Wolfe Research reiterates Nvidia stock rating on datacenter upside

Wolfe Research reiterated an Outperform on Nvidia with a $275 price target and cites visibility into ~$1 trillion of Blackwell/Blackwell Ultra and Rubin revenue through CY2027. Wolfe's base and bull-case EPS estimates are ~$12.50 and $14 (FY2028), well above consensus EPS of $8.23 for FY2027, and the firm sees ~14-17% upside to datacenter estimates through FY2028, noting each 10% attach of new products could add ~$50B in datacenter revenue. Other brokers were bullish: Raymond James raised its target to $323, Truist to $287, Argus maintained $220, and William Blair reiterated Outperform; Nvidia also announced a collaboration with Qnity Electronics to accelerate AI-driven materials research.

Analysis

The immediate market implication is not just a demand story for one SKU set but a rising ecosystem consumption pattern: higher GPU attach rates mechanically lift demand for HBM, advanced nodes, and system-level components (PSUs, cooling, NICs), concentrating margin capture upstream at foundries and memory suppliers while increasing bargaining power for the GPU leader over OEMs. That supply-chain pull can create capacity-driven pricing power for a cycle, but it also makes growth lumpy — wafer/backlog constraints or a single-supplier bottleneck will translate quickly into forward guidance misses for customers and OEMs.

Key risks live on two horizons. Over the next 1–3 quarters, the dominant risks are inventory cycles and cloud capex pacing — large customers can transiently pull or accelerate orders, creating sharp re-rating events. Over 12–36 months, structural risks include export control escalation, a meaningful improvement in model compute efficiency (algorithmic or sparsity gains), or credible alternative architectures that reduce unit compute intensity, any of which would compress long-term TAM assumptions and valuation multiples.

From a positioning standpoint, the cheapest way to express conviction is to play the multi-node winners rather than a single-name binary: capture upside from wafer and HBM tightness while hedging model/market-share risk with selective short or hedge positions. Conversely, protection against geopolitical or inventory shock is inexpensive via short-dated downside options ahead of known catalysts (earnings, trade-policy windows). The consensus view under-weights the operational friction of scaling exascale systems — software, thermal, and power infrastructure upgrades will create both multi-year revenue tails for some suppliers and near-term margin leakage for customers funding rapid capacity buildouts.

Key Decisions for Investors

• Overweight NVDA (NVDA) via a funded LEAP call spread: buy 12–18 month LEAP calls and sell nearer-term calls to reduce cost; target entering on any >15% pullback. R/R: asymmetric upside if adoption continues; downside limited to premium paid (use sold calls to finance).
• Play supply-chain beneficiaries with a 9–18 month horizon: long TSMC (TSM) and Micron (MU) to capture node/HBM tightness and ASP tailwinds. R/R: captures structural fab/memory tightness but exposed to capex normalization and wafer-cycle downside.
• Pair trade to isolate share/risk: long NVDA / short AMD (AMD) in roughly market-value neutral size for 6–12 months to express system-level moat vs competitive unit-share gains. R/R: benefits if software and system co-design maintain pricing power; risk if AMD wins share in datacenter designs.
• Hedge event/geopolitical risk: buy 30–90 day NVDA puts sized to cover 25–40% of equity exposure around earnings or trade-policy windows. R/R: small premium protects against headline-driven 20–40% drawdowns at peak risk moments.