yes — ingest-based and per-user pricing published |
| Billing basis | Per host, per product — separate charges for Infrastructure, APM, Logs, RUM, Synthetics, and Security, plus per-user fees for full platform access | Per Host Unit (HU), scaled by monitored-host memory size; a separate usage-based surface bills logs, traces, and metrics per GB with no per-host or per-user fee shown | Per GB of data ingested, plus per-user or per-concurrent-user-hour seat fees; unlimited hosts and agents included regardless of count |
| Reported core rates | Infrastructure: $15–40/host/month depending on tier. APM: $23–40/host/month (Pro vs. Enterprise). A comparable combined stack: $46–58/host/month before log or APM overages | Full-Stack Monitoring: $69–74/Host Unit/month at list. Infrastructure-only (no APM): $21–25/Host Unit/month. Separately, usage-based logs at $0.42/GB, traces at $0.16/GB, metrics at $0.05/GB | Ingest: $0.30/GB (Standard) to $0.55/GB (Data Plus). User seats: $49–349+/month (Pro/Full Platform tiers), or billed as concurrent user-hours at roughly $0.25/hour in one reported framing |
| Free tier | None for production use | Not confirmed as a distinct free tier in available sources | Yes — 100 GB/month genuinely free, reported as a real, usable allowance for small teams |
| AI feature cost | Not separately itemized — bundled into the base platform tiers in the sources reviewed | Not separately itemized — bundled into the base platform, now positioned under Dynatrace's broader "Dynatrace Intelligence" layer | Not separately itemized in available sources |
| Reported relative positioning | Positioned as "developer-friendly all-in-one," with the broadest integration catalog and most mature interface of the three | Reported 15–25% more expensive than Datadog at equivalent coverage, positioned as the premium, most automated option | Reported 3–5x cheaper than Datadog at typical enterprise scale (100 hosts), positioned as "the highest-value option" |
| Structural cost risk | Costs spike with custom-metric cardinality and the sheer number of separately-billed products a deployment activates | Per-host-hour and per-GiB-hour billing scales sharply with host memory size — a 32 GiB host costs roughly 4x an 8 GiB host under this model | Costs spike specifically with user count under the "Data + User" model — a real structural risk for organizations with large engineering teams needing broad platform access |
What headline pricing excludes
None of the three vendors' AI features (Bits AI, Davis AI, New Relic AI) appear to carry a separate, itemized price in available public pricing information — this is itself the most useful finding for anyone specifically budgeting "how much does AI cost on top of observability." The honest answer, based on available evidence, is that it doesn't cost anything additional beyond whichever base platform tier already includes it — the real cost driver in this category is the underlying host, ingest, and user pricing, not the AI layer riding on top of it.
Datadog's true cost for a comparable full-stack deployment is reported at $46–58/host/month "before log or APM overages" — a real, explicit caveat that the headline per-product rates don't represent a deployment's total cost, since logs and APM overage charges are separate, additional line items on top.
Dynatrace's per-host pricing scales with host memory size, not just host count — a detail easy to miss when comparing "per host" rates across vendors, since a fleet of larger-memory hosts costs meaningfully more under Dynatrace's model (reportedly up to 4x more for a 32 GiB host versus an 8 GiB host) than an equivalent host count would suggest from the headline rate alone.
New Relic's per-user or per-concurrent-user-hour pricing means its "unlimited hosts" advantage is genuinely offset by a real, separate cost risk tied to team size. One detailed, specific illustration: a 100-person engineering team on New Relic's Pro tier would pay $34,900/month in user fees alone — described as "more than many organizations' entire Dynatrace commitment." A company evaluating New Relic specifically because of its unlimited-host model should model its actual expected user-seat count with equal care.
Hidden costs
- Datadog has historically raised per-host and per-span pricing annually by 5–10%, per one detailed procurement-focused source — a real, recurring cost-drift risk that should be negotiated with a cap at initial contract signing.
- Dynatrace's Davis Platform Subscription (DPS) consumption model is reported to create "cost escalation dynamics that catch enterprises off guard at renewal" — a genuine, named risk specific to Dynatrace's newer billing architecture (migrated from per-host/DDU licensing between 2022 and 2024).
- Dynatrace's own sales team is reported to specifically emphasize Davis AI's automated causal analysis, positioning Datadog as "developer tools, not enterprise observability" rather than a true peer competitor in large enterprise (Fortune 500) negotiations — a real positioning dynamic worth knowing when using Datadog as negotiating leverage against Dynatrace, since Dynatrace is reported to hold firmer against Datadog-based competitive pressure than against New Relic-based pressure specifically.
- Realistic negotiated discounts differ meaningfully by vendor: 25–35% off list is reported achievable for both Datadog and New Relic, while Dynatrace's realistic discount is reported somewhat lower at 20–30%, though a documented competitive evaluation against Datadog or New Relic is reported to push Dynatrace discounts as high as 25–40% off list DPS rates in competitive renewal situations specifically.
- Davis AI's root-cause analysis is reported to have a genuine failure mode worth knowing about: when Dynatrace's underlying topology data is stale or incorrect, Davis AI is described as capable of "confidently" pointing at the wrong service — described as potentially "worse than no AI at all" in that specific failure scenario, a caveat about capability rather than cost, but one that affects whether Dynatrace's reported price premium is actually justified for a given environment's data quality.
Worked scenarios
Because New Relic's core billing driver (ingest volume plus user seats) doesn't scale by host count the way Datadog's and Dynatrace's pricing does, this article does not force New Relic into the same per-host framework used for the other two — doing so would misrepresent its actual cost structure.
25 hosts
| Estimated monthly cost | |
|---|---|
| Datadog (comparable full-stack rate) | 25 × $46–58 = $1,150–1,450/month, before log or APM overage |
| Dynatrace (Full-Stack Monitoring) | 25 × $69–74 = $1,725–1,850/month, consistent with its reported 15–25% premium over Datadog |
| New Relic | Not meaningfully expressed as a per-host figure — cost depends on actual data ingest volume and the number of engineers needing seat access, both of which are independent of the 25-host count itself |
At small scale, a small engineering team (say, 5 engineers) using New Relic could plausibly stay within or close to its free 100 GB/month tier for ingest, making user-seat cost the dominant factor rather than host count.
250 hosts
| Estimated monthly cost | |
|---|---|
| Datadog (comparable full-stack rate) | 250 × $46–58 = $11,500–14,500/month, before overage |
| Dynatrace (Full-Stack Monitoring) | 250 × $69–74 = $17,250–18,500/month |
| New Relic | Still not a direct per-host calculation; a reference point from a comparable analysis puts a 100-host environment's New Relic cost at roughly $15,000–30,000/year total (ingest plus users) versus $80,000–120,000/year for a comparable Datadog deployment at that same 100-host reference — a reported 3–5x gap in New Relic's favor at that specific scale |
The 100-host reference point above illustrates the scale of the reported gap, though it should not be linearly extrapolated to 250 hosts, since New Relic's cost driver (ingest and users) doesn't move in lockstep with host count the way Datadog's does.
2,500 hosts
| Estimated monthly cost | |
|---|---|
| Datadog (comparable full-stack rate, before negotiated discount) | 2,500 × $46–58 = $115,000–145,000/month, reducible by a reported 25–35% with negotiation |
| Dynatrace (Full-Stack Monitoring, before negotiated discount) | 2,500 × $69–74 = $172,500–185,000/month, reducible by a reported 20–40% depending on competitive pressure applied during negotiation |
| New Relic | At this scale, New Relic's genuine structural advantage (unlimited hosts included) becomes most pronounced — its cost is driven by total ingest volume across the fleet and by user-seat count, not multiplied by host count the way the other two are, though a large engineering team accessing the platform would need to model the user-fee risk explicitly: a 100-person team alone is reported at $34,900/month in user fees under the Pro tier, a cost that exists independent of host count entirely |
Normalizing across billing units
Datadog's per-host-per-product pricing and Dynatrace's per-Host-Unit pricing are similar enough in structure to compare directly, provided a buyer accounts for Dynatrace's memory-based scaling within the Host Unit itself. New Relic's ingest-plus-user model measures something fundamentally different — it doesn't charge more as host count grows, but does charge more as either data volume or team size grows, meaning a fair comparison requires modeling a specific organization's actual ingest volume and expected user-seat count against New Relic's rates, rather than trying to force a per-host equivalent that its billing model doesn't actually use.
Break-even and crossover
The clearest, most calculable crossover in this category involves New Relic's own user-based pricing risk rather than a cross-vendor host-count comparison: New Relic's unlimited-host advantage is a genuine structural benefit for an organization with a large infrastructure footprint but a small engineering team, but becomes a genuine structural disadvantage for an organization with a large team needing broad platform access. One detailed source frames this precisely: for organizations needing observability access for 50 or more engineers, Dynatrace's zero per-seat cost becomes "a genuine structural advantage" specifically because New Relic's per-user model at that team size can exceed what "many organizations' entire Dynatrace commitment" would cost. This is a real, quantifiable crossover — it just depends on team size rather than host count, which is not the axis most observability pricing comparisons default to examining.
Who pays more, and when
- An organization with a large host count but a small engineering team accessing the platform benefits most from New Relic's ingest-and-user model or Dynatrace's no-per-seat structure, and is likely to find Datadog's per-host-per-product billing the least favorable of the three at this specific profile.
- An organization with a large engineering team (50+ people) needing broad observability access should weight Dynatrace's zero per-seat cost heavily, despite its reported 15–25% premium over Datadog on pure host-based pricing, since New Relic's per-user fees are reported to scale into a genuinely large, sometimes prohibitive cost at this team size.
- An organization prioritizing cost predictability above all else is reported to find New Relic's ingest-based model, and Dynatrace's per-host model, both easier to forecast than Datadog's per-product-per-host structure, which is described as "the most complex of the three" given how many separate products it bills independently.
- An organization running complex, high-scale microservices environments where automated root-cause analysis has high value may find Dynatrace's Davis AI and Smartscape topology mapping worth its reported cost premium — provided the underlying topology data stays current, since stale data is reported to make Davis AI's automated conclusions actively misleading rather than merely unhelpful.
- Any organization negotiating with any of the three should bring a genuine competing evaluation into the conversation — reported discount ranges of 25–35% (Datadog, New Relic) and 20–40% (Dynatrace, with the higher end specifically tied to documented competitive evaluations) make this the single most effective lever across the category, more so than any specific published rate.
Limitations and uncertainty
All three vendors publish pricing for their core observability platforms, though available pricing reports show some spread in specific rates (Datadog APM at $23–40/host/month depending on tier and source; Dynatrace Full-Stack Monitoring at $69–74/Host Unit/month, alongside a separately described usage-based logs/traces/metrics pricing surface that may represent a different or evolving billing option rather than a direct alternative to the Host Unit model). None of the three vendors' AI capabilities (Bits AI, Davis AI, New Relic AI) appears to carry a separately published or reported price distinct from the base platform tier. The reported 3–5x cost gap between New Relic and Datadog, and the $34,900/month New Relic user-fee illustration, both come from specific detailed comparative sources and should be read as illustrative reference points for the scale of the pattern, not universal figures applicable to every deployment.
Official sources
- datadoghq.com/pricing (confirmed official modular pricing)
- dynatrace.com/pricing (confirmed official Host Unit and usage-based pricing)
- newrelic.com/pricing (confirmed official ingest-and-user-based pricing)