Forget CoreWeave. 3 Profitable AI Stocks That Don't Burn $30B a Year

CoreWeave plans up to $35 billion of spend in 2026 versus a $69 billion market cap, carries $29 billion of liabilities, reported Q4 interest expense of $388M (+160% YoY) and guided Q1 interest expense to $510–$590M, highlighting leverage and liquidity risk. The author recommends reallocating to MasTec (Q4 revenue $3.94B; FY 2025 revenue $14.3B; backlog $19B; guidance ~19% revenue growth to ~$17B and adjusted EBITDA $1.45B), Vertiv (FY 2025 revenue $10.2B, 26% organic growth, adj. EPS ~$4.2, $15B backlog, ~41x forward P/E) and nVent (data center sales ~$1B in 2025, backlog $2.3B, ~27x P/E). Implication: avoid highly leveraged, capex-heavy AI-focused data-center builders in a potential slowdown and favor diversified infrastructure operators with stronger earnings and backlog.

Analysis

The market is bifurcating between capital‑intensive, high‑duration buildouts and diversified infrastructure providers that monetize retrofit, replacement, and broad energy needs. That creates a durable advantage for firms with multi‑segment revenue streams and balance‑sheet optionality: they convert backlog into cash more predictably and can redeploy crews/equipment across end markets when hyperscaler demand lags.

Key tail risks sit outside product demand — financing and utilization. Firms that rely on continual capital raises or covenant relief are exposed to a 6–18 month cliff if rates stay elevated or if hyperscaler guidance turns conservative; simultaneously, a meaningful drop in rack/GPU utilization would quickly compress margins for pure builders while leaving retrofit suppliers (power, cooling, electrical protection) with sticky aftermarket demand.

Actionable trades should express balance‑sheet differentiation and optionality rather than pure exposure to AI capex. Use pairs and option structures to isolate operating/financing risk from secular demand: overweight diversified infrastructure contractors and OEMs that capture both new build and aftermarket wallet share, hedge macro rate exposure with limited‑risk downside protection on high‑beta builders, and take a longer time horizon (9–18 months) for conviction to materialize.

Contrarian framing: consensus is overpaying for perpetual growth optionality in pure data‑center builders while under‑rewarding companies that win through penetration of the physical infrastructure stack. If capex growth decelerates modestly, expect multiple compression to hit levered builders far harder than it does suppliers with diverse end markets — giving asymmetric upside in the latter and convex downside in the former.

Key Decisions for Investors

• Pair trade (6–12 months): Long MTZ (size 2–4% portfolio) vs short CRWV via 1:1 long‑dated put spread on CRWV (buy 12–18 month puts, sell lower strike puts to finance). Target = +30–40% on MTZ / payoff asymmetric on CRWV; risk = capped to premium paid on spread (~3:1 reward:risk if MTZ outperforms materially).
• Long VRT (3–9 months): Buy shares size 1.5–3% and sell 6–9 month OTM calls to fund basis (buy‑write). Thesis realized if organic orders sustain; target +20–30%, cut to stop −12% on a guidance miss or order deterioration.
• Long NVT as lower‑beta exposure (12–18 months): Accumulate on pullbacks (stagger buys over 6 weeks) size 1–2%. Target +25% if infrastructure retrofit cycle continues; use a protective 9–12 month put (cost ~1–2% of position) to limit downside from macro shock.
• Tactical bearish option (9–18 months) on CRWV: Buy long‑dated put spread (cap risk) sized small (0.5–1% portfolio). This expresses funding/covenant risk without unlimited short‑stock exposure — aim for 4x+ payoff if market re‑rates or liquidity tightens, losing only premium if AI demand stays unabated.