UBS turns bullish on five AI data center stocks

UBS initiated coverage of five AI data-center colocation companies—Hut 8, TeraWulf, Applied Digital, Core Scientific and Cipher Digital—with Buy ratings. The bank cited constrained AI compute capacity and the companies' favorable grid and power access, inherited from their cryptomining operations, as advantages supporting rapid growth as AI adoption expands.

Analysis

The investable distinction is not generic AI exposure but contracted, financeable megawatts. Among the group, CORZ and APLD appear best positioned to convert power interconnection into visible infrastructure revenue; WULF, CIFR and HUT retain greater execution sensitivity to build schedules, customer concentration and the cost of funding large-scale GPU-ready capacity. The likely second-order beneficiaries are electrical-equipment vendors such as VRT, ETN and GEV, but their diversified revenue bases make them lower-beta expressions of the same buildout.

Near term, a new sell-side coverage cycle can compress the sector's unusually wide valuation dispersion, particularly in names with limited institutional ownership. Over 1-3 months, the decisive catalyst is not further AI rhetoric but disclosures on signed leases, tenant credit quality, project-level financing and energized-MW milestones. A single delayed interconnect, construction overrun, or equity raise would matter more than an incremental hyperscaler demand forecast because these businesses remain capital-intensive and balance-sheet constrained.

Consensus may be underpricing the option value of scarce power while overpricing the assumption that all legacy mining sites can be converted economically. AI tenants require redundancy, cooling, network connectivity and delivery certainty—not simply low-cost electricity—so the market should ultimately separate contracted data-center operators from speculative land-and-power stories. The structural upside is 6-18 months, but the path will be volatile as these equities retain residual Bitcoin-beta and refinancing risk.

Key Decisions for Investors

• Prefer long CORZ / short CIFR as a 3-6 month relative-value trade: CORZ offers the cleaner route to contracted AI/HPC cash flows, while CIFR has greater dependence on proving conversion economics. Target 15-25% spread widening; exit if CIFR announces a fully financed, creditworthy long-term HPC lease or CORZ misses an energized-capacity milestone.
• Accumulate APLD only on financing-related pullbacks rather than chase coverage-driven strength; use a 6-12 month horizon and cap sizing given dilution risk. Underwrite only if project financing is non-recourse or lease-backed; a material equity issuance before contracted capacity ramps falsifies the risk/reward.
• Treat WULF and HUT as high-beta watch-list names, not core AI infrastructure longs, until customer contracts disclose duration, pricing and power delivery dates. A 10-15% post-rally pullback with confirmed financing would improve entry; avoid if their valuation becomes primarily a Bitcoin proxy.
• For a lower-volatility AI-power expression, remain long VRT or ETN against a basket short of uncontracted mining-to-HPC converters over 6 months. The pair benefits if data-center capex persists but conversion projects face delays; reassess if order growth at electrical suppliers decelerates materially or converter firms secure financed leases.