United Internet 2025 revenue rises but falls short of forecasts By Investing.com

United Internet reported fiscal 2025 revenue of €6.10B, missing consensus of €6.24B but up 1.9% year‑over‑year; adjusted EPS rose driven mainly by lower tax expenses. The company added 700,000 fee‑based customer contracts and posted EBITDA of €1.28B, while higher depreciation reflects mobile and fiber network investments. For 2026 it guides sales of €6.25B, EBITDA ~€1.45B and cash capex of €600–650M.

Analysis

The real directional lever here is infrastructure-led demand for latency-sensitive compute rather than a simple ARPU/subscriber story. Fiber and dense mobile coverage upgrades materially shift workload mix toward distributed inference and bursty training — a profile that disproportionately favors high-GPU-density chassis and GPU supply (NVDA exposure) over commodity x86 refreshes. Expect a 12–36 month replacement/upsize cycle as operators phase in edge POPs and on-prem clusters to monetize low-latency services.

Second-order winners are server OEMs that tightly integrate multi‑GPU systems and provide rapid fulfillment (SMCI‑style models), plus optical/pluggable suppliers that bridge networks to compute nodes. The supply chain risk is not silicon scarcity per se but power delivery, thermal solutions, and NIC/ASIC availability; these are 3–9 month chokepoints that can throttle deployments even when purchase intent is strong. Conversely, advertising-dependent app monetization players face asymmetric downside if mobile throughput gains don’t translate to higher ad yields — ad budgets remain the marginal cut in a slowing macro.

Key catalysts and tail risks: near-term swings will be driven by guidance language and inventory commentary (days–weeks), while the buildout cadence and permitting/regulatory noise determine multi‑quarter realization (months–years). A macro capex pullback or an unexpected GPU supply surge that forces price resets would reverse the positive setup quickly; conversely, a sustained uptick in latency-sensitive AI services would amplify demand for NVDA/SMCI over the next 12–24 months.

Contrarian read: the market underprices the coupling between European network upgrades and incremental GPU server demand — this is not linear server growth but step changes when new POPs go live. Positioning for concentrated multi‑GPU deployments (not just “more servers”) offers asymmetric upside versus being long broad IT hardware or cyclical ad/platform names.

Key Decisions for Investors

• Long SMCI (12–18 months): buy a 12–18 month call spread sized at 2–3% portfolio — strike width tailored to cost limit — targeting 2.0–3.5x payoff if multi‑GPU rack demand accelerates; max loss = premium (~100% of premium), take-profit at 50–75% of notional gain, stop if company reports inventory build >2x QoQ.
• Long NVDA (9–24 months): purchase longer‑dated calls or a call calendar to capture continued AI-driven GPU demand while limiting theta — size 1–2% portfolio. Risk/reward: target 1.5–3x if data‑center orders reaccelerate; hedge with small out‑of‑the‑money puts if guidance on margins weakens.
• Pair trade (3–12 months): long SMCI equity (or calls) vs short APP equity (or buy puts) to express hardware vs ad‑monetization divergence — target spread widening of 20–40% relative; size net exposure small (1–2% net) with stop if spread compresses >15% within 30 days.
• Risk control & alerts: set event alerts for EU regulatory moves on permitting, NVDA earnings/guidance, SMCI inventory commentary, and two consecutive months of ad‑revenue downside for APP. Trim positions by 25–50% on any quarter‑over‑quarter guidance rollback or inventory destocking signal.