DigitalOcean: The Inference Cloud Thesis Has Still Not Priced In

DigitalOcean reported Q4 2025 revenue of $901M (+15% YoY) and net income of $259M (29% margin), with $120M AI ARR (70% from inference and core cloud). Analyst maintains a Buy and $105 price target, citing a strategic pivot to usage‑based AI inference, 0% churn in million‑dollar cohorts, and rapid expansion in the $500K and $100K customer tiers supporting sustainable growth.

Analysis

DigitalOcean’s pivot to usage-based inference turns the company into a low-friction access point for enterprises that want operational LLMs without hyperscaler lock-in. That dynamic creates two durable optionalities: (1) higher gross-margin revenue from inference workloads as utilization and model multiplexing improve, and (2) a distribution wedge for third-party model vendors and MLOps tooling who will prefer a neutral, developer-friendly host over platform-tied clouds. Expect mix-driven margin improvement to compound over 12–36 months as average revenue per customer rises without linear increases in capex if GPU utilization reaches multi-tenant thresholds.

The supply chain lever is Nvidia and emerging lower-cost accelerator options. If DigitalOcean scales inference meaningfully, it will amplify GPU spot-market tightness during procurement cycles, pushing hyperscalers and chip vendors to reprice capacity or prioritize long-term contracts. Conversely, rapid commoditization via model quantization and inference-specialized silicon (2–4 years) could cap pricing power; the near-term risk is execution — getting to multi-tenant, low-latency inference at scale is an ops problem as much as a sales one.

A bullish path is not binary: the clearest reversal would be hyperscaler-led price aggression or an enterprise pivot back to on-prem inference for latency/security reasons, both of which could compress multiple and growth expectations within 6–18 months. The most underappreciated upside is M&A optionality — a hyperscaler or chip vendor could find owning a developer-centric inference distribution channel attractive, creating a takeover-premium scenario within 12–36 months if growth and retention metrics remain pristine.

Key Decisions for Investors

• Long DOCN equity (ticker: DOCN): build position in two tranches—50% now, 50% on a 10–15% pullback. Timeframe 12–18 months; target asymmetric upside of ~35–50% with a hard stop at ~18% below entry to control execution risk.
• DOCN option collar to express leveraged upside: buy 12–18 month (Jan–Dec 2027) 30% OTM calls while selling 3-month ATM calls to fund half the premium. Purpose: capture multi-quarter adoption while protecting cost; worst-case loss limited to premium paid if thesis fails in 12–18 months.
• Macro/infra play: long Nvidia (NVDA) via 6–12 month call spread to capture incremental GPU demand driven by inference growth at emerging clouds. Use a modest notional (10–15% of DOCN equity exposure) and expect semiconductor cyclicity risk—take profits on >40% move.
• Relative-value pair: long DOCN / short AMZN or GOOGL (equal notional) to express outperformance of neutral, developer-focused inference delivery vs hyperscaler incumbents. Timeframe 12 months; peel exposure if hyperscalers announce large unconditional price cuts or aggressive free-tier inference initiatives.