What is Datadog gross margin trajectory through 2028?

Direct Answer

Datadog has held subscription gross margin in the ~81-82% non-GAAP band through FY25, near the high end of large-cap infrastructure SaaS. Through FY28 we model 100-200bps of compression to a 79-81% range, with CFO David Obstler's publicly stated 80% long-term floor acting as the anchor. Three forces pull GM down: (1) AI inference cost passthrough from Bits AI and LLM Observability, (2) sovereign cloud + regional duplication as Datadog opens EU/APAC/GovCloud regions, and (3) hyperscaler infra costs scaling faster than blended pricing. Two forces protect the floor: (1) compute-architecture efficiency (ingestion-tier optimization, custom storage, query pushdown) and (2) multi-year customer commits that lock unit economics. Net: expect FY26 ~81%, FY27 80-81%, FY28 79.5-80.5% in the base case — a controlled glide, not a cliff.

What Datadog's GM Looked Like FY24-FY26

• FY24 subscription GM: ~82% non-GAAP, ~80% GAAP — peak efficiency post-COVID re-optimization wave
• FY25 subscription GM: ~81-82% non-GAAP, held flat despite Bits AI launch and 4 new region buildouts
• Q1 FY26 commentary (Obstler): AI inference cost pressure is real but offset by ongoing infrastructure efficiency work; reiterated 80% long-term floor
• GAAP-to-non-GAAP gap: ~150-200bps, driven by stock-based comp allocation and amortization of acquired tech (Eppo, Quickwit, Metaplane)
• Pomel discipline: Datadog has never traded GM for growth the way Snowflake or MongoDB have — culturally, GM is a board-level KPI, not a residual

What Compresses GM Through FY28

• Bits AI inference passthrough: Anthropic + OpenAI inference costs flow through COGS at near-pass-through margin until volume justifies dedicated capacity reservation pricing
• LLM Observability product: ingesting + analyzing token-level traces is structurally more expensive per GB than APM or logs
• Sovereign cloud + regional expansion: EU sovereign, GovCloud, APAC region duplication means Datadog runs ~3-5x the regional footprint vs FY22 with sub-scale utilization in newer regions
• AWS/Azure/GCP commit step-ups: as Datadog grows, hyperscaler enterprise discount tiers stop compounding linearly — diminishing returns on volume discounting
• International support + SE headcount in COGS: customer success and solutions engineering scale with international ARR, weighing on services GM
• Acquisition integration drag: Eppo, Quickwit, Metaplane carry sub-corporate-average GM in year 1-2 of integration

What Protects The GM Floor

• Compute-architecture efficiency: Datadog's infra team continues to extract 15-25% annual unit-cost reductions through ingestion-tier work, custom storage formats, and query pushdown — historically the single biggest GM defender
• Multi-year customer commits: ~75%+ of ARR on multi-year contracts locks revenue per unit while Datadog drives cost per unit down
• Bits AI value-add pricing: if Bits AI lands as a premium SKU rather than a feature giveaway, AI COGS becomes margin-accretive rather than dilutive (open question through FY26)
• Pomel + Obstler discipline: unlike growth-at-all-costs peers, Datadog leadership has publicly defended the 80% floor on every earnings call since the IPO
• Ingestion repricing levers: Datadog has historically repriced legacy SKUs (logs, custom metrics) when unit costs drift — pricing is a live lever, not a static input

The 3 Scenarios FY27-FY28

• Bear (~78% subscription GM): Bits AI lands as feature giveaway, sovereign cloud buildout outruns utilization, hyperscaler committed-spend discounts plateau, no offsetting repricing
• Base (~80% subscription GM): controlled compression, Bits AI partially monetized as premium SKU, infra efficiency offsets ~half of inference drag, regions reach utilization on schedule
• Bull (~82% subscription GM): Bits AI monetizes cleanly as platform-wide upsell, regional utilization hits target by FY27, custom inference infrastructure (dedicated Anthropic capacity, in-house inference where viable) cuts AI COGS 30%+

What Investors Should Watch

• Subscription GM disclosure each quarter — Datadog breaks subscription GM separately; watch the Q-over-Q delta, not the headline
• Obstler's prepared-remarks language on AI cost — shifts from "managing" to "pressure on" signal real compression
• Bits AI revenue disclosure — if Datadog starts breaking out Bits AI ARR, that is a sign the SKU is monetizing
• Regional revenue mix — international ARR growth ahead of US ARR signals regional utilization improving
• CapEx + infrastructure commitments in the 10-Q cash flow statement — leading indicator of regional buildout pace
• Mention of dedicated inference capacity (Anthropic enterprise, Azure OpenAI committed capacity) — signals AI COGS structural improvement

Comparable GM Trajectory

• ServiceNow (~83% subscription GM): higher floor, lower COGS variability — workflow SaaS does not carry ingestion COGS the way observability does
• Snowflake (~74-77% product GM): structurally lower because storage + compute pass-through dominates COGS; useful as the *floor* comparable for inference-heavy SaaS
• MongoDB (~73-75% subscription GM): Atlas hosting drag — Datadog should *not* trend toward MongoDB territory unless AI inference cost is mismanaged
• Cloudflare (~77-78% GM): edge network capex weighs on GM — different model, but useful comp for "GM under expansion pressure"
• Datadog's relative position: sits between ServiceNow (workflow ceiling) and Snowflake (data-platform floor) — and should *stay* in the upper half of that band through FY28

GM Trajectory Table

GM Compression Drivers

Bottom Line

Datadog's GM trajectory through FY28 is a controlled glide from 82% to 79-81%, anchored by Obstler's public 80% floor and protected by best-in-class compute efficiency. The single biggest swing factor is whether Bits AI monetizes as a premium SKU or ships as a feature giveaway — that one call probably moves FY28 GM by 200bps. Watch the subscription GM disclosure and Obstler's earnings-call language quarter by quarter; both are leading indicators ahead of the print.

Related: [q1671](/knowledge.html?id=q1671) (Datadog AI cost structure), [q1681](/knowledge.html?id=q1681) (Datadog sovereign cloud economics), [q1686](/knowledge.html?id=q1686) (Datadog Bits AI monetization model).