type: pattern tags: [ai-infrastructure, capex, depreciation, eps-trough, pe-trap, valuation, leverage, capacity-ramp] confidence: medium created: 2026-03-25 source: DOCN stock-analysis 2026-03 persona: bert provenance: legacy source_analysis_path: null source_paragraph_quote: null source_transcript_span: null source_loss_log_path: null

AI Infrastructure Capacity Expansion Creates an EPS Trough That Traps P/E Investors

When AI infrastructure companies deploy large tranches of new capacity (GPUs, data centers), EPS can fall 50%+ in the ramp year even as revenue is accelerating. The mechanism: depreciation on new GPU/infrastructure assets begins immediately upon deployment while revenue from that capacity ramps over 2-4 quarters; simultaneously, debt service on the financing used to fund expansion hits the income statement from day one. The result is a trough year where P/E multiples look alarming but EV/EBITDA on forward estimates and the revenue trajectory tell a much more constructive story.

Evidence

• DOCN FY26: Non-GAAP EPS guided to 0.75−1.00 vs FY25 actual $2.12 — a >50% decline — despite revenue accelerating from 15.5% to ~21% growth and Rule of 40 expanding from 34 to 36-38
• Mechanism: (1) depreciation on 31 MW of new GPU capacity hitting P&L before revenue ramps; (2) interest expense on $606M 2030 convertible notes + $376M credit facility drawn to finance the buildout; (3) new data center operating expenses 2-3 quarters ahead of revenue
• At the same time, forward EV/EBITDA on FY27 estimates (when capacity is fully ramped): 16.1× for a 30% grower — clearly attractive
• Management language: "FY26 is a trough year for margins" — explicitly flagging the transient nature

Implication

For AI infrastructure companies building capacity ahead of demand, do not use P/E as a primary valuation screen during the expansion year — it will systematically flag the stock as expensive precisely when the investment opportunity is best. The correct framework: (1) EV/EBITDA on the out-year (when capacity is fully ramped and billing); (2) EV/S vs. the post-ramp growth cohort; (3) assess whether the EPS trough is timing-driven (depreciation + interest before revenue) or structural (pricing deterioration or utilization problems). A >50% EPS decline accompanied by accelerating revenue and expanding EBITDA margins is a signal to buy, not avoid.