Nvidia's $63 Billion Investment Portfolio Has a New No. 2 Holding

Nvidia disclosed its latest 13F showing a major new No. 2 public investment in SpaceX: 122,764,805 shares worth nearly $21B at end-June, alongside a roughly $10B investment in AI start-up xAI in January 2026 (from a $20B round). However, SpaceX is emerging as a risk: despite 92% higher Q2 sales, it reported a large loss and steep data-center capex, with AI capex rising from $3.32B (first six months 2025) to $23.55B (first six months 2026). The article flags potential dilution/overhang from accelerated, staggered lockups that could increase float post-IPO, making the SpaceX outlook “murky” for Nvidia’s new holding.

Analysis

The market is overfocusing on the novelty of the cross-holding and underfocusing on the capital-structure math. A newly public name with staggered float expansion and ongoing capex needs usually transitions from story-stock to financing-stock; that is negative for SPCX first, but also for adjacent AI/infra comps whose valuation depends on an uninterrupted primary-market window. In that regime, the highest-quality balance sheets are the relative winners: NVDA’s operating earnings matter far more than the mark-to-market of its venture book, so the headline link is more sentiment noise than fundamental drag.

Near term, the trade is about flow, not fundamentals. Unlocks, passive index inclusion, and any follow-on financing can pressure shares for weeks even if management can still point to revenue growth; the risk is that public investors start demanding proof of self-funding, not just ARR or sales acceleration. Over 1-3 months, a dilutive round or down-round would likely spill into the broader late-stage AI complex, especially names with balance-sheet dependence or high implied growth like CRWV and NBIS.

Contrarian view: the consensus is likely underestimating how fast the market punishes “AI optionality” when it comes with visible cash burn and insider supply. If the company can fund capex without repeated equity issuance, the bear case weakens quickly; that is the main falsifier. Absent that, the structural effect over 6-18 months is a valuation reset for speculative AI infrastructure, while NVDA remains comparatively insulated as a cash-generating core beneficiary rather than a financing-dependent story.

Key Decisions for Investors

• Short SPCX into unlock windows / any post-IPO strength over the next 1-3 months; target 15-25% downside into the first clear financing overhang. Cover if management secures non-dilutive capital or the stock reclaims IPO highs on expanding volume.
• Pair trade: long NVDA / short SPCX for 3-6 months. NVDA keeps the AI compute upside with far less balance-sheet or funding risk; SPCX carries the capital-structure penalty. Best if SPCX borrow is manageable.
• Use weakness in CRWV or NBIS to initiate a short basket against NVDA if the market starts repricing late-stage AI funding risk. This is a cleaner expression of ‘AI capex without earnings’ than a single-name directional bet.
• If listed options are liquid, buy 2-4 month SPCX put spreads rather than outright puts to define risk into unlock/funding catalysts. Avoid naked premium if implied vol is already elevated.
• Set an alert on any announced follow-on, convertible issuance, or secondary sale from insiders; that is the event most likely to convert a headline-driven selloff into a sustained multi-week trend.