type: pattern tags: [cloud-infrastructure, ai-inference, gross-margin, mix-shift, structural-compression, gpu, traditional-cloud] confidence: medium created: 2026-04-01 source: DOCN stock-analysis 2026-04 persona: bear provenance: legacy source_analysis_path: null source_paragraph_quote: null source_transcript_span: null source_loss_log_path: null

AI Inference Mix Shift Is Structurally Margin-Dilutive for Traditional Cloud Providers

When a traditional cloud provider (compute/storage/networking) grows AI inference revenue as a share of total revenue, blended gross margins compress structurally. The mechanism: GPU-based inference services carry lower margins than established core cloud services (CPU compute, object storage, networking), because (1) GPU hardware is more expensive per dollar of revenue than legacy server infrastructure, and (2) inference pricing commoditizes rapidly with 30+ providers competing on benchmark performance. Even "full-stack" managed inference (inference APIs, serverless, GPU droplets) runs at lower margins than the provider's own legacy cloud services. This is distinct from the bare-metal vs. full-stack comparison (where full-stack is the winner): within a single provider's P&L, AI inference dilutes the blended gross margin as its revenue share grows.

Evidence

• DOCN FY22 → Q4 FY25: GAAP gross margin compressed from ~63% to 58.7% (-400bps over three years), tracking the rise of AI workloads in the revenue mix.
• DOCN CFO explicitly confirmed in Q4 FY25 earnings: "AI margins are lower than core cloud margins" — management disclosure, not a model assumption.
• DOCN AI ARR $120M growing 150% YoY in Q4 FY25 (vs. total ARR $970M at 18% YoY). AI becoming a fast-growing share ensures the mix headwind intensifies into FY26-FY27.
• Management guided FY26 adjusted EBITDA margin of 36-38%, down from 41.6% in FY25 — the margin compression year is the capacity ramp into AI, not legacy business deterioration.
• Contrast with the existing ai-inference-revenue-per-mw-managed-services-attach pattern: DOCN still earns ~2× revenue/MW vs. pure bare-metal operators, and EBITDA margins remain ~40%+. The dilution is real but bounded — this is not a race-to-zero, it is a structural ceiling lower than the legacy cloud business.

Implication

When analyzing any traditional cloud provider pivoting to AI inference, treat AI revenue growth as a gross margin headwind by default, unless management explicitly demonstrates AI-specific margins above their own core cloud margins. The watch signal is CFO disclosure: "AI margins are [lower/higher/in line with] core margins." Track AI revenue as a percentage of total alongside gross margin trajectory — a company growing AI 100%+ while GAAP gross margins decline 100-150bps/year per point of AI mix growth is experiencing structural, not temporary, compression. This pattern applies regardless of how strong the managed-services attach ratio is (even at 70% managed attach, DOCN sees this headwind). Do not assume AI pivot = margin expansion in the cloud infrastructure sector.