Baron Opportunity Fund Q2 2026 Top Contributors And Detractors

Nvidia reported April-quarter data-center revenue of $75.2 billion, up 92% year over year, and total revenue of $81.6 billion, up 85%; it guided July-quarter revenue to increase by at least $9.5 billion sequentially. The article also highlights SpaceX's execution of more than 170 launches in 2025, including 165 Falcon 9 missions, with boosters achieving up to 35 re-flights, supporting a high-conviction view on AI and space-technology secular growth.

Analysis

NVDA’s setup is no longer principally a semiconductor-cycle trade; the key variable is whether hyperscaler capex remains broad enough to absorb successive platform transitions without a digestion quarter. Near-term upside should accrue disproportionately to the AI infrastructure bottlenecks—TSM (advanced packaging), VRT (data-center thermal/power), ETN (electrical distribution), and ANET (networking)—where revenue recognition can lag GPU orders by one to three quarters. The risk is that NVDA’s exceptional earnings conversion has already pulled forward estimates: even a beat-and-raise can produce multiple compression if forward growth decelerates faster than investors expect.

The launch cadence demonstrates a cost and reliability moat that widens the gap between SpaceX and listed launch competitors, but its more investable implication is downstream. Cheaper, more frequent access to orbit lowers the capital hurdle for satellite connectivity and Earth-observation constellations, supporting demand for space-qualified components while intensifying pricing pressure on standalone launch and legacy geostationary communications models. RKLB should be viewed as a differentiated execution option rather than a direct beneficiary: rising industry launch activity validates demand, but SpaceX’s scale limits RKLB’s ability to earn scarcity pricing on launch services.

Consensus is likely underestimating the power-grid constraint on AI deployment. GPU procurement announcements do not translate one-for-one into deployed compute when utility interconnects, switchgear, transformers, and cooling systems remain gating items; this can defer NVDA system revenue while extending the earnings runway for VRT and ETN. Over 6-18 months, the more durable AI beneficiaries may therefore be physical-infrastructure suppliers with lower valuation sensitivity than NVDA, provided data-center construction pipelines remain intact.

Key Decisions for Investors

• Maintain NVDA core exposure but avoid adding aggressively into near-term earnings strength; use a 1-3 month pullback or evidence of sustained hyperscaler capex guidance to add. Thesis is falsified by a material reduction in next-quarter revenue guide or signs that supply availability is no longer the binding constraint.
• Initiate a 6-12 month basket long VRT and ETN versus a partial NVDA hedge: data-center power and cooling spend should be recognized later in the deployment cycle and offers a second-order AI exposure. Target 2:1 upside/downside; exit if hyperscaler capex guidance weakens broadly or backlog conversion slips for two consecutive quarters.
• Prefer TSM over a pure NVDA add for 6-12 months, as leading-edge fabrication and advanced packaging capture multi-customer AI demand rather than single-vendor GPU concentration. Key risk is a foundry-capacity normalization or a sharp slowdown in non-AI premium-node demand.
• Do not treat SPCX as a tradable public-equity catalyst. For liquid space exposure, keep RKLB on a watchlist rather than initiate solely on industry activity; require improving launch-service gross margin and contracted backlog growth before buying, as SpaceX’s scale remains the dominant pricing risk.