How does Datadog ARPU change post-AI agent rollout?

Direct Answer

Datadog's blended ARPU sits near ~$113K/customer today (~$3.4B subscription revenue / ~30K customers, FY25 anchors), and the math says it should rise 15-25% by FY27 as Bits AI consumption, Cloud SIEM cross-sell, and LLM Observability stack new revenue lines on top of the existing per-host APM envelope. The 3 forces lifting ARPU: (1) Bits AI per-investigation consumption adds metered revenue above bundled allowance, (2) Cloud SIEM cross-sells into the ~3,800 $100K+ infra customers who don't yet pay for security, (3) LLM Observability adds a brand-new revenue line per AI-workload customer that didn't exist in FY24. The 2 forces dragging: Datadog for Startups dilutes the customer count denominator with low-ARPU logos, and Microsoft (Azure Monitor + Sentinel + Copilot) compresses pricing at the SMB end. Net read: blended ARPU moves from ~$113K toward ~$130-140K by FY27, with the $1M+ club ($3-5M average, ~340 customers) leading the lift. All figures are estimates derived from public reporting, not Datadog disclosures.

The ARPU Math Today

• Blended ARPU — ~$3.4B FY25 subscription revenue / ~30,000 customers = ~$113K/customer/year (estimate from Q4 FY25 earnings + 10-K customer count)
• $1M+ club — ~340 customers, average $3-5M ARR each = ~$1.0-1.7B ARR concentrated in <2% of logos (drives ~30-50% of total revenue)
• $100K+ club — ~3,800 customers, average $300-500K ARR each = ~$1.1-1.9B ARR concentrated in ~13% of logos
• Long tail — ~26K customers below $100K ARR average ~$15-25K each, contributing the remaining ~$400-650M
• NRR anchor — 115% net revenue retention means the existing book grows ~15%/year *before* new logo wins, and that compounding is where ARPU expansion shows up

The 3 Forces Lifting ARPU FY26-FY27

• Bits AI per-investigation consumption — bundled at launch with caps, then metered overage at ~$0.50-$2.00/investigation; heavy users (the $1M+ club running thousands of incidents/month) generate $50-200K/year incremental ARR per logo on top of APM spend
• Cloud SIEM cross-sell — only ~30% of Datadog infra customers currently buy Cloud SIEM; closing that gap into the $100K+ club at ~$0.20/GB ingested + per-detection adds $80-300K/year per cross-sold logo, lifting cohort ARPU 20-40%
• LLM Observability — a brand-new SKU launched FY25 priced per-trace + per-token-monitored; every customer building production AI workloads becomes a buyer, adding $30-150K/year per AI-workload logo and creating a third growth vector beyond infra + security

The 2 Forces Dragging ARPU

• Datadog for Startups dilution — the free/discounted startup program adds thousands of low-ARPU logos to the denominator; great for top-of-funnel and developer mindshare, but mathematically drags blended ARPU down 3-5%
• Microsoft compression at SMB — Azure Monitor + Sentinel + Copilot bundled with Azure consumption credits pulls SMB and mid-market price-sensitive customers toward "good enough" alternatives, capping per-host pricing power at the bottom of the market
• Multi-product discount stacking — as customers buy 4+ products (the strategic goal), bundle discounts grow — protects retention but trims per-product realized ARPU
• Inference COGS pass-through resistance — customers push back on consumption pricing for AI features, forcing Datadog to absorb more inference cost in bundled tiers and limiting the metered-overage upside
• Currency + macro — non-USD revenue (~30% of total) translates softer in a strong-dollar environment, and any 2026 IT-budget tightening compresses expansion deals

Customer Cohort ARPU

• $1M+ club (~340 customers) — FY25 average $3-5M, FY27 target $4-6.5M (Bits AI overage + Cloud SIEM + LLM Obs all stack here; biggest absolute dollar lift)
• $100K+ club (~3,800 customers) — FY25 average $300-500K, FY27 target $400-650K (Cloud SIEM cross-sell is the dominant lift; LLM Obs adoption emerging)
• Long tail (~26K customers) — FY25 average $15-25K, FY27 target $18-30K (modest lift from APM expansion; Bits AI bundled but rarely overages)
• Startup program logos — FY25 average <$5K, structurally low; graduation rate to paid tiers is the metric that matters, not ARPU
• New logo cohort FY26-27 — landing larger as multi-product platform pitch matures; new-logo ARPU likely $80-120K average, well above blended

What Comparable SaaS Companies Are Doing

• Snowflake — consumption pricing means AI-driven ARPU lift is automatic when customers run Cortex AI workloads; FY25 saw average customer consumption rise as AI use cases (RAG, embedding pipelines) added to the warehouse spend without new SKUs
• ServiceNow — Now Assist Pro Plus uplift (60% premium over Pro) drove ARPU expansion in FY25; the bundled-uplift model is what Datadog is studying for Bits AI Premium tiering
• MongoDB — Atlas Vector Search added a new revenue line per AI-workload customer, lifting Atlas ARPU as customers built RAG/embedding apps on the same cluster; same playbook Datadog is running with LLM Obs
• Palo Alto Networks — "platformization" strategy bundles XSIAM + Prisma + Cortex AI for enterprise customers, lifting ARPU through cross-product attach rather than per-product price hikes
• CrowdStrike — Charlotte AI bundled into Falcon platform tiers; ARPU lift comes from module attach (8+ modules per customer goal) rather than standalone AI SKU pricing

What Pomel Should Optimize For

• Module attach rate, not per-product ARPU — the $1M+ club already buys 8+ products; the lever is getting the $100K+ club from 4 products to 6+ (Cloud SIEM + LLM Obs are the two highest-leverage adds)
• Bits AI bundled cap calibration — set caps high enough to drive adoption stickiness, low enough that the $1M+ club generates metered overage; the tail rarely hits the cap and stays happy
• LLM Observability as land-and-expand wedge — every AI-workload customer is a future $1M+ club member; price low to land, expand into APM + Cloud SIEM + Bits AI from there
• Multi-year commit + inference credit pools — locks the per-host envelope before AI substitution can compress it, while giving customers a reason to commit to consumption pricing
• Startup program discipline — track graduation rate (free → paid) as the leading indicator; if graduation stays >25%/year, dilution is acceptable because the cohort eventually expands
• Net revenue retention as the headline metric — lifting NRR from ~115% toward 120-125% does more for ARPU math than any single SKU price increase, and it's the number Wall Street rewards

Customer Cohort ARPU Trajectory

ARPU Drivers Map

Bottom Line

Datadog's ARPU story for FY26-27 is a mix-shift story, not a price-hike story. Bits AI consumption, Cloud SIEM cross-sell, and LLM Observability each add revenue lines on top of the existing per-host APM envelope, lifting blended ARPU from ~$113K toward ~$130-140K by FY27 — a 15-25% expansion concentrated in the $1M+ and $100K+ cohorts where multi-product attach drives the math. The drag from Startups dilution and Microsoft SMB compression is real but mathematically smaller than the lift. Pomel's job is to keep module attach rates climbing in the upper cohorts while protecting the long tail from tier-down. All figures are estimates derived from public Datadog reporting, comparable SaaS disclosures, and industry analyst frameworks. See also: [q1673 Datadog cohort revenue concentration](/library/?q=q1673), [q1681 Datadog AI inference COGS structure](/library/?q=q1681), [q1691 Datadog Bits AI pricing without cannibalization](/library/?q=q1691).